Search Results (5,470)

Background: Former and current smokers in lung cancer screening remain at elevated risk for lung cancer despite smoking cessation. We and others have shown that curcumin (CUR) exhibits anti-inflammatory and antiproliferative effects but is limited by poor bioavailability. However, since CUR is lipophilic, co-administration with ω-3 FAs represents a mechanistically rational strategy to enhance delivery and target complementary pathways, including signal transducer and activator of transcription 3 (STAT3) and the transcription factor NF-κB (NF-κB) signaling for lung cancer chemoprevention. Methods: We conducted a randomized, single-blind, placebo-controlled Phase II pilot study evaluating CUR combined with ω-3 FAs in high-risk former and current smokers with CT-detected pulmonary nodules. Participants received intervention agents with active-dose groups (low dose = 3; high dose = 9) or a placebo (n = 7) for 6 months. Primary endpoints included radiologic changes in nodule size, number, and density. Secondary endpoints included safety, adherence to the study agent and exploratory biomarker analyses. Correlation analyses of imaging-derived metrics were performed to assess relationships among LDCT parameters. Results: Nineteen participants were enrolled (intervention, n = 12; placebo, n = 7). Eighteen (11 intervention, 7 placebo) subjects completed post-intervention imaging. One subject was unable to complete follow-up and study-related procedures. Data from the treatment arms were pooled for analysis and comparison with the placebo arm. No statistically significant between-group differences were observed in primary imaging endpoints. The intervention was well tolerated, with predominantly grade 1 adverse events. Exploratory analyses demonstrated consistent positive correlations among established imaging biomarkers, with clustering of size-based metrics (mean diameter, volume, sum of longest diameters) and density-based parameters. Multidimensional scaling supported this structure, indicating internal coherence among imaging-derived endpoints. Conclusions: Although no statistically significant treatment effect on the image biomarkers was observed, this pilot study demonstrates feasibility challenges and identifies coherent imaging biomarkers that may serve as intermediate endpoints in early-phase chemoprevention trials. These results support further investigation of strategies utilizing agent combinations with enhanced bioavailability and safety and refinement of trial design in high-risk lung cancer patient populations. Full article

Background: Colorectal neoplasia in adults younger than 50 years has attracted increasing clinical attention because of the rising incidence of early-onset colorectal cancer. However, the clinicopathological characteristics of colorectal lesions detected in younger adults remain incompletely described. This study aimed to characterize the clinical, anatomical, and histopathological features of colorectal neoplasia in adults younger than 50 years undergoing colonoscopy. Methods: This retrospective single-center study included adults aged 18–49 years who underwent complete colonoscopy between January 2018 and December 2023. After exclusion of individuals with normal colonoscopy findings, 152 patients with at least one colorectal lesion were included. Lesions were classified as benign, advanced, or malignant according to established histopathological criteria. Demographic, endoscopic, and pathological characteristics were analyzed and compared across lesion categories. Results: Among 152 patients, 83 (54.6%) had benign lesions, 64 (42.1%) had advanced neoplasia, and 5 (3.3%) had malignant lesions. Advanced lesions were observed predominantly in individuals aged 40–49 years. Most lesions were detected in symptomatic patients undergoing clinically indicated colonoscopy. Histopathological evaluation demonstrated a higher frequency of villous/tubulovillous adenomas, sessile serrated lesions, and high-grade dysplasia among advanced lesions. Larger lesion size was strongly associated with advanced pathological features (p < 0.001). In multivariable analysis, patients aged 30–39 years had significantly lower odds of advanced or malignant neoplasia than those aged 40–49 years, whereas no independent associations were observed for sex, smoking status, alcohol use, family history of colorectal cancer, or lesion location. Conclusions: In this selected cohort of adults younger than 50 years with detected colorectal lesions undergoing clinically indicated colonoscopy, advanced neoplasia represented a substantial proportion of cases. These findings should not be extrapolated to the general population of adults younger than 50 years or to screening cohorts. Full article

