Search Results (2,202)

Background/Objectives: Chest X-ray (CXR) images are a widely used first-line screening tool for pulmonary tuberculosis (TB) detection but are difficult to interpret, which has increased demand for an automated screening tool. Deep-learning-based computer-aided diagnosis systems have demonstrated a classification performance comparable to that of trained radiologists, but they rely on dense annotations such as lesion-level or pixel-level labels, which are costly and difficult to obtain in routine clinical workflows. We developed CPM-XNet, an annotation-efficient framework for lesion-annotation-free downstream TB classification in CXR images. Methods: CPM-XNet incorporates a compressing–projecting mask (CPM) to provide soft lung-aware modulation while preserving global contextual information. The CPM-modulated images are then used for downstream classification with multiple convolutional neural network backbones and a vision transformer baseline. Results: Experiments were conducted using an internal hospital dataset and public TB datasets, and CPM-XNet showed improved performance compared with baseline models trained on unmodulated images. In a repeated-seed evaluation of the main ResNet-101 configuration on the Tung cohort, CPM-ResNet101 showed higher and more stable performance than the non-CPM counterpart and demonstrated significant paired improvement using McNemar’s exact test. An ablation analysis indicated that CPM modulation was the main contributor to performance improvement while data augmentation and the classifier architecture further influenced the overall robustness. Conclusions: CPM-XNet provides an annotation-efficient strategy for lesion-annotation-free downstream TB classification in CXR images. The findings support preliminary technical feasibility, although larger, naturally imbalanced, cross-institutional validation is required before clinical deployment can be inferred. Full article

Purpose: To investigate the agreement on perfusion parameters derived from two different commercially available solutions for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in patients with prostate cancer (PCa). Methods: A total of 50 patients (mean age, 71.6; range 56–86) who had undergone radical prostatectomy between December 2021 and September 2022 were included in this retrospective study. All patients had undergone DCE-MRI on a single 3T-MR scanner. Tumor segmentation on MR images was performed by two radiologists in consensus after radiologic-pathologic correlation using topographic maps as a reference standard. Subsequently, four perfusion parameters were calculated by dedicated commercially available solutions from two different vendors. Both solutions adopted a population-based arterial input function and an extended Tofts model as the pharmacokinetic model. The perfusion parameters were as follows; volume transfer constant (K^trans), rate constant (k_ep), volume fraction of extravascular extracellular space (v_e), and volume fraction of plasma (v_p). The differences between paired measurements were compared by Bland–Altman analyses and the reproducibility was evaluated using the intraclass correlation coefficient (ICC). Results: The study population consisted of Gleason score (GS) 6 (n = 12), GS 7 (n = 34), GS 8 (n = 1), and GS 9 (n = 3). Significant differences were found for all parameters (p < 0.0001). Mean differences were as follows: K^trans, −0.2102 (95% confidence interval; −0.2687 to −0.1518); k_ep, −0.7632 (−0.9005 to −0.6258); v_e, −0.1507 (−0.2422 to −0.05907); v_p, −0.02929 (−0.03383 to −0.02476). ICCs for average measures were as follows: K^trans, 0.2989 (−0.2355 to 0.6021); k_ep, 0.6883 (0.4507 to 0.8231); v_e, −0.1331 (−0.9967 to 0.3570); v_p, 0.2653 (−0.3106 to 0.5881). Conclusion: All perfusion parameters were significantly different between the two solutions. Therefore, comparison of perfusion parameters across different solutions is not recommended. Full article

Lung cancer remains the leading cause of cancer-related mortality worldwide, largely due to late-stage diagnosis, biological heterogeneity, and persistent challenges in staging and treatment selection. This narrative review summarizes current and emerging applications of AI across lung cancer screening and early detection, imaging-based staging and prognostication, tissue and liquid biopsy-based tumor characterization, treatment planning, surgical and intraoperative guidance, and drug discovery. In imaging, deep learning models have demonstrated high performance in pulmonary nodule detection, risk stratification, and prediction of molecular alterations, while also showing promise in improving screening efficiency and reducing interpretive variability. In pathology and liquid biopsy domains, AI enables prediction of driver mutations, immunotherapy response, and survival outcomes directly from histopathology slides, circulating tumor DNA, and other blood-based biomarkers, facilitating minimally invasive precision oncology approaches. In treatment planning and delivery, AI systems are being developed to support clinical decision-making, surgical planning (through advanced image segmentation and delineation of operative anatomy), and intraoperative navigation through robotic and computer vision-enabled platforms. Despite these advances, significant barriers remain, including limited real-world validation, algorithmic biases, workflow integration issues, and unresolved ethical and legal concerns. Future progress will depend on the development of transparent, clinically validated, and generalizable AI systems that augment rather than replace the expertise of clinical providers and healthcare teams. Active engagement from pulmonologists, oncologists, radiologists, and thoracic surgeons will be essential in guiding safe implementation and ensuring that AI-driven innovations translate into meaningful improvements in patient outcomes. Full article

