Beyond the Diatom Test: Emerging Diatom-Based Approaches and Medico-Legal Validation in Forensic Drowning Investigation—A PRISMA-ScR Scoping Review

Abstract

Forensic diagnosis of drowning remains challenging because postmortem findings are often supportive rather than pathognomonic. The classical diatom test has long been used as an ancillary tool, but its interpretation remains controversial because of contamination, false positives, environmental ubiquity, and heterogeneous analytical workflows. This PRISMA-ScR scoping review mapped emerging diatom-based and adjacent aquatic biological approaches proposed beyond the classical test, with emphasis on analytical innovation, forensic applicability, and medico-legal validation. Scopus, PubMed, and Web of Science Core Collection were searched for literature published between 2014 and 2026. After deduplication and screening, 165 studies were included. The literature covered targeted molecular and environmental-biological approaches, drowning-site inference, automated image analysis and scanning electron microscopy (SEM)/automated scanning electron microscopy (Auto-SEM) workflows, quantitative and tissue-to-medium comparisons, alternative substrates, contamination assessment, and medico-legal interpretation. Most evidence consisted of controlled experimental or comparative validation studies and case-based forensic applications; no study was classified as multicenter, externally validated, or operationally mature Level 5 evidence. Emerging approaches may strengthen drowning investigations by improving detection, comparison, quantification, and workflow reproducibility. However, they should currently be regarded as adjunctive tools rather than standalone proof of drowning or precise site attribution. Harmonized protocols, standardized environmental reference sampling, explicit contamination controls, external validation, and fit-for-purpose medico-legal interpretation remain central priorities.

1. Introduction

Drowning remains a difficult postmortem diagnosis because autopsy findings are usually supportive rather than pathognomonic. Drowning is also a major public-health concern, with an estimated 300,000 fatal drowning events worldwide in 2021 [1], and a recent systematic review has reaffirmed that no single autopsy finding is pathognomonic for drowning, so the diagnosis remains one of integration and exclusion [2]. Independent Global Burden of Disease estimates for the same year report 274,230 fatal drowning events and a 64% reduction in the age-standardized mortality rate since 1990, with persisting geographic and demographic heterogeneity [3]. Drowning remains the third leading cause of death among children aged 5–14 years globally, and 92% of fatal drowning events occur in low- and middle-income countries, where mortality rates are 3.2 times higher than in high-income settings. A further dimension particularly relevant to medico-legal practice is the documented but largely uncounted burden of fatal drowning among migrants, with more than 39,000 deaths reported during unsafe migration journeys since 2014 [1]. In bodies recovered from water, the forensic question often extends beyond the immediate cause of death to the timing of submersion, the compatibility between the recovery site and the site of drowning, and the influence of postmortem or environmental factors. Interpretation therefore depends on the convergence of scene information, autopsy and histology, toxicology, environmental data, and ancillary laboratory tests [2,4,5,6]. Among these ancillary approaches, diatom analysis has remained prominent because it links biological traces recovered from the body with organisms present in the aquatic environment [4,5,6,7].

The classical diatom test is based on the premise that, during active drowning, water containing diatoms or other planktonic material may be aspirated, enter the lungs, and in some circumstances disseminate to distant tissues. Diatoms are attractive forensic traces because their siliceous frustules can survive laboratory processing and may be compared with assemblages in the suspected drowning medium [4,5,6,7]. However, the test is also controversial: diatoms are environmentally widespread, may be introduced by contamination, and may be detected in low numbers or in contexts that do not establish drowning. Sampling strategies, tissue masses, digestion and filtration protocols, light microscopy or scanning electron microscopy (SEM) workflows, taxonomic expertise, and interpretive thresholds all influence evidentiary value [6,8,9,10].

Recent work has expanded the field beyond conventional microscopy-based presence/absence testing. Molecular and environmental DNA (eDNA) approaches have been explored to improve detection and environmental comparison [11,12,13,14,15,16]; image-analysis, deep-learning, artificial intelligence (AI)-assisted, You Only Look Once (YOLO)- or DiatomNet-type systems, and SEM/automated SEM (Auto-SEM) workflows aim to reduce observer dependence [17,18,19,20,21,22,23]; and quantitative strategies, including tissue-to-medium comparisons and lung-to-drowning-medium ratios (L/D ratio), seek to make interpretation more reproducible [24,25,26,27,28]. These developments differ in biological targets and forensic purposes. Some remain close to classical forensic diatomology, whereas others use diatom DNA, mixed planktonic signatures, microbial profiles, automated imaging, or standardized quantitative comparisons.

This diversity is where the phycological dimension becomes central, but also where terminology requires precision. Diatoms are not generic particles: they are algae with siliceous frustules that can be preserved through many laboratory procedures, and their morphology, taxonomy, and ecological distribution vary with the water body, season, hydrology, substrate, salinity, and local environmental conditions. When diatoms and other microalgae are translated into forensic evidence, algal morphology, ecology, molecular signatures, environmental reference sampling, and medico-legal reasoning become part of the same interpretive pathway.

Existing reviews have clarified the history, diagnostic controversy, and medico-legal interpretation of the classical diatom test [4,5,6,7,8,9,10]. The present review addresses a different question: how the broader methodological landscape emerging around forensic phycology and related aquatic biological evidence can be organized by forensic target, methodological family, and validation maturity. In this review, forensic phycology refers primarily to diatoms, algae, and other microalgae, whereas bacterioplankton, mixed microbial profiles, eDNA, metagenomic, and metabolomic approaches are treated as related aquatic or postmortem biological evidence. Its focus is not to re-evaluate the classical test alone, but to map newer diatom-based and adjacent aquatic biological approaches in relation to medico-legal readiness.

A scoping-review design was used to organize this heterogeneous literature rather than to estimate pooled diagnostic accuracy [29]. This review aimed to map emerging approaches proposed beyond the classical diatom test in forensic drowning investigations, with particular attention to forensic purpose, analytical workflow, validation maturity, and medico-legal interpretability.

2. Materials and Methods

2.1. Review Design

This scoping review mapped the extent, range, and characteristics of the literature on emerging diatom-based, algal/microalgal, molecular, metagenomic, automated, AI-assisted, quantitative, standardized, and related aquatic biological approaches in forensic drowning investigation. The review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) [29]. No protocol was registered; however, the review questions, eligibility criteria, and charting framework were defined before the full-text eligibility assessment.

The purpose of the review was not to estimate pooled diagnostic accuracy or to perform meta-analysis. Instead, it was designed to organize a heterogeneous body of forensic, phycological, molecular, computational, and medico-legal literature into categories relevant to drowning investigations and to identify the main methodological and validation gaps.

2.2. Review Questions

The review was guided by the following questions:

• Which emerging diatom-based, molecular, metagenomic, automated, AI-assisted, quantitative, and standardized approaches have been proposed beyond the classical diatom test?
• Which forensic functions do these approaches address, including drowning diagnosis, drowning-medium comparison, site inference, detection/classification, standardization, contamination control, and medico-legal interpretation?
• Which methodological, interpretive, and validation gaps limit operational or court-oriented use?

2.3. Information Sources and Search Strategy

The final search used three bibliographic sources: Scopus, PubMed, and Web of Science Core Collection. Records published from 1 January 2014 to 15 May 2026 were considered. English-language filtering was applied where available. Article and review filters were applied where available; PubMed article-type filtering was not applied at the export stage and publication types were assessed during screening. The search combined drowning/submersion terms with diatom, planktonic, microalgal, molecular, eDNA/metagenomic, automation, SEM/Auto-SEM, quantitative, validation, contamination, and medico-legal interpretation concepts.

