AI Medical Service Obtains Manufacturing and Marketing Approval for Endoscopic Image Diagnosis Support Software, gastroAI model-G3
AI Medical Service Inc. has received regulatory approval for 'gastroAI model-G3,' an AI software that detects suspected early gastric cancer and adenomas during endoscopies. Launch is set for summer 2026.
📋 Article Processing Timeline
- 📰 Published: May 16, 2026 at 00:00
- 🔍 Collected: May 15, 2026 at 15:32
- 🤖 AI Analyzed: May 15, 2026 at 18:58 (3h 26m after Collected)
AI Medical Service Inc. (Headquarters: Toshima-ku, Tokyo, CEOs: Tomohiro Tada, Yasushi Takigawa, hereinafter "AIM"), whose mission is "to save patients worldwide," announces that it has obtained manufacturing and marketing approval on May 13, 2026, for "gastroAI model-G3". This is an endoscopic image diagnosis support software that extracts areas suspected of early gastric cancer and adenomas [1] in gastric endoscopic videos acquired by doctors for the purpose of lesion detection, showing the videos to the doctors to alert them to the suspected areas, thereby assisting doctors in interpreting the videos and detecting lesions [2]. Sales of this product will begin sequentially from around summer 2026.
About gastroAI model-G3
Product Overview
gastroAI model-G3 detects areas suspected of early gastric cancer and adenomas in the stomach, superimposing the results on the video output from the endoscopic equipment. When detecting a lesion image, it displays up to three suspected lesion marks on the suspected area.
Sales Name: Endoscopic Image Diagnosis Support Software AP01
General Name: Endoscopic Image Diagnosis Support Program for Lesion Detection
Approval Number: 30800BZX00135000
About Accuracy
In addition to expanding the variation of training data, optimizing the AI model, and optimizing the determination process linked to multiple frames, it achieves both high detection performance and a reduction in false positives by featuring an automatic recognition function that detects images unsuitable for analysis and pauses the analysis. This reduces excessive detection and supports efficient diagnosis of areas that truly require confirmation.
[Test Results]
Results of a retrospective performance evaluation test [3] targeting lesions diagnosed as early gastric cancer (protruding, flat, and depressed types) and adenomas based on gastric endoscopic videos and pathological information. (Sample size: 1099)
Sensitivity (%)
(97.5% one-sided confidence interval lower limit [4])
94.6
(92.4)
Specificity (%)
(97.5% one-sided confidence interval lower limit)
94.3
(92.0)
Background of Development
In Japan, gastric cancer continues to rank high in both morbidity and mortality. Approximately 100,000 patients are newly diagnosed with gastric cancer each year, ranking 3rd after colorectal and lung cancer, and 4th in terms of deaths [5]. While gastric cancer is a highly treatable disease if found early, a key characteristic is that the survival rate drops significantly as the stage advances; the 5-year relative survival rate is over 95% if discovered at Stage I, but falls below 50% if discovered at Stage III or later. Therefore, early detection of gastric cancer is crucial, but early gastric cancer is difficult to distinguish, and it is said that about 4.5-25.8% of cases are missed [6].
The gastroAI model-G3, which recently received manufacturing and marketing approval, makes it possible to analyze endoscopic videos during examinations. This allows the AI to continuously run alongside the doctor from the beginning to the end of the examination. Against the backdrop of challenges such as the uneven distribution of doctors regionally, gastroAI model-G3 targets approximately 18,000 medical institutions [7] nationwide that perform endoscopies, and we anticipate future expansion, including enhancing the value provided to existing customer facilities.
AI Medical Service will continue to strive toward realizing its mission to "save patients worldwide."
About AI Medical Service
AI Medical Service is a medical startup company with a mission to "save patients worldwide." Japan leads the world in endoscopic medicine, accumulating world-class data in both quality and quantity. We are engaged in the research and development of products utilizing this data, and by delivering these results to medical sites as quickly as possible, we aim to save patients worldwide.
[1] Benign tumors. Since they can become cancerous, large ones may be treated endoscopically.
