Developed a Surgical-Specific Generative AI that Brings "Intelligence" to the Operating Room

As part of "GENIAC (Generative AI Accelerator Challenge)," a project implemented by NEDO and the Ministry of Economy, Trade and Industry aimed at strengthening generative AI development capabilities, Direava Inc. has been working on "Development of a Vision-Language Integrated AI Foundation Model for Surgical Support" (hereinafter, "this project"). We have now developed Japan's first surgical-specific generative AI (Surgical VLM) that understands and interacts with surgical situations in real-time.

Conventional generative AI used in the surgical field has been limited to image recognition (the "eye") for assisting in the recognition of patient organs and affected areas. The surgical-specific generative AI developed this time is an innovative system that understands "what to do next" from the surgical situation, much like a "doctor's brain," and engages in advanced dialogue with surgeons. On February 20, 2026, a demonstration test was conducted during an actual gastric cancer surgery at Keio University Hospital, successfully confirming its usefulness by meeting the required standards for medical education (anatomical accuracy, clinical utility, and textual fluency).

Recently, fostering young talent who can inherit advanced medical technologies has become a challenge in the medical field. Moving forward, Direava will fully initiate preparations for commercialization (service launch) by 2026 and will also proceed with demonstrations for surgeries other than gastric cancer. This will strongly support the training of surgeons and advanced surgical education, contributing to future improvements in medical quality and ensuring safety.

1. Background

Since fiscal year 2024, NEDO has been promoting the GENIAC project with the Ministry of Economy, Trade and Industry to accelerate research and development and support for generative AI. GENIAC provides support for the development of various domain-specific foundation models, from general-purpose foundation models, and as of the end of March 2026, a total of 53 companies have been developing generative AI foundation models.

Direava, which achieved these results, is one of these companies and a medical startup originating from Keio University School of Medicine. To date, it has developed medical AI and systems utilizing it, including obtaining manufacturing and sales approval from the Minister of Health, Labour and Welfare on December 9, 2025, for its AI-powered programmed medical device "Surgical Video Recognition Program Kinosura" ※1.

The company was selected for GENIAC in July 2025. For AI development in the surgical field, it developed Japan's first ※2 surgical-specific generative AI, "Surgical VLM," implementing unprecedented functions by training its self-developed foundation model with a large dataset of intraoperative images and natural language captions. The results were demonstrated in an actual surgical setting. For the evaluation of the developed AI, with the cooperation of Keio University Hospital, items considered important not only for accurate understanding of surgical situations but also for the training of surgeons were established.

Figure: Evaluation items and performance of the developed surgical-specific generative AI

2. Achievements

(1) Development of Generative AI with a "Brain" that Understands Surgical Context

Until now, the use of generative AI in the surgical field has focused on "anatomical recognition (i.e., the function of the eye)" to identify patient organs and blood vessels through image recognition. However, "advanced situational understanding (i.e., the function of the brain)," which can infer and interact with recognized information to comprehend complex surgical procedures, the condition of the affected area, and determine "what to do next" or "what to be careful about," has remained unimplemented and is a major challenge for such AI.

Against this backdrop, Direava has been working on the development of surgical-specific generative AI that meets the required standards of medical education. On February 20, 2026, this AI was tested in an actual surgical environment...