All complete database-specific search strings, restrictions, search dates, and exported record numbers are reported in Supplementary Table S1; these operational details are not repeated in the main text.

All final database exports were imported into a master screening dataset before deduplication.

2.4. Eligibility Criteria

Records were eligible when they addressed forensic drowning investigations or postmortem diatom/plankton analysis and examined at least one method, marker, interpretation problem, or validation issue relevant to the review objective. Eligible evidence sources included original research articles, experimental studies, methodological studies, validation studies, case reports or case series, computational or AI studies, database or mapping studies, reviews, systematic reviews, and other high-relevance interpretive papers.

Records were included when they addressed one or more of the following: diatom test or postmortem diatom analysis; diatom-based, plankton-based, microalgal, molecular, DNA-based, polymerase chain reaction (PCR)/quantitative polymerase chain reaction (qPCR), 18S ribosomal ribonucleic acid/deoxyribonucleic acid (18S rRNA/rDNA) or ribulose-1,5-bisphosphate carboxylase/oxygenase large-subunit gene (rbcL), microarray, sequencing, metabarcoding, or metagenomic methods relevant to drowning; drowning-site inference or drowning-medium comparison; automated, AI-assisted, deep learning, machine learning, image-analysis, Auto-SEM, or classification methods applied to diatoms or drowning evidence; quantitative or standardized indicators such as the L/D ratio, tissue-to-medium comparisons, standardized sampling, digestion/filtration workflows, or method validation; and medico-legal interpretation, contamination, false positives, reliability, reproducibility, or evidentiary validation of diatom, plankton, or microalgal evidence.

Records were excluded when they were clearly unrelated to forensic drowning or aquatic forensic evidence, including non-forensic algal ecology, paleontology, paleoecology, geology, carbonate-platform drowning, stratigraphy, harmful algal bloom ecology or toxicology without forensic drowning relevance, generic drowning pathology without a relevant diatom/plankton/microalgal/molecular/automated component, generic forensic AI unrelated to drowning evidence, and publication types without sufficient methodological content. Full texts were also excluded when they were the wrong records, duplicate publications, outside the time frame, or not usable for extraction.

2.5. Selection Process

Database exports were merged and deduplicated using identifiers and bibliographic metadata, including DOI, PMID, database identifiers, normalized titles, authors, year, and journal. Title/abstract screening followed conservative scoping-review logic: records were included if clearly relevant, excluded only if clearly irrelevant, and marked as uncertain when relevance could not be determined from the title and abstract alone.

Records classified as Include or Uncertain at title/abstract screening were taken forward for full-text retrieval. Full texts were retrieved through reference-management software, institutional or legal access routes, manual searching, publisher pages, repository pages, author-uploaded sources, and document-delivery attempts when necessary. Reports that could not be retrieved after these attempts were classified only as reports sought for retrieval but not retrieved and were not assessed for content eligibility on the basis of the title or abstract alone.

Retrieved full texts were assessed against the eligibility criteria, and reasons for full-text exclusion were recorded using standardized categories. Title/abstract screening, full-text eligibility assessment, initial data charting, and initial validation-level assignment were performed by one reviewer. Uncertain records, borderline eligibility decisions, thematic-domain assignments, and validation-level classifications were subsequently checked by the other authors with relevant forensic, phycological, molecular, and medico-legal expertise. Disagreements or interpretive uncertainties were resolved by discussion and consensus.

2.6. Data Charting

Data were charted from each included study using a structured extraction form. The charted fields included record identifier, title, authors, year, journal, country or setting, dominant evidence descriptor, sample type, organism or marker, analytical method, forensic target, main findings, reported advantages, reported limitations, validation level, thematic domain, relevance to the review, source file, and notes for synthesis. High-relevance reviews and systematic reviews were retained when they informed methodological context, interpretive frameworks, validation concerns, or reference tracking; their conclusions were not treated as independent original diagnostic-performance evidence.

Validation levels were categorized as follows: 1 = conceptual or narrative; 2 = proof-of-concept; 3 = experimental controlled validation; 4 = case-based forensic application; and 5 = multicenter, externally validated, or operationally mature. This validation-level classification was used as a pragmatic descriptive maturity-mapping framework developed for this scoping review to summarize apparent methodological maturity. It was not intended as a formal risk-of-bias tool, a validated certainty-of-evidence scale, or a pooled diagnostic-accuracy assessment.

2.7. Synthesis of Results

The results were synthesized descriptively and thematically. Because the included literature was methodologically heterogeneous, no pooled quantitative synthesis was attempted. Five descriptive layers were kept distinct: thematic domains described substantive contributions; methodological families described technical platforms or workflows; dominant evidence descriptors summarized the primary contribution of each record for descriptive counting; validation levels described apparent methodological maturity; and relevance ratings described direct relevance to the review questions.

Thematic-domain and methodological-family assignments were non-mutually exclusive because individual studies could contribute simultaneously to method development, quantitative standardization, and medico-legal interpretation. Dominant evidence descriptors were used only to summarize the primary contribution of each record; they should not be read as mutually exclusive analytical categories. Summary tables were prepared to describe thematic distribution, methodological families, forensic applications, validation levels, and recurrent limitations.

3. Results

3.1. Study Selection

The database searches identified 487 records after the application of database filters. After the removal of 268 duplicate records, 219 unique records were screened at the title/abstract level. Thirty-four records were excluded at the title/abstract screening, leaving 185 reports sought for full-text retrieval. Of these, 18 reports could not be retrieved after EndNote-assisted retrieval, manual searching, institutional or library attempts, and legal web searching. These 18 reports were not assessed for content eligibility.

A total of 167 reports were retrieved and assessed for full-text eligibility. Two full texts were excluded: one generic forensic AI review unrelated to diatoms or drowning evidence, and one general drowning-pathology discussion without a specific diatom, plankton, microalgal, molecular, or automated analytical focus. Therefore, 165 studies were included in this scoping review. The study-selection process is summarized in Figure 1.

Figure 1. PRISMA-ScR flow diagram. Flow diagram summarizing record identification, deduplication, title/abstract screening, full-text retrieval, full-text eligibility assessment, and final inclusion in the scoping review. Reports not retrieved were counted as retrieval failures and were not assessed for full-text eligibility.

3.2. General Characteristics of Included Studies

The 165 included studies were published between 2014 and 2026. The annual distribution was uneven, with the largest number of included studies charted in 2022 (n = 28), followed by 2025 (n = 17), 2021 (n = 16), 2023 (n = 14), 2024 (n = 13), 2018 (n = 12), and 2019 (n = 12). The remaining years contributed between 7 and 11 studies each.

The charted country or setting most frequently recorded was China (n = 79), followed by Japan (n = 22), India (n = 20), Italy (n = 14), and the United Kingdom (n = 9). Smaller numbers of studies were charted from Romania, Australia, Portugal, Hungary, the United States, Sweden, Canada, South Korea, and Poland; five studies had a country or setting that was not clearly reported.

Using a dominant evidence descriptor assigned to each record for descriptive counting, the included corpus was led by experimental animal studies (n = 76) and molecular assay or metagenomic studies (n = 44). AI or computational studies accounted for 22 records, case reports or case series for 11, reviews for 6, systematic reviews for 3, methodological or validation studies for 2, and database or mapping studies for 1. These descriptors summarize the primary contribution of each record; they should not be interpreted as mutually exclusive technical categories, because animal experiments, molecular assays, AI workflows, validation designs, and review articles may overlap analytically or thematically. The complete charting of included studies is reported in Supplementary Table S2.