Background: Mismatch repair deficiency (MMRd) is a central molecular determinant of colorectal cancer (CRC) biology, prognosis, and treatment response, and Universal Tumour Screening (UTS) is advocated for Lynch syndrome (LS) detection; yet real-world performance across clinical subgroups remains limited. We evaluated MMRd distribution and UTS-based LS detection in a large consecutive surgical cohort. Methods: We retrospectively analyzed 1022 consecutive CRC patients undergoing surgical resection at the University Hospital of Padua (2015–2023). MMR status was assessed by immunohistochemistry; MMRd cases underwent reflex BRAF mutation testing and, when available, MLH1 promoter methylation analysis, followed by germline multigene panel testing for suspected LS. Clinicopathological features were compared by MMR status, age at onset, and tumour location. Results: MMR testing was performed in 875 patients (85.6%), rising from 67.0% (2015–2017) to 97.4% (2021–2023). MMRd was identified in 139 tumors (15.9%) and was independently associated with age ≥ 70 years, colonic location, and stage 0–II. Of 22 patients with confirmed LS, 13 (59.1%) were newly identified through UTS; family history showed no significant univariate association with LS status and was not independently associated with MMRd after multivariable adjustment. MMRd prevalence was numerically higher in early- than late-onset CRC (20.0% vs. 15.4%), approaching significance after multivariable adjustment (OR 1.90, 95% CI 0.99–3.64; p = 0.054); hereditary syndromes were also more frequent in early-onset disease. MMRd was markedly rarer in rectal than colonic cancer (4.1% vs. 22.5%; p < 0.0001), though MMRd rectal cancers arose in younger patients. Conclusions: UTS identified a substantial proportion of LS carriers missed by age- or family-history criteria. The relationship between age and MMRd prevalence proved more nuanced than a simple comparison would suggest, reinforcing the value of universal over selective testing across the age spectrum, while MMRd rectal cancer shows a distinct younger profile relevant to immunotherapy-based organ preservation. Full article

Early detection of colorectal cancer (CRC) is a major determinant of patient prognosis, as survival strongly depends on disease stage at diagnosis. Despite advances in screening programs, a significant proportion of CRC cases are still diagnosed at advanced stages, underscoring the need for improved early detection strategies. Most sporadic CRCs arise through the adenoma–carcinoma sequence over 10 to 15 years, providing a window for the detection of premalignant lesions, such as advanced adenomas. Current screening approaches are based on colonoscopy or its combination with stool-based tests. Although colonoscopy is the gold standard, it is an invasive technique with high associated costs and limited patient compliance. Stool-based tests are non-invasive and more widely accepted but lack specificity and sufficient sensitivity for detecting premalignant lesions. In this context, liquid biopsies have emerged as a promising minimally invasive alternative for identifying tumor-derived biomarkers in biological fluids such as blood or stool. Small non-coding RNAs (sncRNAs), and particularly microRNAs (miRNAs), have gained considerable attention as non-invasive biomarkers for their highly stability, resistance to handling conditions, and reliable quantification even in low-input samples. Single miRNAs and miRNA signatures detected in biofluids and combined with clinical parameters have shown promise for CRC detection. However, their utility for detecting advanced adenomas remains insufficiently characterized. Further validation in large, independent cohorts and standardization of analytical methods are required before their clinical implementation. Despite these challenges, sncRNA-based liquid biopsies represent a promising approach for improving early detection of CRC and, consequently, its prognosis. Full article