[2] The positioning of this product is to "assist physicians in detecting lesions," and it is not intended to support qualitative diagnosis or determine treatment policies based solely on the analysis results of this product.
[3] A performance evaluation test using anonymized real clinical data previously collected from medical institutions.
[4] Statistically, the lower limit estimated with high confidence (97.5%) that the true value is equal to or greater than this value.
[5] Source: National Cancer Center Japan, Cancer Information Service, "Cancer Statistics" (National Cancer Registry)
[6] Source: Hosokawa et al. Hepatogastroenterology 2007 Mar;54(74):442-4.
[7] Source: Ministry of Health, Labour and Welfare, "Survey of Medical Institutions" (e-Stat)
About gastroAI model-G3
Product Overview
gastroAI model-G3 detects areas suspected of early gastric cancer and adenomas in the stomach, superimposing the results on the video output from the endoscopic equipment. When detecting a lesion image, it displays up to three suspected lesion marks on the suspected area.
Sales Name: Endoscopic Image Diagnosis Support Software AP01
General Name: Endoscopic Image Diagnosis Support Program for Lesion Detection
Approval Number: 30800BZX00135000
About Accuracy
In addition to expanding the variation of training data, optimizing the AI model, and optimizing the determination process linked to multiple frames, it achieves both high detection performance and a reduction in false positives by featuring an automatic recognition function that detects images unsuitable for analysis and pauses the analysis. This reduces excessive detection and supports efficient diagnosis of areas that truly require confirmation.
[Test Results]
Results of a retrospective performance evaluation test [3] targeting lesions diagnosed as early gastric cancer (protruding, flat, and depressed types) and adenomas based on gastric endoscopic videos and pathological information. (Sample size: 1099)
Sensitivity (%)
(97.5% one-sided confidence interval lower limit [4])
94.6
(92.4)
Specificity (%)
(97.5% one-sided confidence interval lower limit)
94.3
(92.0)
Background of Development
In Japan, gastric cancer continues to rank high in both morbidity and mortality. Approximately 100,000 patients are newly diagnosed with gastric cancer each year, ranking 3rd after colorectal and lung cancer, and 4th in terms of deaths [5]. While gastric cancer is a highly treatable disease if found early, a key characteristic is that the survival rate drops significantly as the stage advances; the 5-year relative survival rate is over 95% if discovered at Stage I, but falls below 50% if discovered at Stage III or later. Therefore, early detection of gastric cancer is crucial, but early gastric cancer is difficult to distinguish, and it is said that about 4.5-25.8% of cases are missed [6].
The gastroAI model-G3, which recently received manufacturing and marketing approval, makes it possible to analyze endoscopic videos during examinations. This allows the AI to continuously run alongside the doctor from the beginning to the end of the examination. Against the backdrop of challenges such as the uneven distribution of doctors regionally, gastroAI model-G3 targets approximately 18,000 medical institutions [7] nationwide that perform endoscopies, and we anticipate future expansion, including enhancing the value provided to existing customer facilities.
AI Medical Service will continue to strive toward realizing its mission to "save patients worldwide."
About AI Medical Service
AI Medical Service is a medical startup company with a mission to "save patients worldwide." Japan leads the world in endoscopic medicine, accumulating world-class data in both quality and quantity. We are engaged in the research and development of products utilizing this data, and by delivering these results to medical sites as quickly as possible, we aim to save patients worldwide.
[1] Benign tumors. Since they can become cancerous, large ones may be treated endoscopically.
[2] The positioning of this product is to "assist physicians in detecting lesions," and it is not intended to support qualitative diagnosis or determine treatment policies based solely on the analysis results of this product.
[3] A performance evaluation test using anonymized real clinical data previously collected from medical institutions.
[4] Statistically, the lower limit estimated with high confidence (97.5%) that the true value is equal to or greater than this value.
[5] Source: National Cancer Center Japan, Cancer Information Service, "Cancer Statistics" (National Cancer Registry)
[6] Source: Hosokawa et al. Hepatogastroenterology 2007 Mar;54(74):442-4.
[7] Source: Ministry of Health, Labour and Welfare, "Survey of Medical Institutions" (e-Stat)