3.3. Thematic Distribution of Included Studies

The evidence did not form a single methodological stream. Instead, the included studies clustered around four partly overlapping thematic functions: molecular or environmental comparison, automation and classification, quantitative standardization, and medico-legal interpretation. Domain assignment was therefore non-mutually exclusive and should be distinguished from the methodological-family synthesis, which describes technical platforms or workflows rather than thematic contributions. Quantitative or standardized indicators were the most represented area (Domain C; n = 148, 89.7%), followed by medico-legal interpretation, validation, contamination, and reliability (Domain D; n = 133, 80.6%). Molecular, metagenomic, DNA-based, or site-inference approaches (Domain A) were represented by 83 studies (50.3%), and automation, image analysis, and AI-assisted detection or classification (Domain B) by 51 studies (30.9%). No study remained in the unclear/other category after re-audit.

The overlap between domains was itself informative. The most frequent combinations were C; D (n = 40, 24.2%), A; C; D (n = 37, 22.4%), B; C; D (n = 23, 13.9%), and A; B; C; D (n = 19, 11.5%). These combinations show that many studies did not simply introduce a technique; they also addressed reproducibility, comparison, contamination, or interpretation. The thematic distribution is summarized in Table 1.

Table 1. Thematic distribution of included studies.

Domain/Combination	Label	n	%
A	Molecular/metagenomic/DNA-based/site inference	83	50.3
B	Automation/image analysis/AI-assisted detection or classification	51	30.9
C	Quantitative/standardized indicators beyond simple presence/absence	148	89.7
D	Medico-legal interpretation/validation/contamination/reliability	133	80.6
E	Unclear/other	0	0.0
Most frequent domain combinations
C; D	Quantitative/standardized indicators + medico-legal interpretation/validation	40	24.2
A; C; D	Molecular/site-inference + quantitative/standardized + medico-legal interpretation	37	22.4
B; C; D	Automation/image analysis/AI + quantitative/standardized + medico-legal interpretation	23	13.9
A; B; C; D	All four thematic domains	19	11.5

Domain assignment was non-mutually exclusive and thematic; therefore, individual studies could contribute to more than one domain. Domains should not be read as study-design categories or as one-to-one methodological platforms. Only the most frequent domain combinations are shown. The complete domain-combination table is provided in Supplementary Table S3; complete study charting is provided in Supplementary Table S2.

The methodological-family synthesis is set out in Table 2.

Table 2. Methodological families and forensic applications. Summary of the main methodological families identified among included studies, including representative techniques, principal forensic applications, and recurrent caveats or validation gaps. Methodological families were treated as non-mutually exclusive technical/workflow categories, as the same study could contribute to more than one analytical or interpretive category. SEM/Auto-SEM was charted as a methodological family, whereas Domain B captures automation, image analysis, and AI-assisted detection/classification as thematic contributions; the two counts should therefore not be read as one-to-one equivalents.

Methodological Family	n (%)	Representative Techniques	Main Forensic Applications	Main Caveats/Validation Gaps
Morphological microscopy/digestion-filtration workflows	88 (53.3%)	Light microscopy, tissue digestion, filtration, conventional diatom counting	Drowning diagnosis; tissue/drowning-medium comparison; method optimization	Highly dependent on sampling, digestion/filtration workflow, taxonomic expertise, and contamination control.
SEM/Auto-SEM workflows	78 (47.3%)	SEM, Auto-SEM, MD-VF-Auto SEM, vacuum filtration	Higher-resolution detection; automated or semi-automated workflows; standardized counting	Requires expensive, specialized equipment; trained operation, maintenance, and calibration; validated cleaning or single-use digestion/filtration components; procedural blanks, recovery controls, and contamination controls; assessment of digestion-induced frustule fragmentation; harmonized rules for counting intact frustules versus fragments; independent replication, multicenter evidence, inter-laboratory reproducibility, and evaluation of routine-laboratory generalizability.
Contamination/false positives/medico-legal interpretation	50 (30.3%)	Contamination studies, false-positive analysis, review/interpretive papers	Court-oriented interpretation; reliability; evidentiary caution	Findings support adjunctive rather than standalone use until stronger validation standards exist.
Quantitative indicators/standardized comparisons	40 (24.2%)	L/D ratio, tissue-to-medium comparisons, counts, dispersion, standardized sampling	Beyond presence/absence; reproducibility and diagnostic support	Cut-offs and comparability differ across organs, media, digestion protocols, and case contexts.
Site inference/environmental mapping/forensic databases	38 (23%)	Diatomological mapping, waterbody databases, site comparison, distribution studies	Drowning site inference; source attribution; environmental reference sampling	Environmental heterogeneity and temporal variation limit portability across regions and seasons.
Molecular PCR/qPCR/microarray assays	36 (21.8%)	PCR, qPCR, multiplex PCR-capillary electrophoresis, microarray, 18S/rbcL markers	Molecular detection of diatoms/plankton and support for drowning diagnosis	Marker selection, inhibition, degradation, and interpretation thresholds require validation.
Automation/image analysis/AI-assisted classification	31 (18.8%)	Deep learning, YOLO, DiatomNet, automated image analysis	Automated detection and classification of diatoms; reduced observer dependence	Dataset annotation, taxonomic uncertainty, external validation, model transparency, and deployment across laboratories remain key barriers.
Sequencing/metabarcoding/metagenomics/microbial profiling	14 (8.5%)	DNA sequencing, metabarcoding, metagenomic profiling, microbial community analysis	Site inference; environmental comparison; differentiation of drowning media	Promising for broad environmental discrimination, but precise geographic attribution remains limited by temporal/spatial heterogeneity and transferability.
Trace transfer/clothing/alternative substrates	10 (6.1%)	Clothing/fabrics, maggots, bone, tooth, vitreous, synovial fluid, closed organs	Alternative matrices; trace evidence; postmortem/submersion context	Often preliminary and matrix-specific; transfer/persistence and diagnostic meaning need further validation.

3.4. Molecular, Microbial, Environmental-Mapping, and Site-Inference Approaches

These studies addressed two related forensic functions: detecting aquatic biological signals in postmortem samples and comparing them with environmental sources. The molecular and environmental-biological component of the corpus was methodologically diverse, ranging from targeted PCR/qPCR and marker-based assays to sequencing, metabarcoding, metagenomic diatom analysis, microbial profiling, diatomological mapping, water-body databases, and environmental comparisons.

Targeted molecular and molecular-adjacent work focused on whether aquatic biological signals could be recovered or interpreted more reproducibly than by morphology alone. This subgroup included DNA-binding, PCR-based, silicon/tissue-distribution, quantitative, veterinary, and morphology-adjacent studies relevant to biological signal recovery and drowning interpretations [30,31,32,33,34]. Related studies addressed organ–medium correlations, the comparison of microwave digestion-vacuum filtration-automated scanning electron microscopy (MD-VF-Auto SEM) with multiplex PCR, digestion variables, diatom DNA extraction, and closed-organ or case-based interpretations [35,36,37,38,39]. Case-based and contamination-oriented studies, including mud aspiration, resuscitation or postmortem changes, and maggot-based evidence, broadened this evidence base [40,41,42,43].

