Artificial Intelligence Advances in Interventional Oncology: Year in Review

Artificial intelligence (AI) is rapidly moving from research environments into the interventional oncology (IO) suite, particularly in liver-directed therapies. Recent prospective trials and multicenter validation studies published within the past year suggest AI may improve ablation precision, standardize procedural planning, and enhance prognostic modeling.¹^-⁴ Together, these developments signal a shift toward real-time, AI-supported decision-making in image-guided cancer care.

In a randomized phase 2 trial published in the May 2025 issue of The Lancet Gastroenterology & Hepatology, Odisio et al evaluated AI-enabled software for minimal ablative margin (MAM) assessment during computed tomography (CT)-guided thermal ablation of liver tumors.¹ The COVER-ALL trial compared software-based margin assessment using deformable image registration and automated segmentation with conventional visual assessment. Patients in the AI-assisted arm achieved significantly improved minimal margin coverage without increased complications, demonstrating procedural safety and technical superiority.¹

In Cell Reports Medicine, Balaguer-Montero et al reported development of a fully automated deep learning tool trained on 1598 CT scans comprising 4908 liver tumors.² The system demonstrated strong external validation performance for lesion detection and segmentation, at times even exceeding the assessments of expert radiologists and state-of-the-art models. Authors suggested the tool may support longitudinal monitoring and interventional planning.²

Additionally, AI is influencing procedural decision support. In a 2025 prospective multicenter cohort study, Ding et al evaluated an AI-based microwave ablation planning system (MWA-PS) for hepatocellular carcinoma (HCC).³ The system generates ablation parameter recommendations based on imaging and clinical inputs with the aim to equalize operator performance regardless of experience level. Among less-experienced operators, AI assistance was associated with local tumor progression rates comparable to experienced operators.³

Beyond procedural guidance, machine learning-driven radiomics models have shown promise for outcome prediction, specifically in patients with HCC. In a retrospective study of 108 patients with transarterial chemoembolization (TACE)-resistant HCC, a machine learning model accurately predicted prognosis after continued TACE, outperforming clinical- or radiomics-only models.⁴ Another study showed a deep-learning model successfully predicting early recurrence rates after thermal ablation in this population.⁵

For IOs, AI applications are converging on 3 high-impact domains: automated tumor segmentation, intraprocedural quality assurance, and predictive analytics. Standardized margin assessment addresses one of the most critical determinants of local tumor control following ablation.¹ Improved reproducibility could translate into lower recurrence rates and more consistent outcomes across centers.

Automated CT-based segmentation may reduce workflow burden and interobserver variability in tumor measurement, facilitating more precise treatment planning and response assessment.² Meanwhile, AI-guided procedural planning tools may help mitigate experience-dependent variability.³

Importantly, regulatory and reporting frameworks are evolving in parallel. The Society of Interventional Radiology’s iCARE checklist, published in Journal of Vascular and Interventional Radiology in 2025, outlines standards for dataset transparency, bias mitigation, and validation in AI research.⁶ The FDA’s 2025 guidance on a “Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions” further clarifies pathways for updating AI-enabled device software while maintaining safety oversight.⁷ Together, these measures aim to ensure responsible clinical integration.

As AI advances across the medical landscape, clinicians must match its technical progress with their own expertise to ensure high-quality care—a shift interventional radiologists (IRs) are learning to embrace. Upon attending a 2024 conference and surveying its participants, Dr Judy Wawira Gichoya, an associate professor of IR at Emory University, reported the general IR attitude toward robotics as “indifference.” ⁸ In a 2025 article for Interventional News, she noted that many AI products are not created with the interventional side of radiology in mind; although IRs use these tools, they often do so in ancillary rather than central roles. Now, as the integration of AI into all aspects of radiology is impossible to ignore, IRs must take action to ensure it is implemented responsibly and effectively. Technology will never cease to advance, but for purposeful progress to be achieved, it requires “the input of the interventional radiologist to not continue to develop monolithic applications in isolation, but envision an enabled future that allows us to provide excellent care to all patients.”⁸

Emerging evidence from randomized trials and prospective studies indicates that AI in IO is transitioning from experimental validation to procedural augmentation. As regulatory guidance and reporting standards mature, AI-assisted therapies may become integral to precision cancer treatment, so long as operators and developers work towards the same goal of patient-centered care.

1. Odisio BC, Albuquerque J, Lin YM, et al. Software-based versus visual assessment of the minimal ablative margin in patients with liver tumours undergoing percutaneous thermal ablation (COVER-ALL): a randomised phase 2 trial. Lancet Gastroenterol Hepatol. 2025;10(5):442-451. doi:10.1016/S2468-1253(25)00024-X
2. Balaguer-Montero M, Marcos Morales A, Ligero M, et al. A CT-based deep learning-driven tool for automatic liver tumor detection and delineation in patients with cancer. Cell Rep Med. 2025;6(4):102032. doi:10.1016/j.xcrm.2025.102032
3. Ding W, Zheng L, Wei H, et al. Artificial intelligence system assists doctors to improve the treatment effect of microwave ablation on hepatocellular carcinoma: a prospective cohort study. Digital Engineering. 2025;6:100054. doi:10.1016/j.dte.2025.100054
4. Li H, Fan Y, Li Y, Ren W. Artificial intelligence-driven CT radiomics model predicts prognosis in TACE-refractory hepatocellular carcinoma. Abdom Radiol (NY). 2025. doi:10.1007/s00261-025-05285-0
5. Kong Q, Li K. Predicting early recurrence of hepatocellular carcinoma after thermal ablation based on longitudinal MRI with a deep learning approach. Oncologist. 2025;30(3):oyaf013. doi:10.1093/oncolo/oyaf013
6. Anibal JT, Huth HB, Boeken T, et al. Interventional radiology reporting standards and checklist for artificial intelligence research evaluation (iCARE). J Vasc Interv Radiol. 2025;36(9):1381-1388.e4. doi: 10.1016/j.jvir.2024.12.585
7. US Food and Drug Administration. Marketing submission recommendations for a predetermined change control plan for artificial intelligence-enabled device software functions. August 2025. Accessed February 18, 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/marketing-submission-recommendations-predetermined-change-control-plan-artificial-intelligence
8. Gichoya JW. Are we there yet? Computational advancements in interventional radiology. Interventional News. Apr 10, 2025. Accessed February 18, 2026. https://interventionalnews.com/are-we-there-yet-computational-advancements-in-interventional-radiology/