Ovarian cancer (OC) remains the most lethal gynecological malignancy, mostly due to its frequent diagnosis at advanced stages, early peritoneal dissemination, ascites formation, and limited sensitivity of currently available approaches for early detection. Extracellular vesicles (EVs), particularly exosomes, mediate intercellular communication through the transfer of proteins, lipids, metabolites, and nucleic acids. In OC, EV-associated protein profiles reflect both tumor-cell-intrinsic programs and the complex interactions between malignant cells and the peritoneal microenvironment. This review summarizes current evidence regarding the involvement of exosomal proteins in OC progression, with particular emphasis on epithelial–mesenchymal transition, mesothelial reprogramming, extracellular matrix remodeling, angiogenesis, immune suppression, peritoneal dissemination, and platinum resistance. Mechanistic studies indicate that exosomal proteins, including CD44, the integrin α5β1/asparaginyl endopeptidase complex, annexin A2, low-density lipoprotein receptor-related protein 1, and programmed death-ligand 1, can directly contribute to metastatic niche formation and tumor progression. In parallel, proteomic studies of plasma-, serum-, ascites-, peritoneal-fluid-, and uterine-lavage-derived EVs have identified candidate liquid-biopsy biomarkers, including MUC1, EpCAM, FOLR1, integrins, complement- and coagulation-related proteins, and proteins associated with treatment resistance. To integrate the biological significance of proteins reported in OC-associated exosomes, we additionally performed protein–protein interaction and functional enrichment analyses. These analyses revealed interconnected protein groups associated with cell adhesion, oxidative stress adaptation, secretory remodeling, lipid metabolism, extracellular matrix organization, and inflammatory signaling. Taken together, the available evidence supports exosomal proteome profiling as a promising approach for investigating OC dissemination and developing minimally invasive diagnostic and prognostic tools. However, standardized EV isolation, quantitative proteomics, functional validation, and independent clinical cohorts remain essential for translation into clinical practice. Full article

Background and objectives: Myeloproliferative neoplasms (MPNs) are blood cancers characterized by elevated blood cell counts, bone marrow fibrosis (BMF), and chronic inflammation, which drives disease progression. BMF results from disrupted collagen turnover in the bone marrow extracellular matrix (ECM), making its reduction or stabilization a key therapeutic goal. Non-invasive biomarkers reflecting collagen turnover could potentially improve early detection of BMF and monitor disease activity. Methods: We evaluated serum biomarkers of collagen turnover in 130 MPN patients (MPN subtypes: ET = 49, PV = 60, pre-PMF = 8, PMF = 13) included in the DALIAH trial (ClinicalTrials.gov identifier: #NCT01387763). Type I and type III collagen formation (PRO-C1 and PRO-C3) and MMP-degraded type I, III, and IV collagens (C1M, C3M, and C4M) were measured by ELISA in serum. Biomarker levels were compared to age- and sex-matched healthy individuals and were assessed according to disease subtypes, somatic driver mutations, JAK2V617F VAF, and fibrosis grade. Furthermore, we studied correlations between the biomarker levels and conventional hematological markers for disease activity, such as hemoglobin, white blood cell count (WBCs), platelet counts, and lactate dehydrogenase (LDH). Results: Baseline PRO-C3 (p = 0.0005) and C1M (p = 0.0102) were elevated in MPN patients compared to healthy individuals, whereas C3M was decreased (p < 0.0001). Patients with primary myelofibrosis (PMF) had higher levels of PRO-C3 compared to patients with ET (p = 0.0041) and PV (p = 0.0172), correlated with higher JAK2V617F VAF (≥50%) (p = 0.0069) and advanced fibrosis grade (p = 0.0002). In addition, PRO-C3 was positively correlated to LDH, which is a biomarker of disease activity in MPNs (r = 0.5754, p < 0.0001). Conclusions: Taken together, these findings emphasize the role of ECM remodeling in MPN pathophysiology and the potential of soluble ECM neoepitopes as biologically plausible disease markers. Full article