A second cluster used broader environmental-biological signals for comparison with drowning media or potential drowning sites. Bacterioplankton/PCR and microbiota-based studies examined non-diatom microbial indicators relevant to drowning diagnosis or interpretations [44,45,46]. Algal, planktonic, and related animal or case-context studies contributed complementary biological detection evidence [47,48,49,50,51]. Chloroplast, cyanobacterial, multiplex-PCR, and diatom DNA barcode workflows further contributed detection and environmental-comparison evidence [52,53,54,55]. Further studies addressed contamination blind spots, sternal aspirates, diatom DNA extraction from water and tissues, and the comparison of plankton testing methods [56,57,58,59,60].

Site inference and environmental comparisons were also pursued through diatomological mapping, water-body databases, and regional reference resources. These studies covered rivers, lakes, dams, drowning media, and other local aquatic settings [61,62,63,64,65,66,67,68]. Water-body database studies and comparative sampling protocols characterized spatial structure in diatom assemblages [69,70,71,72,73,74,75,76], while regional mapping and site-specific profiles expanded the evidence across additional aquatic contexts [77,78,79,80,81,82,83,84]. Additional regional freshwater applications were also represented [85]. AI-supported database work and water-body characterization studies further developed reference resources for site-oriented interpretations [86,87,88,89,90,91,92,93].

Studies based primarily on bacterial or mixed microbial profiles were treated as adjacent aquatic biological approaches relevant to forensic site inference, rather than as strictly diatom- or microalgal-based evidence. They were retained because they inform environmental comparisons while relying on different biological markers and interpretive assumptions. This adjacent evidence included bacterial composition and random-forest site inferences [94], broader photoautotrophic microbial perspectives [95], object-detection and optical-classification applications connected with detection or site-oriented tasks [96,97], and substrate-colonization evidence [98]. Within this subgroup, recent quantitative work has shown that random-forest classifiers trained on 16S ribosomal ribonucleic acid (16S rRNA) microbial-community data may discriminate antemortem drowning from postmortem immersion with an area under the curve (AUC) of 0.96 and a Day-1 accuracy of 98.22%, declining to 86.11% by Day 7 [60], and that microbial site-attribution models can achieve Day-1 accuracies above 95% within a single species while losing substantial performance when transferred across species (down to approximately 61% in a mouse-to-rabbit application) [99]. Further regional and integrative comparison studies confirmed the importance of local environmental context in site-oriented interpretations. [100,101,102].

3.5. Automation, Image Analysis, and AI-Assisted Detection/Classification

Automation-oriented studies addressed a different bottleneck: the time, expertise, and subjectivity required to locate and classify candidate diatoms. Fifty-one studies mapped to this area, including semi-automated detection, image-analysis workflows, AI-assisted classification, deep learning, YOLO-based systems, DiatomNet, automated search systems, and selected SEM/Auto-SEM or microwave digestion-vacuum filtration-automated SEM (MD-VF-Auto SEM) workflows.

The studies mainly pursued faster screening, reduced observer dependence, more reproducible image-based classifications, and the scalable processing of diatom evidence. In the methodological-family synthesis, automation, image analysis, and AI-assisted classification accounted for 31 studies (18.8%), while SEM or Auto-SEM workflows accounted for 78 studies (47.3%). The difference reflects the purpose of the two classifications. Domain B captures studies in which automation, image analysis, or AI-assisted detection/classification was a substantive thematic contribution, whereas SEM/Auto-SEM was charted as a broader methodological family. Some SEM studies overlapped with Domain B when SEM was part of automated search or standardized detection, whereas other SEM studies used SEM primarily as an imaging or detection platform without being AI- or automation-oriented. Recurrent limitations included dataset annotation, model transparency, external validation, equipment dependence, and deployment across laboratories (Table 2).

The computational subgroup included image-analysis and deep-learning classification studies [103,104,105,106,107,108], together with object-detection, DiatomNet/YOLO-type, and AI-assisted search systems [109,110,111,112,113,114,115,116].

3.6. Quantitative and Standardized Indicators

The largest body of evidence concerned the move from simple presence/absence findings toward reproducible quantification and standardized comparisons. Domain C included 148 studies (89.7%) and brought together work on diatom counts, lung-to-drowning-medium logic, tissue-to-medium comparisons, organ distribution, digestion and filtration workflows, SEM/Auto-SEM-based counting, and standardized analytical procedures.

Morphological microscopy and digestion-filtration workflows were the largest methodological family (n = 88, 53.3%). SEM and Auto-SEM workflows accounted for 78 studies (47.3%), and quantitative indicators or standardized comparisons for 40 studies (24.2%). Together, these studies addressed method optimization, tissue/drowning-medium comparison, organ distribution, higher-resolution detection, and more structured interpretations. However, these counts should not be interpreted as evidence that SEM/Auto-SEM has replaced light microscopy; the latter remains historically broader and more accessible, whereas SEM/Auto-SEM requires equipment-intensive workflows and external validation. Recurring caveats were heterogeneity in sampling, digestion and filtration, taxonomic expertise, equipment requirements, organ-specific differences, environmental background variation, and the limited comparability of thresholds across contexts (Table 2).

Quantitative, tissue-to-medium, organ-distribution, the L/D ratio, and abundance-comparison studies formed the central group of structured indicators. Key examples included the quantitative investigation of diatom dispersion in lung tissue [24], quantitative analysis for drowning diagnosis [25], lung tissue-to-drowning-medium comparison [26], concordance analysis of diatom types and patterns in lung tissue and the drowning medium [27], relationships between diatom abundances in rat organs and environmental waters [28], and L/D ratio validation [112]. Automation, SEM-based, and AI-assisted detection studies intersected with quantitative or standardized workflows when they supported countable, reproducible detection or search efficiency [103,104,105,106,107]. Object-detection, AI-search, and related SEM-associated workflows further developed this automation-oriented strand [108,109,110,111,112]. Methodological studies on digestion, filtration, enzymatic, hypochlorite, hydrogen-peroxide/protease, and ethanol-related processing evaluated recovery and comparability [117,118,119,120,121,122,123], while experimental lung-tissue concentration, diagnostic-ratio, and retrospective quantitative studies added further structured evidence [124,125]. Further case-based or interpretive work addressed individualization, methodological limits, trace linkage, diagnostic algorithms, imaging-assisted assessment, and difficult case interpretations [126,127,128,129,130,131,132,133].

A related set of studies broadened the evidentiary setting beyond routine lung or organ sampling. Context-specific evidence included archeological or historical water-death remains, tissue samples, lung-weight findings, and human case series [134,135,136,137,138,139]. Digestion, filtration, and extraction methods were addressed across conventional and newer workflows [140,141,142,143,144,145]. Water-body mapping, databasing, phytoplankton ecology, site comparison, and forensic diatom-flora studies supported environmental interpretations [146,147,148,149,150,151]. Veterinary or non-human applications and trace transfer to clothing or fabrics were represented in separate studies [152,153]. Alternative or specialized substrates and interpretive settings included experimental lung-injection testing [154], closed organs [155], vitreous humor [156], bone marrow or bone [157,158,159], bone and tooth [160], synovial fluid [161], poisoning-related interpretive contexts [162], and maggots [43]. These records emphasize that transfer, persistence, contamination, decomposition, and processing effects require matrix-specific validation rather than direct extrapolation from lung tissue or drowning-medium analysis.

3.7. Medico-Legal Interpretation, Contamination, and Validation

The interpretive evidence base showed that the main medico-legal issue is not simply whether an aquatic biological signal can be detected, but how that signal should be weighed in context. Domain D included 133 studies (80.6%) addressing contamination, false-positive findings, reliability, reproducibility, forensic interpretations, evidentiary caution, case-based applications, and the diagnostic or medico-legal validation of diatom, plankton, microalgal, molecular, or automated methods.