Background: Osteochondral lesions of the talus (OLT) remain a challenging condition due to the limited regenerative potential of articular cartilage. Conventional bone marrow stimulation (BMS) techniques often result in fibrocartilage formation with inferior biomechanical properties. This study aimed to evaluate the safety and preliminary clinical efficacy of an arthroscopically assisted, single-stage injection of a heparin-conjugated fibrin hydrogel (HCFH) for OLT treatment. Methods: Twelve patients with symptomatic OLT underwent arthroscopic debridement, microfracturing, and HCFH injection containing autologous mesenchymal stromal cells (MSCs) and growth factors. Safety was assessed through systematic monitoring of adverse events (graded according to Common Terminology Criteria for Adverse Events criteria), wound healing, and serial laboratory inflammatory markers (leukocytes, erythrocyte sedimentation rate, C-reactive protein) during early and late follow-up. Clinical outcomes were evaluated using the Visual Analog Scale (VAS) and American Orthopedic Foot and Ankle Society score (AOFAS) preoperatively and at 6 and 12 months. Morphological assessment was performed using magnetic resonance imaging (MRI) with the modified Magnetic Resonance Observation of Cartilage Repair Tissue (MOCART) scoring system, evaluated independently by two blinded musculoskeletal radiologists. Results: No serious adverse events (Grade III–IV) were observed during the 12-month follow-up. All adverse events were mild (Grade I) and self-limited. A transient postoperative elevation in inflammatory markers was observed, returning to clinically acceptable levels by day 14. Significant improvements were noted in pain (VAS decreased from 6.0 to 2.0) and ankle function (AOFAS increased from 70.0 to 90.6) (p < 0.001). MRI demonstrated progressive morphological improvement, with the MOCART score increasing from 34.16 ± 17.1 at 6 months to 75 ± 5.43 at 12 months (p < 0.001). This increase corresponded with imaging features consistent with tissue maturation over time. The favorable MOCART outcomes observed in this study may be explained by the regenerative properties of heparin-conjugated fibrin hydrogels; however, larger randomized controlled trials with longer follow-up are needed to confirm the durability of the regenerated tissue. Interobserver agreement was substantial to almost perfect for MOCART scoring (κ = 0.68–0.84), with perfect agreement observed for surface assessment, bony defect/overgrowth, and cysts. Conclusions: Within the limitations of this study, single-stage HCFH injection demonstrated an acceptable safety profile and favorable preliminary clinical and radiological outcomes at 12 months. These findings suggest potential regenerative capability; however, controlled studies with larger cohorts and longer follow-up are required to determine comparative efficacy and long-term durability. Full article

Background: Basal ganglia calcifications (BGCs) are frequently detected on brain CT scans in older adults, but their clinical relevance for mobility and fall risk is unclear. This study investigated the association of BGCs with impaired mobility and recurrent falls. Methods: In this cross-sectional study, all consecutive patients referred to the mobility clinic of a regional teaching hospital between 2019 and 2021 were included. Mobility was assessed using the Performance-Oriented Mobility Assessment (POMA) for balance, gait and overall mobility, and the Timed Up and Go (TUG) test for functional mobility. All assessments were performed by a trained physiotherapist. Recurrent falls were defined as self-reported occurrence of more than one fall in the past 12 months. Brain CT scans were evaluated for BGCs by a trained senior radiologist and were scored by severity. Univariable and multivariable logistic regression analyses were performed, adjusting for age, sex, and history of cardiovascular events. Results: A total of 253 participants were included (median age 82 years; 58% female), of whom 31% had BGCs. Falls data were available for 246 participants, and 70% reported recurrent falls. In both univariable and multivariable analyses, there was no evidence of a statistically significant association between the presence of BGCs and impaired balance, gait, overall mobility, functional mobility, or recurrent falls. Conclusions: No evidence of a statistically significant association was found between incidentally detected BGCs and impaired mobility or recurrent falls in older adults. Further longitudinal research is needed to confirm these findings and clarify whether BGCs are clinically relevant for mobility and fall risk assessment. Full article

