Advances in AI — October 2025

The integration of artificial intelligence (AI) into radiology continues to evolve, demonstrating utility in tumor characterization, acute triage and long-term patient management.

Recent studies published in October and FDA clearances for emerging software highlight the role of deep learning models as sophisticated adjunctive tools designed to enhance diagnostic efficiency and potentially improve adherence to patient follow-up protocols.

Can AI Enhance Tumor Characterization?

In thoracic imaging, AI-enhanced radiomics are being utilized to differentiate subtypes of pulmonary nodules on chest computed tomography (CT). In a retrospective study, researchers found that a deep learning AI platform (uAI-Chest-Care, V1.0, United Imaging), which incorporated 3D morphometrics and CT attenuation metrics, achieved a high area under the curve (AUC) for detecting invasive adenocarcinoma (IAC) at 93.6 percent and atypical adenomatous hyperplasia and adenocarcinoma in situ (AAH + AIS) at 88.4 percent.^1,2

The study authors suggested these findings may represent an imaging continuum from AAH + AIS to IAC, marked by progressive increases in CT attenuation and solid components. However, performance was notably reduced for minimally invasive adenocarcinoma (MIA) with an AUC of 70.7 percent, indicating a “gray zone” where overlapping radiomic features necessitate the future integration of clinical and longitudinal CT data for improved accuracy.^1,2

A meta-analysis comparing MRI-based deep learning (DL) to unassisted radiologist interpretation for detecting lymph node metastasis in colorectal cancer found that DL models exhibited significantly higher diagnostic performance. In internal validation cohorts, DL achieved a pooled sensitivity of 89 percent and an AUC of 93 percent in comparison to 65 percent sensitivity and 76 percent AUC for unassisted radiologists.^3,4

While DL showed higher overall sensitivity, its specificity was not statistically different from that of senior radiologists, suggesting its primary role may be as a high-performance triage or concurrent reader to prioritize suspicious cases and streamline workflow.^3,4

Furthermore, AI models are expanding beyond detection capability in screening. Research presented on digital breast tomosynthesis (DBT)-based AI demonstrated high sensitivity (93 percent) and also suggested the AI score could serve as an imaging biomarker. In a recent interview with Diagnostic Imaging, lead study author Manisha Bahl, M.D., noted that higher AI scores were significantly associated with higher histologic grades and positive lymph node status with lymph node-positive tumors averaging a score of 72 out of 100 in comparison to 60 for lymph node-negative tumors.⁵

Can Emerging AI Advances Bolster Acute Care Triage?

The Food and Drug Administration (FDA) granted 510(k) clearance for the qER-CTA software (Qure.ai) to provide adjunctive AI assessment for possible large vessel occlusions (LVOs) on CT angiography (CTA), specifically targeting the internal carotid artery (ICA) and the M1 segment of the middle cerebral artery (MCA).⁶

Ajith Thomas, M.D., the chairman of the Department of Neurosurgery with Cooper University Healthcare in Camden, N.J., suggested that the capability of the qER-CTA software may facilitate rapid triage and timely notification of specialists.⁶

In other FDA CT news, the FDA issued a breakthrough device designation for an AI-powered multi-triage system from Aidoc. Employing a clinical AI foundation model (Clinical AI Reasoning Engine), the multi-triage platform reportedly detects various conditions on CT within a single workflow, potentially mitigating workflow strain caused by rising imaging volumes and staff shortages.⁷

Can AI Facilitate Improved Follow-Up in the Management of Patients with Aortic Abdominal Aneurysm?

AI also demonstrates utility in improving long-term patient follow-up. A retrospective study evaluating CT-based AI in the detection and management of aortic abdominal aneurysms (AAA) found that adjunctive AI use resulted in a nearly fourfold reduction in the AAA evaluation timeline (from 83 days to 22 days). ^8,9

The AI integration also improved follow-up adherence, increasing the percentage of scheduled long-term monitoring appointments from 65 percent to 99 percent compared to unassisted CT interpretation. For patients with AAA > 5 cm, the time from imaging to surgical repair was reduced from 270 days to 58 days in the AI-assisted cohort.^8,9

References

1. Zhang H, Liu K, Ding Y, et al. Predictive value of artificial intelligence-based quantitative CT feature analysis for diagnosing the pathological types of pulmonary nodules. Eur Radiol. 2025 Oct 11. doi: 10.1007/s00330-025-12050-w. Online ahead of print.

2. Hall J. Can AI-enhanced radiomics improve differentiation of pulmonary nodules on chest CT? Diagn Imaging. Available at: https://www.diagnosticimaging.com/view/ai-enhanced-radiomics-differentiation-pulmonary-nodules-chest-ct- . Published October 13, 2025. Accessed October 25, 2025.

3. Wang F, Deng W, Zhang Z. et al. MRI-based deep learning algorithms vs. radiologists for lymph node metastasis in colorectal cancer: a systematic review and meta-analysis. Acad Radiol. 2025 Oct 16:S1076-6332(25)00954-7. doi: 10.1016/j.acra.2025.09.048. Online ahead of print.

4. Hall J. MRI-based deep learning for lymph node metastasis detection in colorectal cancer: what a new meta-analysis reveals. Diagn Imaging. Available at: https://www.diagnosticimaging.com/view/mri-based-deep-learning-lymph-node-metastasis-detection-colorectal-cancer-meta-analysis- . Published October 21, 2025. Accessed October 25, 2025.

5. Hall J. Mammography research: can adjunctive AI for DBT provide insight into tumor biology with breast cancer? Diagn Imaging. Published October 6, 2025. Accessed

6. Hall J. FDA clears AI software for large vessel occlusion detection on CT angiography scans. Diag Imaging. Available at: https://www.diagnosticimaging.com/view/fda-ai-software-large-vessel-occlusion-detection-ct-angiography- . Published October 22, 2025. Accessed October 25, 2025.

7. Hall J. FDA issues breakthrough device designation for AI-powered multi-triage system for computed tomography. Diagn Imaging. Available at: https://www.diagnosticimaging.com/view/fda-breakthrough-device-designation-ai-powered-multi-triage-system-computed-tomography . Published October 1, 2025. Accessed October 25, 2025.

8. Kostiuk V, Rodriguz PP, Loh SA, et al. Artificial intelligence-based algorithms improve care of patients with AAA. Ann Vasc Surg. 2025 Oct 6:123:207-214. doi: 10.1016/j.avsg.2025.09.046. Online ahead of print.

9. Hall J. Study suggests benefits of CT-based AI in detecting and managing patients with aortic abdominal aneurysm. Diagn Imaging. Available at: https://www.diagnosticimaging.com/view/study-benefits-ct-based-ai-detecting-managing-patients-with-aortic-abdominal-aneurysm . Published October 9, 2025. Accessed October 25, 2025.

(Editor’s note: This article was written with artificial intelligence and reviewed by human editors by accuracy.)