Contamination, false-positive analysis, and medico-legal interpretations formed a distinct methodological family of 50 studies (30.3%). These studies were mainly relevant to court-oriented interpretations, diagnostic reliability, contamination control, and the limits of evidentiary weight. Across this group, emerging methods were generally framed as supportive or adjunctive rather than as standalone proof of drowning. Recurrent limitations included environmental ubiquity, sampling and laboratory variability, incomplete external validation, limited inter-laboratory comparability, and uncertain operational thresholds.

Across this subset, the recurrent pattern was that analytical positivity required contextual qualification: the studies did not treat detection as self-explanatory, but linked interpretation to contamination control, case circumstances, tissue or matrix selection, and comparison with the drowning environment.

Reviews and studies addressing false-positive results, contamination sources, and medico-legal interpretations provide the main basis for weighing positive diatom findings in context [5,6,8,9,10]. Additional work on contamination blind spots, non-drowning entry pathways, closed-organ interpretations, and laboratory consumables reinforces the need for strict contamination control [56,115,155,163]. Case-based and validation-oriented studies added practical examples of how analytical findings may be interpreted alongside scene, autopsy, environmental, and laboratory information [164,165,166].

3.8. Validation Level and Forensic Applicability

The validation-level synthesis confirmed that the field remains active but not yet operationally mature. Controlled experimental or comparative validation studies dominated the corpus (Level 3; n = 110, 66.7%), followed by case-based forensic applications (Level 4; n = 36, 21.8%), conceptual or narrative studies (Level 1; n = 14, 8.5%), and proof-of-concept studies (Level 2; n = 5, 3.0%). No study was categorized as Level 5, defined as multicenter, externally validated, or operationally mature.

Relevance to the review was categorized as High for 72 studies (43.6%), Moderate for 69 (41.8%), and Low for 24 (14.5%). The absence of Level 5 evidence does not negate the value of experimental or case-based work, but it indicates that the mapped evidence has not yet demonstrated broad multicenter or externally validated operational readiness. Validation levels and medico-legal limitations are summarized in Table 3.

Table 3. Validation level and medico-legal limitations. Classification of included studies according to a pragmatic descriptive maturity-mapping framework and relevance to the review, with associated medico-legal limitations relevant to forensic applicability. The Level 1–5 system was not a formal risk-of-bias or certainty-of-evidence grading tool.

Validation Level/Category	Meaning	n	%	Main Medico-Legal Limitation
1 = conceptual/narrative	Conceptual, review, or narrative synthesis	14	8.5	Useful for identifying interpretive issues and research directions, but not sufficient for operational adoption.
2 = proof-of-concept	Early technical or feasibility work	5	3	Analytical promise but limited case realism, sample size, and external validation.
3 = experimental controlled validation	Controlled experimental or comparative validation	110	66.7	Dominant evidence level; controlled design helps method comparison but may not reproduce real-case heterogeneity.
4 = case-based forensic application	Case report, case series, or applied forensic use	36	21.8	Forensic realism is higher, but evidence is often non-comparative and context-dependent.
5 = multicenter/externally validated/operationally mature	Multicenter/external validation or operational implementation	0	0	Absent in the mapped dataset; this remains the main barrier to court-ready standardization.
Relevance category
High		72	43.6
Moderate		69	41.8
Low		24	14.5

4. Discussion

4.1. Principal Findings

This scoping review shows that recent research on diatom-based evidence in drowning investigations is no longer centered only on the question of whether diatoms can be detected after death. The included literature instead reflects a broader methodological shift: aquatic biological traces are being explored as diagnostic supports, environmental-comparison tools, quantitative indicators, automated image-analysis targets, and medico-legal evidence requiring cautious interpretations.

The classical diatom test remains the conceptual reference point, but most contemporary work attempts to address its weaknesses through molecular or environmental-biological detection, SEM/Auto-SEM workflows, AI-assisted classifications, quantitative tissue-to-medium comparisons, contamination assessment, and validation-oriented designs. The validation profile remains the central limiting factor. Although the corpus was large and methodologically active, no included study was categorized as multicenter, externally validated, or operationally mature Level 5 evidence. This supports a cautious conclusion: emerging approaches may strengthen forensic interpretations, but they are not yet standalone proof of drowning or precise site attribution.

Figure 2 summarizes this integrated fit-for-purpose framework and the main pre-analytical, analytical, and interpretive barriers identified by the scoping map.

Figure 2. Conceptual overview of the transition from classical diatom testing to integrated forensic phycology and related aquatic biological evidence. The scheme illustrates the expansion from conventional microscopy-based diatom testing to a broader framework incorporating molecular/metagenomic approaches, automated or AI-assisted detection, quantitative standardization, medico-legal interpretation, and fit-for-purpose forensic validation. The pre-analytical, analytical, and interpretive layers indicate key conditions influencing evidentiary robustness and forensic applicability. MD-VF-Auto SEM is included as a promising but validation-dependent workflow, not as an established replacement for conventional methods and not as a method that removes contamination or interpretive uncertainty.

4.2. Methodological Advances Beyond the Classical Diatom Test

4.2.1. Molecular, eDNA, Metagenomic, and Microbial Approaches

Molecular, eDNA, metabarcoding, metagenomic, and microbial-profiling approaches are being explored because they address some limitations inherent in morphology-based diatom analyses while introducing different validation requirements. Classical diatom testing depends on the recovery of identifiable frustules, operator expertise, digestion or filtration conditions, the organ sampling strategy, and comparison with environmental material. DNA-based methods offer a complementary route: rather than relying exclusively on morphologically intact diatom structures, they may detect diatom, planktonic, algal, or microbial signals using genetic targets. These methods differ substantially in target, resolution, and validation needs, ranging from targeted PCR/qPCR or marker-based assays to broader sequencing, metabarcoding, metagenomic, and microbial-profiling approaches (Table 2). Examples include PCR-capillary electrophoresis for forensic diatom testing [11], 18S rRNA gene arrays for the forensic detection of diatoms [12], qPCR-based or real-time qPCR approaches [15,16], and rbcL-based pyrosequencing for drowning-site inferences [13].

The potential contribution of these methods lies in their ability to support analytical sensitivity, target specificity, and comparisons between biological samples and drowning media. Molecular assays may be useful when diatoms are scarce, degraded, or difficult to classify morphologically. They may also allow the detection of planktonic or microbial targets that are not limited to classical diatom morphology. This is reflected in studies that examined bacterioplankton or microbial indicators as diagnostic supports [44,46], as well as studies that used microbial community profiling and random forest approaches for site inferences [94,101]. These approaches broaden the evidentiary landscape beyond strict forensic phycology and should be described as related aquatic biological evidence when they rely primarily on bacterial, mixed microbial, or other non-algal signals.

However, the charted evidence also indicates that the forensic interpretation of molecular findings is not straightforward. Detection of a molecular target is not equivalent to proof of drowning unless the sampling context, contamination controls, the postmortem interval, reference-water sampling, and the background distribution of the target are carefully addressed. A positive molecular signal does not necessarily demonstrate intact diatom frustules, and it cannot be directly equated with antemortem drowning. Marker selection is a central issue. Targets such as 18S rRNA/rDNA and rbcL can be informative for diatoms, but they differ in taxonomic resolution, amplification behavior, reference-database availability, and susceptibility to degradation or inhibition. In addition, molecular workflows may be affected by DNA degradation in forensic samples, extraction difficulty, amplification inhibition, contamination, background eDNA signals, incomplete or misannotated barcode databases, cost, and technical platform requirements. This is why molecular methods should be understood as extending and complementing, not replacing, morphological examinations.