Artificial intelligence (AI) is increasingly being integrated into Clinical Decision-Support Systems (CDSSs), shifting attention from algorithmic performance alone to the broader sociotechnical conditions that shape effective human–AI collaboration. In this study, we investigated whether nine displacement-based structured coordination protocols can improve the collective diagnostic decision-making of hybrid human–AI teams (16 board-certified radiologists and a simulated AI model) in a radiological double-reading task for vertebral fracture detection from X-ray images. Among the protocols tested, the Accuracy-Oriented, Confidence-Oriented, and Presumptuous strategies achieved the highest (balanced) accuracy overall, with up to 97% among strong clinicians and 92% among weak ones, significantly outperforming simpler methods like majority voting. Conversely, approaches optimized for a single metric (e.g., sensitivity or specificity) introduced performance trade-offs. Benefits were strongest among less proficient clinicians, which exhibited substantial and consistent improvements, while proficient clinicians showed limited gains and occasional declines. Critically, Kasparov’s Law emerged as a comparative framework for empirically evaluating coordination quality relative to the diagnostic task, clinical objective, and clinician proficiency by identifying situations in which less proficient clinicians supported by superior coordination protocols outperformed more proficient clinicians operating under inferior ones. These findings demonstrate that coordination design is a critical determinant of hybrid human–AI decision-making, highlighting that a well-structured process can be more relevant than individual components’ performance and support process-centered approaches to the development and evaluation of CDSSs. Full article

Accurate segmentation of the ureter on abdominal computed tomography (CT) remains challenging due to its thin tubular structure and limited expert-annotated training data. While recent deep learning approaches have shown promise on non-contrast CT, arterial-phase imaging remains under-researched. We systematically compared nnU-Net-based configurations for ureter segmentation on arterial-phase CT using 25 radiologist-annotated cases from Seoul St. Mary’s Hospital. Seven training strategies were evaluated with five-fold cross-validation: binary ureter-only segmentation, multi-organ training with anatomical context from eight structures, alternative encoder architectures (ResEncM), specialized loss functions (Tversky, clDice), and a multi-phase fusion architecture. Multi-organ training with Tversky-Focal loss (Config 6) achieved the highest mean Dice of 0.743 ± 0.021 with the best clDice connectivity score (0.800 ± 0.046) and lowest fragmentation (6.56 connected components). Multi-phase fusion yielded a mean Dice of 0.713 on the 12-case subset; a controlled arterial-phase single-channel ablation on the identical 12-case subset achieved 0.721, marginally exceeding the two-channel fusion result (0.713). These findings are scoped to a single-institution exploratory cohort and should be interpreted as internally comparative benchmarking results; they may not generalize to other centres, scanners, or patient populations, and do not constitute clinical validation. Full article

