Pooja Jagmohan

Background: Breast cancer remains the most common cancer among women globally. Mammography is a key diagnostic modality; however, interpretation is increasingly challenged by rising imaging volumes, a global shortage of breast radiologists, and variability in reader experience. Artificial intelligence (AI) has been proposed as a potential adjunct to address these issues, particularly in settings with high breast density, such as Asian populations. This study aimed to evaluate the impact of AI assistance on mammographic diagnostic performance among resident and consultant radiologists in Singapore. Objective: To assess whether AI assistance improves diagnostic accuracy in mammographic breast cancer detection across radiologists with varying levels of experience. Methods: A multi-reader, multi-case study was conducted at the National University Hospital, Singapore, from May to August 2023. De-identified digital mammograms from 500 women (250 with cancer and 250 normal or benign) were interpreted by 17 radiologists (4 consultants, 4 senior residents, and 9 junior residents). Each radiologist read all cases over 2 reading sessions: one without AI assistance and another with AI assistance, separated by a 1-month washout period. The AI system (FxMammo) provided heatmaps and malignancy risk scores to support decision-making. Area under the curve of the receiver operating characteristic (AUROC) was used to evaluate diagnostic performance. Results: Among the 500 cases, 250 were malignant and 250 were non-malignant. Of the malignant cases, 16%(80/500) were ductal carcinoma in situ and 84%(420/500) were invasive cancers. Among non-malignant cases, 69.2%(346/500) were normal, 17.6%(88) benign, and 3.6%(18/500) possibly benign but stable on follow-up. Masses (54.4%, 272) and calcifications (10.8%, 54/500) were the most common findings in cancer cases. A majority of both malignant (66.8%, 334/500) and non-malignant (68%, 340/500) cases had heterogeneously or extremely dense breasts (BIRADS [Breast Imaging Reporting and Data System] categories C and D). The AI model achieved an AUROC of 0.93 (95% CI 0.91‐0.95), slightly higher than consultant radiologists (AUROC 0.90, 95% CI 0.89‐0.92; P =.21). With AI assistance, AUROC improved among junior residents (from 0.84 to 0.86; P =.38) and senior residents (from 0.85 to 0.88; P =.13), with senior residents approaching consultant-level performance (AUROC difference 0.02; P =.051). Diagnostic gains with AI were greatest in women with dense breasts and among less experienced radiologists. AI also improved inter-reader agreement and time efficiency, particularly in benign or normal cases. Conclusions: This is the first study in Asia to evaluate AI assistance in mammography interpretation by radiologists of varying experience. AI significantly improved diagnostic performance and efficiency among residents, helping to narrow the experience-performance gap without compromising specificity. These findings suggest a role for AI in enhancing diagnostic consistency, improving workflow, and supporting training. Integration into clinical and educational settings may offer scalable benefits, though careful attention to threshold calibration, feedback loops, and real-world validation remains essential. Further studies in routine screening settings are needed to confirm generalizability and cost-effectiveness.

Background/Objectives: Artificial intelligence (AI) is increasingly applied in breast imaging, with potential to improve diagnostic accuracy and reduce workload in mammographic breast cancer screening. However, real-world integration of AI into national screening programs remains limited, and little is known about radiologists’ perspectives in Asian settings. This study aimed to explore radiologists’ perceptions of AI adoption in Singapore’s breast screening program, focusing on perceived benefits, barriers, and requirements for safe integration. Methods: We conducted a mixed methods study involving a cross-sectional survey of 17 radiologists with prior experience using AI-assisted mammography, followed by semi-structured interviews with 10 radiologists across all three public healthcare clusters. The survey measured confidence in AI, attitudes toward its diagnostic role, and integration preferences. Interviews were analyzed thematically, guided by the Unified Theory of Acceptance and Use of Technology (UTAUT) framework. Results: Among survey respondents, 64.7% recommended AI as a companion reader, though only 29.4% rated its performance as comparable to humans. Confidence was highest when AI was validated on local datasets (mean 9.3/10). Interviews highlighted AI’s strengths in routine, fatigue-prone tasks, but skepticism for complex cases. Concerns included false positives, workflow inefficiencies, medico-legal accountability, and long-term costs. Radiologists emphasized the importance of national guidelines, local validation, and clear role definition to build trust. Conclusions: Radiologists support AI as an adjunct to, but not a replacement for, human readers in breast cancer screening. Adoption will require robust regulatory frameworks, seamless workflow integration, transparent validation on local data, and structured user training to ensure safe and effective implementation.