Site inference represents one of the most attractive but also most demanding applications. Studies on metagenomic diatom analyses and microbial profiling suggest that environmental biological signatures may contribute to identifying or narrowing possible drowning sites [13,14,87,90]. Diatomological mapping and environmental-comparison studies similarly indicate the potential value of structured reference-water or water-body characterization for site-oriented reconstructions [64,70,77,78,89]. Nevertheless, bacterial or mixed microbial-profile studies should be understood as adjacent aquatic biological approaches, not as strictly diatom- or microalgal-based evidence. The distinction between broad environmental discrimination and precise geographic attribution is critical. A method may differentiate between water types or demonstrate similarity between a body sample and a reference medium, yet still fall short of reliably identifying a specific site, particularly when seasonal variation, hydrology, local microhabitats, and sampling intervals are not standardized. Environmental reference sampling must therefore be treated as a core component of the method, not as an ancillary step.

The future value of these molecular and environmental-biological approaches depends on clearer validation designs, including blinded comparisons, defined thresholds, negative controls, cross-laboratory testing, and the standardized reporting of environmental reference samples. Without these elements, molecular sensitivity may increase the amount of information available to the forensic pathologist but may also increase interpretive uncertainties.

4.2.2. Light Microscopy, SEM/Auto-SEM, and AI-Assisted Image Analysis

A separate line of development concerns image-based detection and automation rather than molecular detection. Conventional light microscopy remains the historically broadest and most accessible platform for forensic diatom examinations and has been applied across a wider range of laboratories, countries, and case contexts. Its limitations include dependence on preparation quality, examiner expertise, background debris, damaged or partial frustules, and taxonomic interpretations. SEM and Auto-SEM may offer high-resolution imaging and support standardized search or counting workflows, but they should not be presented as replacements for conventional light microscopy in routine forensic practice without broader validation.

The charted evidence contained a distinct automation and AI cluster. The included studies covered digital whole-slide image analysis using convolutional neural networks [17], optical or light-microscopy-based database and classification approaches [78,97], object detection systems and YOLO-based models [18,19,96,109,116], DiatomNet [20], comparisons among deep learning image-classification algorithms [23], and AI-based automatic diatom identification systems in practical cases [106]. In parallel, SEM, Auto-SEM, and MD-VF-Auto SEM workflows were widely represented as a methodological family (n = 78; 47.3%), reflecting high-resolution imaging and semi-automated or automated searching (Table 2). This representation should be interpreted cautiously: a substantial part of the SEM/Auto-SEM evidence remains concentrated in a limited number of research environments, and external inter-laboratory reproducibility has not yet reached operational maturity.

Within this methodological family, the microwave digestion-vacuum filtration-automated SEM workflow should be interpreted as an emerging and methodologically valuable technical route, rather than as an established routine standard. Microwave digestion may improve tissue removal and downstream filtration, but the high-temperature and high-pressure digestion steps may also damage or fragment diatom frustules. Such fragmentation can reduce morphologically interpretable material, affect taxonomic identification, and introduce bias into quantitative counting if intact valves and fragments are not distinguished by predefined criteria. In addition, SEM scanning and the automated search do not eliminate contamination risks. Residual diatoms from digestion vessels, filtration devices, membranes, slides, reagents, water, or the laboratory environment remain relevant unless controlled through validated cleaning or single-use procedures, procedural blanks, recovery controls, and contamination controls. Reported high yields or striking counts obtained with MD-VF-Auto SEM are informative, but they require independent replication by different laboratories, operators, instruments, sample matrices, and aquatic environments before supporting routine forensic implementation. Cost, access to microwave digestion systems and SEM platforms, maintenance, calibration, and specialized training also limit generalizability to ordinary forensic laboratories. Accordingly, MD-VF-Auto SEM should be regarded as a promising assistive workflow for detection, enrichment, screening, and counting, but not as a method that has resolved contamination, comparability, or medico-legal interpretation problems [5,6,21,22,37,103,105,107,155,163]. Routine use should therefore remain subject to further validation and practical applicability assessments [5,6].

The main value of automation lies in the potential to reduce manual screening time, support standardized detection workflows, and limit dependence on individual examiner experience. Automated image analyses may be useful when large numbers of particles or fields must be screened, when diatom counts are needed for quantitative comparisons, or when rare target structures are embedded in complex backgrounds. In this sense, automation is not merely a technical convenience; it is connected to reproducibility and auditability. However, it should currently be framed as assistive detection, counting, triage, and classification support, not as autonomous forensic proof.

Yet the review also shows why computational performance should not be equated with medico-legal validity. AI systems depend on annotated datasets, image acquisition conditions, taxonomic diversity, preprocessing, and the representativeness of training and testing data. This dependence is particularly important because diatom taxonomy remains complex and evolving: species boundaries, expert criteria, new genera, and annotation standards remain active issues. A detector trained on one set of expert-labeled images, one preparation method, one microscope, one environmental region, or one taxonomic convention may not perform equivalently in another laboratory or case context. External validation and laboratory transferability are therefore essential. In addition, the forensic meaning of an AI output depends on how it is integrated into an evidentiary workflow. An algorithm may classify particles accurately under test conditions, but forensic interpretation still requires knowledge of sample origin, contamination controls, organ selection, comparison with the drowning medium, and case circumstances.

Auto-SEM and image-based automation also raise practical questions. These systems may support standardized counting and higher-throughput screening, but they require equipment availability, calibration, quality control, specialized operation and maintenance, and agreed interpretive thresholds. Their use may reduce observer dependence at one stage of the process while shifting the validation burden to image acquisition, training-data curation, algorithmic decision thresholds, and inter-laboratory reproducibility. Therefore, the future role of AI and Auto-SEM in forensic diatom analyses is likely to be assistive rather than autonomous. AI may help locate, count, classify, or prioritize candidate diatoms, but the final interpretation should remain embedded within a validated forensic pathway.

4.2.3. Quantitative Comparison and Standardization

A third development concerns quantification and standardization. Quantitative and standardization-oriented studies show that the central methodological concern in the recent literature is reproducibility. The simple presence or absence of diatoms is insufficient because diatoms and other planktonic material may be environmentally ubiquitous, may contaminate samples, and may be detected in small numbers under conditions that do not necessarily establish drowning. In addition, different organs, tissues, digestion methods, filtration systems, microscopes, and counting strategies may produce non-comparable results. The movement beyond presence/absence is therefore a necessary step toward making diatom evidence more reproducible.

Quantitative approaches in the charted literature included diatom counts, tissue-to-medium comparisons, the L/D ratio or lung-to-drowning-medium logic, organ-specific distributions, diatom dispersion in lung tissue, and method comparisons involving digestion, filtration, microscopy, SEM, and Auto-SEM (Table 2). Examples include quantitative investigation of diatom dispersion in lung tissue [24], quantitative analysis of diatoms in drowning diagnosis [25], quantitative comparison of diatoms in lung tissue and drowning medium [26], concordance analysis of diatom types and patterns in lung tissue and drowning medium [27], relationships between diatom abundances in rat organs and environmental waters [28], and L/D ratio validation [112]. These studies illustrate the attempt to transform diatom evidence into a more structured comparative signal.

