AI Model Tells Gastric Tumors Apart Before Surgery

A new multimodal nomogram distinguishes gastrointestinal stromal tumors from leiomyomas in lesions under 2 cm, according to a retrospective study.

Updated on Nov. 12, 2026 in Cancer

AI Model Tells Gastric Tumors Apart Before Surgery

Researchers have developed a multimodal AI nomogram that differentiates gastric GISTs from leiomyomas in submucosal lesions smaller than 2 cm. The study, published as accepted early-access research on November 12, 2026, analyzed 408 patients.

Why it matters

GISTs and leiomyomas carry divergent malignant potential, so accurate preoperative identification shapes treatment decisions. Conventional endoscopic ultrasonography has inherent limitations for lesions this small.

The retrospective study of 408 patients, including 198 GISTs and 210 leiomyomas, found the combined nomogram reached a test AUC of 0.901 versus 0.724 for clinical factors alone. Whether these results hold in prospective or multicenter settings remains unassessed.

The players

960th Hospital of Chinese PLA Joint Logistics Support Force

A military-affiliated hospital in China whose Institutional Review Board approved the study under Approval No. 2026-078.

The details

The team built a deep learning-radiomics signature by fusing compressed MobileNet_v2 deep features with handcrafted radiomics features, then combined it with independently identified clinical factors. The nomogram was significantly superior to the standalone signatures by DeLong test (all p < 0.05), and the study was approved by the Institutional Review Board of the 960th Hospital of Chinese PLA Joint Logistics Support Force.

Timeline

  1. • November 12, 2026: Study published as accepted early-access research.

Health Landscape

The approach builds on endoscopic ultrasonography, the conventional tool for gastric submucosal lesions, which the source describes as having inherent limitations for lesions under 2 cm. It offers a radiomics-and-deep-learning alternative aimed at that diagnostic gap.

If validated in larger prospective studies, the nomogram could help clinicians decide which small gastric lesions need invasive procedures. Patients may face fewer unnecessary operations if the tool proves reliable in routine practice.

The takeaway

Preoperative imaging tools that improve on endoscopic ultrasonography could reduce unnecessary surgery for small gastric tumors. Clinicians and patients should watch for prospective validation before expecting changes in routine care.

Further reading

Explore more on Cancer research and diagnostic advances.