Background: Radiology underpins diagnosis and treatment across pediatrics, yet most artificial intelligence (AI) tools are developed for adults and validated on adult datasets only. Of more than 200 AI systems cleared by the United States (U.S.) Food and Drug Administration (FDA), only about 3% include pediatric validation. Because children differ from adults in anatomy, physiology, pathology, epidemiology, and imaging protocols, adult-trained models often perform sub-optimally in pediatric settings. Methods: A narrative review of peer-reviewed literature from 2000 to 2025 was conducted using PubMed, MEDLINE, Google Scholar, and Scopus. Studies involving AI applications in pediatric X-ray, ultrasound, computed tomography (CT), magnetic resonance imaging (MRI), echocardiography, and point-of-care ultrasound with quantitative performance metrics were included. Findings were synthesized by imaging modality, clinical task, and differences between high-income countries (HICs) and low- and middle-income countries (LMICs). Results: AI demonstrated strong performance across multiple pediatric imaging tasks. In X-ray interpretation, AI detected fractures with area under the curve (AUC) values up to 0.96 (sensitivity, 90.8%; specificity, 88.7%). Pneumonia classification achieved 76.5% accuracy, and foreign body aspiration detection showed 95.3% specificity in HICs. In ultrasound, AI improved junior sonographers’ detection of intussusception (AUC 0.857 to 0.966) and reduced scan time by more than 50%. AI-assisted bone age estimation achieved a mean error of 0.39 years. In echocardiography, AI-derived ejection fraction showed excellent agreement with experts’ interclass correlation coefficient (ICC 0.983), and AI support improved atrioventricular septal defect detection (84.4% to 86.5%). In MRI, the use of AI enhanced lesion detection and supported quantitative analysis. Deep-learning models trained on routine T1- and T2-weighted sequences predicted liver stiffness across multi-site datasets, while advanced neuroimaging pipelines improved the identification of subtle epileptogenic lesions that are often missed on conventional pediatric MRI. However, adult-trained models showed limited generalizability to children. Still, excluding children under the age of two years improved the reading accuracy of pediatric chest X-rays (CXRs) by adult-trained models from 88% to 97%. AI faces challenges beyond the development of age-specific models. Substantial heterogeneity, limited pediatric-specific datasets, and unresolved medicolegal responsibility further restrict adoption worldwide. Challenges are amplified in LMICs, where unstable electricity, limited radiology resources, weak digital infrastructure, and scarce pediatric providers limit implementation. Additionally, many large language models underperform and lack inclusive algorithms suitable for pediatric radiology in many LMICs. Conclusions: AI can enhance diagnostic accuracy, efficiency, and access to pediatric imaging, particularly in resource-limited settings, through task-shifting and decision support. However, it cannot replace pediatric radiologists as of today. Safe adoption requires pediatric-specific model development, standardized validation metrics, diverse datasets that include LMIC populations, stronger digital infrastructure, robust radiologist training in AI capabilities, and the establishment of clear guidelines and medicolegal policies. Full article

Background/Objectives: The trigeminal (CN V) and facial (CN VII) nerves are conventionally taught as separate pathways, yet extensive peripheral anastomoses form sensorimotor plexuses throughout the face. These communications provide the anatomical substrate for proprioception in facial muscles that paradoxically lack muscle spindles and Golgi tendon organs. This review aims to synthesise the anatomical, histological, and clinical evidence on these interconnections and to evaluate their implications across surgery, radiology, neurology, and dentistry. Methods: PubMed/MEDLINE, Scopus, and Google Scholar were searched for cadaveric dissection studies, Sihler whole-mount staining investigations, immunohistochemical analyses, quantitative axonal mapping studies, and clinical case series addressing trigeminal–facial communications and their diagnostic significance. Results: Twenty peripheral anastomoses were systematically identified and mapped, with prevalence ranging from reported-constant in multiple cadaveric series (auriculotemporal–facial trunk; mental–marginal mandibular) to variable (29–86%, depending on trigeminal division and method; V2 by cadaveric dissection, V1 by Sihler staining). Immunohistochemical evidence supports sensorimotor fibre interchange, and recent axonal mapping has revealed that the extracranial facial nerve is a mixed nerve containing motor, sympathetic, and afferent components. Clinically, these anastomoses are implicated in spontaneous facial recovery, trigeminal motor branch transfers, perineural tumour spread, local anaesthesia effects, synkinesis, and Ramsay Hunt syndrome. Conclusions: Available anatomical and histological evidence is consistent with the view that the trigeminal and facial nerves form a functionally integrated unit, though the functional significance of specific communications remains method-dependent. Recognition of these communications is relevant for surgeons, radiologists, neurologists, and dental practitioners managing facial conditions. Full article

Multiple North American and European societies now endorse a combined serological and imaging-based clinical care pathway for non-invasive risk stratification of patients with metabolic dysfunction-associated steatotic liver disease (MASLD). A multidisciplinary group of Canadian radiologists, hepatologists, family physicians, and other health professionals have recently published consensus guidelines for identification and risk stratification of patients with suspected MASLD. Screening should be performed with the FIB-4 score, and those with an indeterminate FIB-4 (between 1.32.67) should undergo imaging-based liver stiffness evaluation either with transient elastography (FibroScan), ultrasound shear wave elastography, or magnetic resonance elastography as a second step. While the implementation of these techniques for measuring liver stiffness differ, there is no clinically significant difference in their diagnostic performance. This narrative review, intended for Family Physicians, summarizes recommendations for serological investigations and imaging modalities of liver steatosis and stiffness. Practical guidance includes an algorithm with thresholds. We discuss current challenges and future directions of risk-stratifying patients with MASLD in the community. Full article