Cardio-oncology has emerged as a critical discipline in modern medicine due to the growing population of cancer survivors and the increasing recognition of cardiovascular disease as a major cause of morbidity and mortality in this population. While there have been significant advances in chemotherapy such as advances in chemotherapy, targeted therapies, immunotherapies, and radiation therapy, these treatments are linked to a wide range of cardiovascular toxicities such as myocarditis, arrhythmias, and progressive fibrotic remodeling. These toxicities tend to be caused by interrelated and multifactorial cellular processes including endothelial damage, immunological dysregulation and mitochondrial dysfunction. Therefore, early diagnosis of subclinical harm using multimodal surveillance strategies integrating clinical risk assessment, specialized clinical imaging, and circulating biomarkers has replaced reactive care of overt cardiotoxicity in modern cardio-oncology. Echocardiographic techniques such as longitudinal strain, with the help of biomarker-guided surveillance of natriuretic peptides and cardiac troponins, have made the early detection of cardiac dysfunction more effective. However, the current prediction risk models are still constrained and limited by their dependence on static clinical variables and their partial integration of biological mechanisms. This review examines the current methods for risk stratification and surveillance throughout the cancer care continuum, including baseline assessment, monitoring during active therapy, and long-term survivorship surveillance. Future approaches for tailored cardiovascular care are also highlighted, including new advances in precision cardio-oncology such as multi-omics profiling, molecular biomarkers, artificial intelligence, and mechanism-guided preventative strategies. Therefore, advancing biologically informed and risk-adapted monitoring frameworks may enhance early diagnosis, maximize cardioprotective measures, and lower long-term cardiovascular consequences. Full article

Background: Machine learning (ML) is increasingly applied in medicine, underscoring the need for transparent and clinically relevant models. In gastrointestinal oncology, most ML studies rely on raw imaging data, which limits clinical adoption due to poor interpretability and the difficulty of collecting high-quality, large-scale video and image datasets in routine practice. Endoscopic ultrasound (EUS) plays a central role in the evaluation of pancreatic cancer; however, structured EUS features remain underused in predictive modeling. Objective: To assess the performance and interpretability of ML models for diagnosing pancreatic ductal adenocarcinoma (PDAC) using routinely collected EUS variables. Methods: We conducted a retrospective multicenter study using data from two Italian hospitals (n = 641) for model training and internal validation and from a third hospital (n = 120) for external validation, collected from 2015 to 2023. Decision trees, random forests, naïve Bayes and other classifiers were developed and evaluated. Model performance was assessed in terms of discriminative ability, calibration, and selective prediction. Results: All models demonstrated high discriminative performance (AUC ≥ 0.90). Decision trees provided the most favorable balance between interpretability and accuracy (balanced accuracy = 0.87; sensitivity = 0.89). Calibration and selective prediction analyses confirmed the robustness of the models. Conclusions: These findings demonstrate the feasibility of implementing interpretable yet high-performing ML models for PDAC diagnosis in real-life endoscopic settings. Full article

Background: The increasing use of routine oncological imaging has led to more frequent detection of incidental pulmonary embolism (PE), potentially modifying the contemporary clinical presentation of cancer-associated pulmonary embolism (CAPE). Methods: We performed a retrospective cohort study including 381 consecutive patients hospitalized with acute PE, of whom 58 had active cancer and 323 had no active malignancy. The primary endpoint was a Severe Hemodynamic Presentation Composite Endpoint (SHPCE), defined as shock, systolic blood pressure < 90 mmHg, and/or high-risk PE according to European Society of Cardiology criteria. Clinical characteristics, severity markers, management strategies, and in-hospital outcomes were compared between patients with CAPE and non-cancer PE (NCPE). Multivariable logistic regression analyses were performed to evaluate factors associated with SHPCE and incidental PE. Results: The patients with CAPE had higher PESI (129.8 ± 29.3 vs. 110.1 ± 32.3; p < 0.001) and sPESI scores (3.03 ± 0.56 vs. 2.68 ± 0.81; p < 0.001), and lower hemoglobin levels (11.7 ± 1.9 vs. 13.2 ± 1.8 g/dL; p < 0.001). Incidental PE was more frequent in CAPE than NCPE cases (13.8% vs. 1.9%; OR 8.45, 95% CI 2.81–25.39; p < 0.001). Despite their higher baseline risk scores, patients with CAPE and NCPE showed similar rates of SHPCE (12.1% vs. 17.6%; p = 0.295), ICU admission (10.3% vs. 11.5%; p = 1.000), and in-hospital mortality (10.3% vs. 11.1%; p = 0.858). In multivariable analyses, right ventricular dysfunction (RVD) showed the strongest association with SHPCE (adjusted OR 10.21, 95% CI 5.34–19.52; p < 0.001), whereas active cancer was not associated with severe presentation. Conclusions: Active cancer was associated with higher clinical risk scores and a greater prevalence of incidental PE but not with increased hemodynamic severity or adverse in-hospital outcomes. Acute PE severity appeared to be more closely related to right ventricular involvement than to cancer status, supporting a severity-based rather than cancer-based approach to risk assessment. Full article

