Journal of Radiology and Clinical Research

Background: Early detection of pulmonary nodules is crucial for lung cancer outcomes, but subtle nodules often go undetected in routine radiological evaluations. This study evaluates a deep learning algorithm's effectiveness in detecting subtle pulmonary nodules on chest CT scans compared to radiologist interpretation. Material and Methods: In this retrospective study, we analyzed 1,200 chest CT scans containing pulmonary nodules ≤8mm. A deep learning algorithm was developed and compared against junior (<5 years) and senior (>10 years) radiologists. Performance metrics included sensitivity, specificity, diagnostic accuracy, and time efficiency. Findings The AI model achieved 91.7% sensitivity and 88.3% specificity (AUC-ROC: 0.93), outperforming junior radiologists (sensitivity: 83.3%, specificity: 85.0%) and matching senior radiologists (sensitivity: 89.2%, specificity: 86.7%). Average interpretation time was significantly reduced with AI (1.2 minutes) compared to junior (4.8 minutes) and senior radiologists (3.5 minutes). However, the AI model showed decreased performance with very low-contrast nodules (sensitivity: 72.5%) compared to senior radiologists (sensitivity: 81.4%). The deep learning algorithm demonstrates promising results in pulmonary nodule detection, particularly in reducing diagnostic time and variability. While it shows comparable performance to experienced radiologists in most scenarios, challenges remain in detecting very low-contrast nodules, suggesting its optimal use as a complementary tool rather than a replacement for radiologist expertise.

Abstract: Background: Pancreatic cancer remains one of the most fatal malignancies, where early identification is crucial for better patient prognosis. Advanced imaging methods such as multiphase computed tomography (CT) and magnetic resonance imaging (MRI) are fundamental in detecting and staging pancreatic tumours. However, their comparative effectiveness, especially for detecting small neoplasms and borderline resectable tumours, warrants further study. This research aims to assess and contrast the diagnostic accuracy of MRI and CT in identifying pancreatic abnormalities, determining vascular involvement, and evaluating tumour resectability to facilitate clinical decision-making. Material and Methods: A retrospective analysis was performed on 95 patients who had undergone both MRI and CT scans due to suspected pancreatic cancer. Diagnostic parameters such as sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were examined to compare lesion detection and characterization. Further analysis was conducted on small tumours (<2 cm) and cases with borderline resectability. Statistical evaluation included McNemar’s and chi-square tests. Findings MRI exhibited a greater sensitivity (92.6%) compared to CT (87.4%) in identifying pancreatic lesions, particularly for smaller tumors (<2 cm), where MRI had a significantly higher sensitivity (88.2% vs. 73.5%; p = 0.014). When assessing vascular involvement in borderline resectable tumours, MRI achieved an accuracy of 94.7%, whereas CT reached 89.5%. Although CT demonstrated a slightly superior specificity (85.3% vs. 82.1% for MRI), MRI provided enhanced lesion characterization due to better soft tissue contrast. Both imaging modalities encountered difficulties in identifying small hepatic and peritoneal metastases, although MRI performed marginally better in these instances. The findings of this study underscore MRI’s superior diagnostic efficiency over CT in detecting and characterizing pancreatic malignancies. MRI’s heightened sensitivity and detailed visualization of small lesions and vascular involvement establish it as a preferred imaging technique for presurgical assessments. These findings advocate for the incorporation of MRI into routine diagnostic protocols to enhance patient management and outcomes

Background: Virtual and augmented reality (VR/AR) technologies have emerged as promising tools in radiology training, offering immersive, interactive learning experiences. While traditional teaching methods rely on static 2D images and didactic lectures, VR/AR enables spatial visualization, real-time feedback, and simulated procedural practice. This study evaluates the effectiveness of VR/AR-assisted training in radiology compared to traditional methods, focusing on diagnostic accuracy, procedural competency, knowledge retention, and trainee engagement. This is study to assess whether VR/AR training improves diagnostic interpretation, procedural skill acquisition, and trainee engagement more effectively than conventional training methods in radiology education. Material and Methods: This prospective, comparative study was conducted over 12 months at a tertiary medical institution, enrolling 80 radiology trainees. Participants were randomized into VR/AR-based (n=40) and traditional (n=40) training groups. Training outcomes were assessed through pre- and post-training diagnostic accuracy tests, procedural competency evaluations, and engagement surveys. Statistical analyses included paired and independent t-tests, ANOVA, and regression modelling to evaluate performance trends and subgroup differences. Findings Diagnostic accuracy improved significantly in the VR/AR group (64.2% to 87.3%, p < 0.001) compared to the traditional group (63.9% to 79.1%, p < 0.001). Procedural competency scores (10-point scale) were higher post-training in the VR/AR group (8.4 vs. 6.9, p < 0.001). VR/AR trainees exhibited steeper learning curves (final accuracy: 88% vs. 80%, p < 0.01). Engagement scores were significantly higher in the VR/AR group (4.7/5 vs. 3.8/5, p < 0.01). VR/AR-assisted training significantly enhances diagnostic accuracy, procedural proficiency, and trainee engagement, making it a valuable addition to radiology education. The study supports integrating VR/AR modules into curricula, particularly for junior trainees who demonstrated the greatest benefit.

Background: Coronary artery disease (CAD) is a leading cause of morbidity and mortality, necessitating accurate, early diagnosis. Invasive coronary angiography (ICA) remains the gold standard for detecting significant stenosis, but it carries procedural risks. Coronary computed tomography angiography (CCTA) has emerged as a non-invasive alternative, providing high-resolution imaging with lower risk. This study evaluates the diagnostic accuracy, clinical utility, and safety of CCTA compared to ICA. Material and Methods: A prospective observational study was conducted over 24 months at a tertiary care center. A total of 450 patients with suspected stable CAD underwent CCTA, followed by ICA if clinically indicated. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of CCTA were assessed using ICA as the reference standard. Secondary outcomes included plaque characterization, radiation exposure, contrast-induced nephropathy, and major adverse cardiovascular events (MACE) at six-month follow-up. Statistical analyses included receiver operating characteristic (ROC) curve analysis, inter-reader agreement using Kappa statistics, and Kaplan-Meier survival analysis. Findings CCTA demonstrated a sensitivity of 94.6%, specificity of 87.3%, PPV of 93.4%, and NPV of 89.1%, with an overall accuracy of 91.3% for detecting significant stenosis (≥50%). The ROC curve analysis yielded an AUC of 0.91, confirming excellent diagnostic performance. Radiation exposure was significantly lower in the CCTA group (5.2 ± 1.3 mSv) compared to ICA (7.8 ± 2.1 mSv, p < 0.01). MACE rates at six months were similar between groups (6.9% for CCTA vs. 8.9% for ICA, p = 0.58), demonstrating comparable long-term clinical outcomes. CCTA is a reliable non-invasive alternative to ICA, with high sensitivity, reduced radiation exposure, and comparable clinical outcomes. While CCTA may slightly overestimate stenosis in heavily calcified arteries, advancements in AI-assisted interpretation and FFR-CT may further refine its diagnostic accuracy. Given these findings, CCTA should be considered a first-line imaging modality for stable CAD evaluation, particularly in intermediate-risk patients