The value of such indicators is that they can potentially distinguish between incidental or background findings and findings that are more consistent with the aspiration and circulation of the drowning medium. However, quantitative indicators require a high degree of pre-analytical and analytical harmonization. Counts are affected by tissue weights, the organ sampled, sampling locations, digestion times, filtration media, microscope types, counting fields, operator criteria, and the density and composition of diatoms in the water sample. If these variables are not standardized, a quantitative value may appear more objective than it really is. This is particularly relevant for workflows involving microwave digestion and vacuum filtration, because digestion-induced frustule fragmentation may change the number of countable structures and may therefore affect comparisons between tissues, organs, and drowning media [6,21,37,103,105].

4.2.4. Alternative Matrices and Trace-Transfer Contexts

Alternative matrices also appear in the charted evidence, including clothing and fabrics [9,153], maggots [43], vitreous humor [156], bone marrow and bone [157,158,159], bone and tooth [160], synovial fluid [161], closed organs and postmortem or resuscitation-related contexts [39,42,56,155,162], and other specialized or archeological/veterinary matrices [134,139,151,152,167]. These studies are relevant because they expand the potential evidentiary context of diatom analyses beyond routine lung or organ sampling. However, they also require matrix-specific validations. The behavior of diatoms in clothing, bone, maggots, synovial fluid, vitreous humor, or closed organs cannot be assumed to match their behavior in lung tissue or the drowning medium. Transfer, persistence, contamination, decomposition, and postmortem redistribution may differ across substrates.

4.2.5. Cross-Cutting Standardization Requirements

The main implication is that standardization should not be limited to laboratory protocols. It must also include case documentation, water-reference collection, sampling locations, sample mass, digestion and filtration parameters, imaging and counting criteria, reporting thresholds, contamination controls, and interpretive language. Only when these elements are harmonized can quantitative indicators support comparability across studies and laboratories.

4.3. Standardization, Validation, and Medico-Legal Interpretation

The medico-legal significance of diatom and planktonic evidence lies not only in detection but in interpretations. Domain D was represented by 133 studies (80.6%), and the methodological-family synthesis identified contamination, false-positive analyses, and medico-legal interpretations as a distinct cluster of 50 studies (30.3%) (Table 1 and Table 2). This prominence confirms that the main challenge is not whether diatoms can be detected, but what their detection means in a specific forensic context.

Contamination and false-positive findings remain central concerns. Diatoms may be present in the environment, on surfaces, in laboratory reagents, in instruments, or in water used during sampling or processing. They may also enter samples through mechanisms unrelated to antemortem drowning, depending on decomposition, handling, the aspiration of mud or debris, or case-specific circumstances. The included evidence contained studies and reviews directly addressing false-positive results, contamination, and the medico-legal interpretation of diatom findings [5,6,8,9,10]. These records support a cautious interpretation: analytical detection alone does not determine the manner, cause, or circumstances of death.

A specific and under-discussed interpretive variable is the toxicological status of the deceased. A recent forensic appraisal on this point indicates that alcohol or drug intoxication may alter the macroscopic and histological pattern of drowning findings and may also affect the yield or distribution of diatom evidence through changes in aspiration dynamics, agonal time, respiratory effort, and circulatory functions [168]. Case-based evidence also illustrates that intoxication and unusual case dynamics can complicate the reconstruction of water-related deaths [169]. Toxicological status should therefore be discussed as a contextual confounder that may influence detection patterns, rather than as a direct source of contamination or a standalone explanation for false-positive or false-negative findings. Forensic reports that integrate diatom or planktonic findings should explicitly comment on the toxicological status as part of the overall interpretive framework.

A further, recently described and methodologically instructive source of contamination concerns the laboratory consumables themselves. Commercial microscope slides cleaned with diatomaceous earth have been documented as a previously unreported source of fossil-diatom contamination, capable of producing reproducible false-positive findings even under stringent analytical conditions [163]. This observation reinforces a general principle that emerges across the included literature: contamination control must include not only reagents, water, and sampling instruments, but also slides, vials, filters, and other consumables, with mandatory blank controls processed in parallel with case materials.

The distinction between analytical validity and forensic validity is essential. A method may be analytically sensitive, may detect diatom DNA, may classify images accurately, or may recover diatoms from a particular tissue. Nevertheless, forensic validity requires that the finding be interpretable in relation to the autopsy findings, investigative circumstances, toxicology, environmental evidence, sample handling, and alternative explanations. A technically positive result may have limited probative value if the environmental reference is inadequate, if contamination controls are not reported, or if the detected organism is common across many possible water sources.

This distinction is particularly important for emerging technologies. Molecular assays, metagenomic profiles, and AI-assisted classifications can increase the amount and granularity of information obtained from samples. However, more data do not automatically resolve the medico-legal questions. In some circumstances, additional sensitivity may detect low-level background or contamination; in others, machine learning may provide a visually compelling classification that remains under-validated for courtroom use. Therefore, emerging methods should be reported with explicit uncertainties, including limitations of sampling, detection thresholds, reference data, and external validations.

Accordingly, the key question is no longer only whether diatoms, planktonic material, or related molecular signals can be detected. The central question is whether the complete sampling–analysis–interpretation chain is demonstrably fit for the forensic purpose at hand: supporting a drowning diagnosis, contributing to drowning-medium comparisons or site inferences, assessing contamination or false positives, enabling quantitative comparisons, or informing court-oriented medico-legal interpretations. A workflow that is useful for high-throughput screening may not be sufficiently validated for precise site attribution; likewise, a sensitive molecular assay may require different controls and interpretive thresholds when used for contamination assessments rather than diagnostic support. This fit-for-purpose framing helps separate analytical promise from evidentiary weight.

The evidence base also emphasizes that diatom or planktonic evidence should be integrated with conventional forensic pathology rather than isolated from it. Findings may support drowning diagnoses, strengthen site comparisons, or add environmental context, but they should be interpreted alongside autopsy signs, scene information, toxicological results, the circumstances of recovery, and the exclusion of alternative causes of death. The most defensible medico-legal position is not that a single test proves drowning, but that a coherent pattern of pathological, circumstantial, environmental, and analytical findings may support or weaken a drowning hypothesis.

4.4. Implications for Forensic Phycology, Related Aquatic Biological Evidence, and Future Research

The review is directly relevant to forensic phycology because the evidentiary value of diatoms and other microalgae depends on ecological, morphological, taxonomic, molecular, and environmental knowledge. Here, “forensic phycology” is used for forensic applications involving algae and microalgae, particularly diatoms. When the evidence concerns bacterioplankton, mixed microbial profiles, eDNA, metagenomics, metabolomics, or other non-algal markers, the broader expression “forensic phycology and related aquatic biological evidence” is more appropriate. Diatoms are especially relevant because their siliceous frustules can be well preserved and can support morphological and taxonomic comparisons with environmental assemblages, while their distribution varies with the water body, season, hydrology, salinity, substrate, pollution, and sampling conditions.

The reviewed evidence suggests that the field is moving toward integration rather than replacement. Molecular methods may improve sensitivity or environmental comparisons, but they require reference databases and taxonomic understanding. AI-assisted detection may accelerate classifications, but it requires curated image sets, expert annotation, taxonomic stability, and external validations. Quantitative indicators may improve reproducibility, but they depend on knowledge of environmental abundance and species distributions. Diatomological mapping and site-inference studies are explicitly ecological as well as forensic [64,70,77,87,89,90].