Background/Objectives: Triple-rule-out CT angiography (TRO-CTA) enables simultaneous evaluation of coronary, pulmonary, and aortic causes of acute chest pain, but conventional single-acquisition protocols may compromise vascular enhancement because of conflicting contrast timing requirements. This study evaluated whether a physiology-based sequential helical–axial–helical acquisition strategy could provide consistent tri-territory enhancement in emergency settings. Methods: In this retrospective single-center study, 71 consecutive evaluable emergency department patients (mean age, 66.6 ± 17.0 years; 33 women) with undifferentiated acute chest pain underwent TRO-CTA using a structured sequential protocol (pulmonary, coronary, and aortic phases) guided by individualized test-bolus timing. Objective image quality was assessed using vascular attenuation, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR); subjective image quality was independently graded by two radiologists. Results: Mean vascular attenuation exceeded predefined diagnostic thresholds in all territories (pulmonary 546.7 ± 237.8 HU [95% CI, 490.4–603.0]; coronary 438.8 ± 113.9 HU [95% CI, 411.9–465.8]; aortic 604.3 ± 190.9 HU [95% CI, 559.2–649.5]). Diagnostic interpretability was achieved in all three territories in every technically analyzable examination without repeat contrast-enhanced imaging. Median subjective image-quality scores were 5 (IQR, 4–5) for pulmonary, 4.5 (IQR, 4–5) for coronary, and 4 (IQR, 4–5) for aortic phases; interobserver agreement was good to excellent. Mean total DLP was 461.5 ± 122.5 mGy·cm. Conclusions: A sequential physiology-based TRO-CTA strategy is technically feasible in a tertiary emergency setting and provides consistent tri-territory enhancement. Because this was a single-arm technical validation study, prospective comparative and outcome-based studies are required to confirm its clinical impact. Full article

Background: Aneurysm treatment options are rapidly evolving, as evidenced by the recent introduction and widespread adoption of flow diversion and intrasaccular devices. However, there is a need to understand how these newer technologies are used for difficult-to-treat aneurysms. The main aims of this study were to investigate the variation in aneurysm treatment recommendations among neurosurgeons, interventional radiologists, and interventional neurologists and to generally describe trends in endovascular treatment. Methods: In this survey-based study conducted from June to September 2024, participants were presented with clinical vignettes and asked to choose preferred treatment options, with responses analyzed based on demographic variables including specialty, age, and training prior to and after the introduction of flow diversion. Results: A total of 108 respondents completed the study with a representative mix of specialties—(45 (42.5%) radiologists, 22 (20.8%) neurologists, and 39 (36.8%) neurosurgeons. Sixty-six (61.1%) trained after the introduction of flow diversion. Treatment recommendations were significantly different by specialty (p < 0.001). The Kappa statistic to assess variation in responses showed significant variation in treatment preferences across aneurysm subtypes, ranging from poor (κ = 0.07) to fair (0.31). Treatment of ruptured aneurysms varied by specialty with radiologists opting for stent-assisted coiling at a higher rate than neurologists or neurosurgeons (p < 0.001). There was no significant difference in rates of recommending flow diversion or intrasaccular devices between those who had trained before and after their introduction (p = 0.97). Conclusion: The study highlights the dynamic nature of aneurysm management and considerable variability among different specialties. Further exploration into the rationale for each decision is needed to understand how specialty training affects these decisions. Full article