Background/Objectives: Cutaneous squamous cell carcinoma (cSCC) is the second most common non-melanoma skin cancer (NMSC) and represents the leading cause of NMSC-related deaths. Despite its growing global burden, comprehensive epidemiological and clinicopathological data from Easter Europe remains limited. This study aimed to evaluate the epidemiological, clinical, histopathological, and surgical characteristics of cSCC diagnosed before, during and after the COVID-19 pandemic. Methods: We conducted a retrospective, descriptive observational study including 332 lesions diagnosed between January 2017 and December 2025 at the Clinical Pathology Department of the Mureș Clinical County Hospital. Demographic, epidemiological, topographic, histopathologic, surgical, and volumetric parameters were analyzed. Tumours were stratified into low-, high-, and very-high-risk categories according to the National Comprehensive Cancer Network (NCCN) criteria. Results: The cohort demonstrated a significant male predominance (n = 193 vs. n = 139; p = 0.0355), with females presenting a higher median age (77 vs. 75; p = 0.0489). Lesions were predominantly located in the head and neck region (n = 216; p < 0.0001), which was significantly associated with very-high-risk tumours (p = 0.0051). Low-risk tumours accounted for 62.35% of cases, while high-risk and very-high-risk lesions comprised 19.88% and 17.77%, respectively (p < 0.0001). Ulcerations were strongly associated with very-high-risk tumours (p < 0.0001). Poor differentiation was more frequent outside the head and neck region (p < 0.0001) and varied significantly across the pandemic periods (p = 0.0349). Tumoral and excision volumes were higher in very-high-risk (p = 0.0070; p = 0.0004) and ulcerated tumours (p < 0.001), with a peak in volume during the COVID-19 period (p < 0.0001). A decrease through the years of diagnosis was observed in tumoral volumes (r = −0.2295; p < 0.0001) and patients showed a weak positive correlation with diagnosis year (r = +0.13; p = 0.019). Conclusions: The study provides an epidemiological and clinicopathological characterization of cSCC within one of the largest Romanian cohorts to date. Tumour aggressiveness was primarily driven by histopathological and topographical features rather than demographic factors. While the COVID-19 pandemic did not induce persistent changes in tumour risk profiles or surgical outcomes, it influenced diagnosis timing and tumour burden. These findings highlight the importance of incorporating temporal and emerging systemic factors, such as pandemics, and infectious events, into future epidemiological models to improve preparedness, early detection, future treatment schemes, and risk stratification in cSCC. Full article

In this study, we propose a novel optical sensor architecture based on two-dimensional photonic crystals for the early detection of cervical cancer (HeLa). The structure consists of a central hollow-disk micro-cavity designed to accommodate biosamples, surrounded by a periodic array of GaAs rods. The detection principle relies on variations in the biosample refractive index, inducing a spectral shift in the resonance. To overcome the limitations of conventional 2D-FDTD method parametric sweeps, an artificial intelligence framework was developed to optimize the geometric parameters of the proposed photonic crystal optical sensor. First, a Random Forest algorithm was employed to identify promising regions of the geometric design space. Next, a multilayer artificial neural network (ANN-MLP) was trained as a high-fidelity surrogate model ($R^2$ = 98.58%) and coupled with a Particle Swarm Optimization (PSO) algorithm to determine the optimal structural configuration. The optimized sensor geometry subsequently achieved an average sensitivity of 5512.91 nm/RIU, a quality factor of 6139.15 and a detection limit of $5.64\times {10}^{-5}$ RIU, demonstrating the effectiveness of the proposed AI-assisted design strategy. The optimized design reduces classical performance trade-offs and exhibits high tolerance to nanometric fabrication deviations below ±20 nm. Full article

