News

2024/11/12 Press Releases

Jichi Medical University and DeepEyeVision Develop Automated Fundus Imaging Report System Using a Large Multimodal Model (LMM)

November 12, 2024
Jichi Medical University
DeepEyeVision Inc.



This press release is intended to provide information about our research activities and corporate activities to members of the press and academia. It may contain information related to medical devices and pharmaceuticals, but it is not intended to solicit customers or provide medical advice to the general public.

Jichi Medical University (Shimotsuke City, Tochigi Prefecture, President: Ryozo Nagai; hereinafter referred to as “Jichi Medical University”) and DeepEyeVision Inc. (headquartered in Shimotsuke City, Tochigi Prefecture, CEO: Hidenori Takahashi; hereinafter referred to as “DeepEyeVision”), a venture company originating from Jichi Medical University and the University of Tsukuba, have developed a system that uses a Large Multimodal Model (LMM)*1 to automatically create human-readable reports including disease probabilities and key findings from fundus examination images.*2 This system is expected to significantly reduce the time required for doctors to enter interpretation results into electronic medical records, contributing to the advancement of “work style reform for doctors.”

This work was supported by Cross-ministerial Strategic Innovation Promotion Program (SIP) on “Integrated Health Care System” Grant Number JPJ012425.*3 (hereinafter referred to as “SIP3”). It is scheduled to be operational at Jichi Medical University’s Health Examination Center within the fiscal year 2024.

1. Background of Development

With the rapid aging of the population, the number of patients with ophthalmologic diseases, such as glaucoma, cataracts, and age-related macular degeneration, is increasing, while the number of ophthalmology specialists capable of interpreting fundus images remains chronically insufficient. This has led to the issue of extended working hours for physicians, as significant time is required to input examination findings for fundus screenings conducted during health check-ups.

2. Research Content and Significance

In response, Jichi Medical University and DeepEyeVision have developed a system that uses LMM to automatically generate disease names and interpretation findings from fundus images. The LMM, which can process multiple types of information, integrates fundus images with information from medical records, such as the examinee’s age, gender, and medical history, and outputs an interpretation draft as text. This draft is then reviewed by ophthalmology specialists, creating a workflow that enables a 70-90% reduction in interpretation time compared to previous methods*4, supporting “work style reform for doctors” and improving the understanding and satisfaction of both health check physicians and examinees.

Jichi Medical University and DeepEyeVision aim to link this system with the Japanese medical LMM being developed under SIP3 and commence trial operations at Jichi Medical University’s Health check center by the end of fiscal 2024. The system will also be incorporated into DeepEyeVision’s remote fundus image interpretation service,*5 where quantitative evaluations of its time-saving effects will be conducted.

Screen Display of Fundus Interpretation Generation Function Using LMM

3. Exhibition and Symposium Presentation at the 78th Annual Congress of Japanese Clinical Ophthalmology

DeepEyeVision will demonstrate this AI model*6 at the 78th Annual Congress of Japanese Clinical Ophthalmology.

Event Details:

  • Dates: November 14–17, 2024 (exhibits open until November 16)
  • Location: Kyoto International Conference Center, The Prince Kyoto Takaragaike
  • Booth Number: N-16
  • Exhibitor: DeepEyeVision Inc.

In addition, DeepEyeVision’s CEO, Hidenori Takahashi, and advisor, Makoto Koyama, will give presentations on the following topics:

78th Annual Meeting of the Japanese Society of Clinical Ophthalmology

  • Symposium 19: Ophthalmic Practice and AI Research
  • Date: November 17, 8:50–10:10
  • Location: Venue 1
  • Presentations:
    • Hidenori Takahashi (University of Tsukuba, Jichi Medical University): “Estimation of Non-perfusion Area from Color Ultra-widefield Retinal Images”
    • Makoto Koyama(Minamikoayasu Eye Clinic): “Estimation of Visual Field Thresholds from 3D OCT Data”
  • https://convention.jtbcom.co.jp/78ringan/program.html

Comments on This Research and Development

Comment from Jichi Medical University President Ryozo Nagai (M.D., Ph.D.)