Background: Chest radiography remains central to post-procedural assessment of central venous catheter (CVC) placement in intensive care units. Multimodal large language models (MLLMs) can process medical images, but their reliability for practical radiography tasks remains uncertain. This study assessed the diagnostic performance of MLLMs and intensivists for CVC access classification, CVC tip assessment, and pneumothorax-related radiographic findings. Methods: In this retrospective diagnostic performance study, consecutive portable anteroposterior chest radiographs obtained after CVC placement in adult critically ill patients were independently evaluated by four intensivists and five MLLMs. A radiologist consensus served as the reference standard. Interobserver agreement and diagnostic performance were assessed using Fleiss’ kappa, Gwet AC1, Cohen’s kappa, accuracy, sensitivity, specificity, precision, F1 score, balanced accuracy, and Matthews correlation coefficient. Results: The final cohort included 183 unique radiographs. Intensivist reviewers showed high performance for CVC access classification but lower and more heterogeneous performance for CVC tip-position assessment. Among MLLMs, CVC access accuracy ranged from 0.339 to 0.874, whereas CVC tip assessment was dominated by almost universal classification of tips as appropriate, with near-zero specificity and chance-level balanced accuracy. For pneumothorax-related findings, all MLLMs classified every case as negative. Intensivist reviewers had higher balanced accuracy than MLLMs for CVC access classification (difference, 0.420; 95% CI, 0.349–0.490; p < 0.001) and CVC tip assessment (difference, 0.247; 95% CI, 0.205–0.290; p < 0.001). Pneumothorax analyses were exploratory because only five positive cases were present. Conclusions: The evaluated MLLMs showed unreliable diagnostic performance compared with experienced intensivists. Apparent performance was influenced by class imbalance and dominant-response behavior, supporting cautious task-specific validation and complete diagnostic performance reporting. Full article

Artificial intelligence (AI) is reshaping ultrasound diagnosis by converting operator-dependent grayscale, Doppler, elastography, contrast-enhanced, automated-volume, and video data into reproducible decision support. In breast ultrasound, the most mature evidence involves benign–malignant lesion classification, BI-RADS risk stratification, reduction in unnecessary biopsy in selected low-risk lesions, assistance for less experienced readers, automated breast volume scanning, video-based assessment, axillary staging, and prediction of biologic markers such as molecular subtype, HER2 status, Ki-67 expression, lymphovascular invasion, and nodal metastasis. AI does not replace sonographers, radiologists, pathologists, or clinical judgment; rather, it can standardize feature extraction, prompt second-reader review, quantify uncertainty, and integrate imaging with clinical context. This review summarizes current clinical applications of AI in ultrasound diagnosis, which has a strong recent multicenter evidence base. It also discusses implementation requirements, including standardized acquisition, external validation, calibration, imaging–pathology concordance, workflow integration, data security, and equity across scanners and patient populations. Full article

Background: Treatment monitoring in pulmonary tuberculosis increasingly requires assessment of residual inflammatory burden and structural lung damage beyond microbiologic response alone. High-resolution computed tomography (HRCT) can provide this information, but interpretation of serial examinations is time-consuming and partly subjective. This study did not aim to evaluate AI for the diagnosis of pulmonary tuberculosis. Instead, it explored whether artificial intelligence (AI)-assisted quantitative HRCT analysis could support longitudinal assessment of treatment-related imaging changes in patients with microbiologically confirmed pulmonary tuberculosis. Methods: We conducted a retrospective, single-center, exploratory longitudinal study of patients receiving treatment for pulmonary tuberculosis. HRCT examinations acquired at diagnosis and during follow-up were anonymized, reviewed by an expert thoracic radiologist, and processed using AVIEW Lung Texture (Coreline Soft v2.0). The software quantified total lung volume and six predefined parenchymal categories: normal lung, ground-glass opacity, consolidation, reticulation, honeycombing, and emphysema. Results: Ninety-six patients contributed 256 HRCT examinations. The most frequent software-detected abnormalities were ground-glass opacity, consolidation, and emphysema-labeled low-attenuation areas. Ground-glass opacity and consolidation showed the clearest decline across serial examinations, consistent with regression of active inflammatory disease during treatment. Reticulation showed a heterogeneous course, likely reflecting both inflammatory resolution and residual structural remodeling. Honeycombing was infrequent and quantitatively limited. Lung volume changed variably and did not consistently parallel visual improvement. A key methodological limitation was the absence of a dedicated cavity class. As a result, emphysema-labeled low-attenuation areas should not be interpreted as conventional emphysema alone, because tuberculous cavities and post-destructive abnormalities were frequently included in this category. Conclusions: AI-assisted HRCT quantification may support longitudinal assessment of pulmonary tuberculosis by providing structured and reproducible measures of interval change. However, tuberculosis-specific interpretation remains dependent on expert radiologic oversight, particularly in cavitary disease. Full article