For forensic pathology, this creates both opportunities and responsibilities. The opportunity is that phycological, molecular, microbial, and computational evidence can provide information that conventional autopsy findings cannot provide alone, particularly in equivocal drowning cases or site-inference questions. The responsibility is that such evidence must be interpreted within a medico-legal framework that recognizes uncertainties, alternative explanations, and the limits of validation. A multidisciplinary model involving forensic pathologists, phycologists, molecular biologists, microscopists, bioinformaticians, and legal medicine specialists is therefore more appropriate than a single-test model.

This integration is especially important because the end user of the evidence may be a court, not a laboratory. Court-oriented interpretations require transparent methods, defined limitations, reproducible workflows, and language that does not overstate the strength of the evidence. Forensic phycology and related aquatic biological evidence should therefore aim not only to improve detection but also to improve reporting standards and interpretive frameworks.

The most important research gap identified by this scoping review is the absence of studies categorized as multicenter, externally validated, or operationally mature (Level 5) (Table 3). This does not mean that the field lacks useful evidence. Rather, it means that much of the evidence remains controlled, local, single-laboratory, case-based, or proof-of-concept. Future research should therefore prioritize designs that test whether promising methods remain reliable across laboratories, water bodies, sample types, instruments, operators, and case conditions.

Several priorities follow. Harmonized protocols are needed for tissue and organ sampling, drowning-medium collection, sample masses, anatomical sites, timing, storage, contamination control, and documentation. Curated molecular and image-reference datasets are also needed, together with standardized environmental sampling for site-inference studies. DNA-based methods require reference data that support interpretations across regions and water types, while AI models require annotated image sets that include taxonomic diversity, debris, damaged frustules, variable preparations, and external test data. Future validations should also include explicit contamination controls, blinded designs, real forensic casework where possible, and more structured reporting of methods, controls, uncertainties, and interpretive limits.

Several of these priorities have been articulated with operational detail in a recent systematic review of the diatom test for fatal drowning, which screened 372 records and included only 17 studies meeting predefined scientific-quality criteria [6]. Among the methodological gaps explicitly identified are the near-absent use of positive (spike) controls to quantify diatom loss during digestion and centrifugation, the inconsistent comparison between tissue and water-medium profiles (performed in only 53% of included studies), and the limited involvement of certified diatomologists in taxonomic classifications (reported in only one of seventeen studies). The same review proposed concrete recommendations for forensic laboratories: the systematic use of single-use sterile instruments with bi- or tri-distilled water along the entire processing chain; adoption of recovery-rate testing through diatom spiking; the standardization of tissue volumes and replicate sampling; alignment with European standards (EN) 13946 [170] for diatom digestion and EN 14407 [171] for taxonomic enumeration; inter-laboratory intercalibration exercises analogous to existing freshwater-algae networks; and routine collaboration with independent expert diatomologists [6]. These recommendations are directly relevant to forensic phycology and to broader aquatic biological evidence workflows, and provide a concrete starting point for future harmonization efforts across the methodological families mapped in this review.

Looking beyond strictly phycological evidence, a parallel methodological frontier is emerging from postmortem metabolomics combined with machine-learning classifiers, which have recently been proposed to discriminate drowning from other causes of death on the basis of high-dimensional metabolic signatures rather than morphological or genetic markers. In a large forensic cohort of 503 drowning cases and four matched control groups (hangings, chronic heart disease, intoxications, and trauma; n = 497–516 each), binary orthogonal partial least-squares discriminant analysis (OPLS-DA) models distinguished drowning from each comparator with sensitivities of 83–87% and specificities of 78–89%, and validation-set AUC values for drowning of 0.87. Notably, postmortem submersion time and post-submersion intervals had minimal impact on metabolomic classifications, while the main source of misclassification was prolonged in-hospital care before death. Significantly enriched pathways included glycerophospholipid metabolism, steroid hormone biosynthesis, fatty-acid activation, and cytochrome P450 drug metabolism, suggesting that the metabolic signature of drowning reflects pathway-level biological changes rather than only circumstantial grouping [172]. Such approaches do not replace forensic phycology and are not phycological evidence in the strict sense. Instead, they exemplify a broader trajectory: the future of forensic drowning diagnosis is likely to rely on the convergence of complementary signatures (diatoms, eDNA, microbial profiles, metabolomic fingerprints, automated imaging) interpreted within a multidisciplinary, validated framework rather than on any single ancillary test.

4.5. Strengths and Limitations

This review used a PRISMA-ScR framework to map a heterogeneous and interdisciplinary body of evidence across three major bibliographic sources: Scopus, PubMed, and Web of Science Core Collection. The workflow included deduplication, title/abstract screening, full-text retrieval, eligibility assessment, standardized exclusion reasons, and structured data charting, while integrating technical, phycological, computational, and medico-legal domains rather than focusing on one analytical platform alone.

A further strength is the distinction between methodological families and validation maturity. Grouping conventional microscopy, SEM/Auto-SEM, PCR/qPCR, sequencing, AI-assisted image analyses, environmental mapping, and medico-legal interpretations under a single “diatom test” label would obscure important differences in forensic targets and validation needs.

The review also has limitations. Eighteen reports sought for retrieval were not obtained and were not assessed for content eligibility, which may have led to the underrepresentation of some regional or specialized evidence. Because the search was limited to Scopus, PubMed, and the Web of Science Core Collection, studies available only through national or language-specific databases may have been missed. Some records required additional bibliographic matching because of title variants, scanned files, translated materials, or incomplete metadata. The included studies were also heterogeneous in design, sample types, analytical platforms, forensic targets, and reporting quality, which precluded a meta-analysis and limits direct comparison across studies.

As a scoping review, this study mapped evidence rather than grading certainty or estimating diagnostic accuracy. It did not perform pooled sensitivity or specificity analyses, nor a formal risk-of-bias assessment; the Level 1–5 categories were used only to describe the mapped methodological maturity. Title/abstract screening, full-text eligibility assessment, initial charting, and validation-level assignment were performed by one reviewer and subsequently checked by the other authors for uncertain or interpretively important records. This approach provided consistency across a heterogeneous evidence base, but it is less robust than fully independent duplicate screening and should be considered when interpreting the map. The findings should therefore be read as a structured map of available evidence and gaps, not as a quantitative validation of any single method.

5. Conclusions

This scoping review indicates that forensic diatom evidence has moved beyond the classical presence/absence test. Contemporary research increasingly addresses how diatom-based and related aquatic biological evidence should be generated, compared, quantified, standardized, and interpreted for forensic purposes.

The main implication is integrative rather than replacement-based. Classical morphology, phycological expertise, molecular methods, computational tools, and medico-legal reasoning may provide complementary information when embedded in transparent and validated workflows. Within such workflows, the quality and reliability of the final interpretations remain closely linked to the expertise of the human operator, particularly in specimen handling, microscopic recognition, comparative assessments, and medico-legal contextualization. However, analytical promise should not be confused with forensic proof. Because no included study reached multicenter, externally validated, or operationally mature Level 5 evidence, emerging approaches should remain adjunctive tools rather than standalone proof of drowning or precise site attribution. No single methodological family, including MD-VF-Auto SEM, SEM/Auto-SEM, AI-assisted detection, or DNA-based analyses, should currently be treated as a replacement for validated, context-integrated forensic interpretations.

Future work should prioritize harmonized sampling and processing protocols, standardized environmental reference collection, explicit contamination controls, curated molecular and image-reference datasets, the external validation of AI models, inter-laboratory reproducibility testing, and the transparent reporting of uncertainties. The most robust path forward is a validated framework for forensic phycology and related aquatic biological evidence, in which microscopy, phycology, molecular biology, bioinformatics, automation, and forensic pathology provide complementary evidence while preserving medico-legal caution.