Despite the rapid progress in AI technology, social implementation in the Japanese medical field remains limited. In this context, the presence of DeepEyeVision, the first venture from Jichi Medical University, which has already obtained medical device certifications and is making steady progress in AI’s social implementation under Prof. Takahashi, is highly reassuring. I hope that the research and development achievements of our university will contribute to better healthcare through the activities of this company.


*1 LMM (Large Multimodal Model) is a model developed based on Large Language Models (LLMs) capable of handling multiple modalities of input and output, such as images, audio, and video.

*2 Fundus images are photographs of the retina, allowing non-invasive imaging of blood vessels and detection of systemic diseases such as hypertension and diabetes.

*3 Cabinet Office’s SIP Phase 3 Supplemental Budget “Utilization of Generative AI in the Construction of Integrated Healthcare Systems”
Institution: University of Tokyo, Research Center for Advanced Science and Technology
Research Theme: Development and Implementation of High-Performance, Reliable Medical LMMs
Research Project: Development of a Medical LMM for Fundus Image Interpretation, with participation from the clinical medicine team at Jichi Medical University
https://sip3.ncgm.go.jp/

*4 Nagaoka et al. Impact of AI Heatmaps with Judgment Basis on Fundus Interpretation by Ophthalmology Residents, 126th Annual Meeting of the Japanese Ophthalmological Society, Abstract Collection, 2022
Hashimoto et al. Impact of AI Heatmaps on Ophthalmologist Fundus Interpretation, 3rd Annual Meeting of the Japanese Society of Artificial Intelligence in Ophthalmology, 2022

*5 This remote interpretation service by DeepEyeVision is provided on the platforms of ESFILL Corporation. and Siemens Healthcare K.K. The core system is used solely within DeepEyeVision for data analysis and does not constitute a medical device. Its system architecture, learning data, and algorithms are distinct from DeepEyeVision’s certified medical device programs.

*6 This model, presented as an unapproved medical device (non-medical device), will be showcased following the society’s application, and is not available for sale or transfer.

About Jichi Medical University

Jichi Medical University, in collaboration with Japan’s prefectures, was established to educate dedicated doctors of ethical standing in order to provide highly skilled medical care in remote areas of Japan where medical aid is limited, to promote the health and well-being for people living in rural communities, and to contribute to the advancement of medicine.

About DeepEyeVision Inc.

DeepEyeVision Inc. is a venture company originating from Jichi Medical University, established by Dr. Hidenori Takahashi, M.D., Ph.D., who is a Professor at the Cybermedicine Research Center of the University of Tsukuba and an Associate Professor in the Department of Ophthalmology at Jichi Medical University. The company was founded with the mission of bringing medical AI technologies developed in academic research to the broader society. Leveraging AI-based imaging diagnostics, DeepEyeVision is committed to advancing not only ophthalmology but also various other medical fields through its research and development efforts.

– CEO: Hidenori Takahashi, M.D., Ph.D.
– Headquarters: C/O Department of Ophthalmology, Jichi Medical University, 3311-1 Yakushiji, Shimotsuke, Tochigi Prefecture, Japan.
– Website: https://deepeyevision.com/en/

Certifications

– Second-class medical device manufacturing and sales (License No. 09B2X10014)
– Medical Device Manufacturing (Registration No. 09BZ200083)
– Controlled Medical Device Sales Business Notification (Notification No. B42917)

Affiliated Organizations

– Japan Ophthalmic Instrument Association (JOIR)
– Japan Medical Imaging and Radiological Systems Industries Association (JIRA)
– Council for AI Medical Devices
– Japan Ocular Imaging Registry
– Japanese Society of Artificial Intelligence in Ophthalmology

Contact

DeepEyeVision Inc.
E-mail: pr@deepeyevison.com