Background: Oral cancer remains a major global health concern, with increasing incidence and mortality, especially in developing countries. Public awareness is crucial for early detection and improved outcomes. Objective: To assess oral cancer awareness, recognition of risk factors and early signs, screening behavior, and care-seeking preferences among patients attending a university dental hospital in Türkiye. Methods: This cross-sectional, facility-based survey included 1360 volunteer patients who attended the Faculty of Dentistry between January and December 2025. Consecutive sampling was used to include all eligible and consenting individuals. A structured questionnaire assessed participants’ sociodemographic characteristics, oral cancer awareness, knowledge of symptoms and risk factors, screening history, and lifestyle habits. Data were analyzed using IBM SPSS Statistics 26.0 (IBM Corp., Armonk, NY, USA). Descriptive statistics, chi-square tests, and multivariable logistic regression analyses were performed. Statistical significance was set at p < 0.05. Main outcomes and measures: The primary outcome was oral cancer awareness (measured by the question “Have you ever heard of oral cancer?”). Secondary outcomes included knowledge of risk factors and early clinical signs, screening history, preferred healthcare contact for suspected oral cancer, and lifestyle factors (tobacco and alcohol). Associations between sociodemographic variables and oral cancer awareness were also evaluated. Results: The sample consisted of 1360 participants (56.4% female; mean age: 33.44 ± 14.60 years; range: 18–70). Of the 1377 questionnaires collected, 1360 were complete and included in the final analysis (questionnaire completion rate: 98.8%). Breast cancer was the most recognized cancer (96.9%), whereas only 50.8% of participants had heard of oral cancer. Dentists (39.4%) were the primary source of information. Tobacco use (92.2%) and alcohol consumption (80.3%) were the most frequently identified risk factors, whereas only 58.3% recognized human papillomavirus as a risk factor for oral cancer. In univariate analyses, oral cancer awareness was significantly associated with age, marital status, education level, monthly income, occupation, alcohol use, and the timing of the most recent dental visit (p < 0.05). In multivariable logistic regression, undergraduate education (OR = 2.008, 95% CI: 1.135–3.553), higher monthly income (>100,000 TRY vs. <25,000 TRY, OR = 2.398, 95% CI: 1.571–3.658), alcohol use (OR = 1.521, 95% CI: 1.082–2.139), and a dental visit within the previous year (OR = 4.130, 95% CI: 1.590–10.730) remained independently associated with oral cancer awareness. Conclusions: Oral cancer awareness in this study population was limited. The findings suggest that targeted educational strategies may be beneficial for improving oral cancer awareness. Strengthening dentists’ preventive role and promoting education through social media may help improve oral cancer awareness and facilitate earlier detection. Full article

Background/Objectives: Automated dermoscopic image analysis can assist research on early skin cancer detection, but deep learning models may learn acquisition-related shortcuts, including terminal hairs, illumination gradients, gel bubbles, shadows and measurement markings. This study evaluates a causal uncertainty-decomposed ensemble (CUDE) for confounding-aware skin-lesion classification within the ISIC 2019 benchmark setting. Methods: CUDE uses a structured variational autoencoder (SVAE) to impose separate lesion-relevant and acquisition-related nuisance latent heads. Three stochastic latent-space experts are trained on complementary streams and are integrated by a Dirichlet-based fusion module trained with nuisance sampling. The causal graph is used as a modeling prior rather than proof of causal identification. Results: Using a stratified internal split of the labeled ISIC 2019 training collection, CUDE achieved a balanced accuracy of 89.7% and an AUROC of 0.972. Under the predefined synthetic artifact protocol, CUDE showed a relative performance drop of 5.0%, compared with 9.6% for the deep ensemble. Deferring the 10% most uncertain cases increased non-deferred accuracy from 89.7% to 93.2% and reduced false-negative rates for melanoma, BCC and SCC in the non-deferred subset. Conclusions: Within the evaluated ISIC 2019 split and controlled synthetic corruption protocol, structured latent separation, stochastic expert diversity and Dirichlet fusion were associated with improved benchmark robustness and calibration. These results should not be interpreted as evidence of broad real-world robustness or clinical deployment readiness; external multi-center, device-diverse and skin-tone-diverse validation remains required. Full article