Background: Early diagnosis of metastatic epidural spinal cord compression (MESCC) is vital to expedite therapy and prevent paralysis. Staging CT is performed routinely in cancer patients and presents an opportunity for earlier diagnosis. Methods: This retrospective study included 123 CT scans from 101 patients who underwent spine MRI within 30 days, excluding 549 CT scans from 216 patients due to CT performed post-MRI, non-contrast CT, or a gap greater than 30 days between modalities. Reference standard MESCC gradings on CT were provided in consensus via two spine radiologists (11 and 7 years of experience) analyzing the MRI scans. CT scans were labeled using the original reports and by three radiologists (3, 13, and 14 years of experience) using dedicated CT windowing. Results: For normal/none versus low/high-grade MESCC per CT scan, all radiologists demonstrated almost perfect agreement with kappa values ranging from 0.866 (95% CI 0.787–0.945) to 0.947 (95% CI 0.899–0.995), compared to slight agreement for the reports (kappa = 0.095, 95%CI −0.098–0.287). Radiologists also showed high sensitivities ranging from 91.51 (95% CI 84.49–96.04) to 98.11 (95% CI 93.35–99.77), compared to 44.34 (95% CI 34.69–54.31) for the reports. Conclusion: Dedicated radiologist review for MESCC on CT showed high interobserver agreement and sensitivity compared to the current standard of care.

A deep learning model trained for pneumothorax detection with a large dataset of chest radiographs generalized well to multiple external datasets that were independent of the training data source. Purpose To assess the generalizability of a deep learning pneumothorax detection model on datasets from multiple external institutions and examine patient and acquisition factors that might influen...

Mammography is extensively used for breast cancer screening but has high false-positive rates. Here, prospectively collected blood samples were used to identify circulating microRNA (miRNA) biomarkers to discriminate between malignant and benign breast lesions among women with abnormal mammograms. The Discovery cohort comprised 72 patients with breast cancer and 197 patients with benign breast lesions, while the Validation cohort had 73 and 196 cancer and benign cases, respectively. Absolute expression levels of 324 miRNAs were determined using RT-qPCR. miRNA biomarker panels were identified by: (1) determining differential expression between malignant and benign breast lesions, (2) focusing on top differentially expressed miRNAs, and (3) building panels from an unbiased search among all expressed miRNAs. Two-fold cross-validation incorporating a feature selection algorithm and logistic regression was performed. A six-miRNA biomarker panel identified by the third strategy, had an area under the curve (AUC) of 0.785 and 0.774 in the Discovery and Validation cohorts, respectively, and an AUC of 0.881 when differentiating between cases versus those with benign lesions or healthy individuals with normal mammograms. Biomarker panel scores increased with tumor size, stage and number of lymph nodes involved. Our work demonstrates that circulating miRNA signatures can potentially be used with mammography to differentiate between patients with malignant and benign breast lesions.

Objective: The aim of this study was to externally validate the feasibility and robustness of a risk-stratification score for B3 lesions based on clinical, pathological, and radiological data for improved clinical decision making. Methods: 129 consecutive histologically confirmed B3 lesions diagnosed at ultrasound-guided biopsy at our institution were included in this retrospective study. Patient- and lesion-related variables were independently assessed by two blinded breast radiologists (R1, R2), by assigning each feature a score from 0 to 2 (maximum sum-score of 5). Sensitivity, specificity, positive and negative predictive values were calculated at two different thresholds (≥1 and 2). Categorical variables were compared using Chi-squared and Fisher exact tests. The diagnostic accuracy of the score to distinguish benign from malignant B3 lesions was assessed by receiver operating characteristic (ROC) analysis. Results: Surgery was performed on 117/129 (90.6%) lesions and 11 of these 117 (9.4%) lesions were malignant. No cancers were found at follow-up of at least 24 months. Area under the ROC-curve was 0.736 (R1) to 0.747 (R2), with no significant difference between the two readers (p = 0.5015). Using a threshold of ≥1, a sensitivity, specificity, PPV, and NPV of 90%/90% (R1/R2), 39%/38% (R1/R2), 11%/12% (R1/R2) and 97%/98% (R1/R2) were identified. Both readers classified 47 lesions with a score ≤1 (low risk of associated malignancy). Of these, only one malignant lesion was underdiagnosed (Ductal carcinoma in situ-G1). Conclusions: In our external validation, the score showed a high negative predictive value and has the potential to reduce unnecessary surgeries or re-biopsies for ultrasound-detected B3-lesions by up to 39%.