Background/Objectives: Cancer remains a major public health challenge, and public awareness of warning signs, risk factors, and prevention methods is essential for early detection and primary prevention. This study aimed to assess cancer knowledge among adults in Poland and identify sociodemographic factors associated with self-reported awareness. Methods: A nationwide cross-sectional survey was conducted in January 2026 among 1087 adults in Poland using computer-assisted web interviewing (CAWI). Non-probability quota sampling was applied based on sex, age, and place of residence. Results: Only 12.7% of respondents reported rather high or very high cancer knowledge. The most frequently recognized warning sign was a lump, mass, or thickening (66.9%). Tobacco use (64.9%) and genetic or familial predisposition (61.8%) were the most commonly identified risk factors, whereas 41.9% recognized overweight or obesity as a cancer risk factor. Smoking cessation and participation in cancer screening programs were each identified as preventive measures by 57.5%. However, 20.8% incorrectly believed that dietary supplements protect against cancer, and 16.2% endorsed “detox” beverages as cancer-preventive. In multivariable analysis, higher education (aOR: 1.67; 95% CI: 1.13–2.46), occupational activity (aOR: 1.56; 95% CI: 1.02–2.37), personal history of cancer (aOR: 4.27; 95% CI: 2.75–6.64), and family history of cancer (aOR: 1.85; 95% CI: 1.25–2.73) were independently associated with higher self-reported cancer knowledge. Conclusions: The findings indicate insufficient cancer awareness among the surveyed sample of Polish adults, particularly among men, younger adults, and individuals with lower educational attainment. Targeted educational initiatives may help improve knowledge of cancer symptoms, risk factors, and prevention strategies. Full article

Colorectal cancer remains a leading cause of cancer-related mortality worldwide, with automated polyp classification from endoscopic images offering a promising avenue for improving early detection. Existing approaches rely on single convolutional neural network (CNN) backbones with manually designed classification heads, limiting both representational capacity and deployment flexibility. This paper presents a swarm intelligence-augmented multi-backbone deep learning framework for eight-class gastrointestinal lesion classification on the Kvasir benchmark. Four CNN backbones (ResNet50, DenseNet121, MobileNetV2, EfficientNetB3) are independently fine-tuned using a two-phase transfer learning protocol and their penultimate-layer features concatenated into a 5888-dimensional representation, reduced to 256 dimensions via PCA. Five swarm intelligence algorithms—Particle Swarm Optimization, Artificial Bee Colony, JADE, L-SHADE, and CMA-ES—are benchmarked on the classification head architecture search task; all independently converge to tanh activation, a consistent pattern across independently initialized algorithms that is suggestive of, though not conclusive evidence for, particular geometric properties of PCA-transformed deep feature spaces. The PSO-optimized single-layer head (284 units, tanh) outperforms a manually designed three-layer baseline by 0.75% while using 67% fewer parameters. SI-guided class weight optimization yields targeted F1 improvements on the two most clinically significant classes (polyps: +0.015, ulcerative-colitis: +0.013). The fixed-head classifier trained on fused four-backbone features achieves 91.08% accuracy on Kvasir v2 (multi-seed mean 91.47% ± 0.49 across nine converging seeds; one seed failed to converge and is disclosed rather than excluded), below end-to-end DenseNet121 (92.25%; Wilcoxon p = 0.31, not statistically significant), while enabling classifier updates in under 30 s; a three-backbone subset dropping the weakest backbone (EfficientNetB3) reaches 92.33%, exceeding the full four-backbone fusion. Cross-dataset evaluation on Kvasir v1-to-v2 confirms near-zero generalization gaps across dataset scales; a restricted two-class evaluation on HyperKvasir (the only two of eight classes with usable labeled data) reaches 96.28% accuracy, and dual Grad-CAM with SI minimal sufficient region analysis, validated quantitatively against Kvasir-SEG ground-truth masks, provides spatially grounded, clinically interpretable explanations. Full article