Introducing the DOCTOR EYE project.

TECHNOLOGY PLATFORM

PIEGENT

An integrated medical AI technology platform delivering safe and precise analytics

Biomarker analysis images

PLATFORM MODULE

OASIS

(Ophthalmic Analysis Segmentation Image System)

Advanced analysis technology for biomarkers including anatomical structures, blood vessels, and nerves within fundus images.
Accurately segments and analyzes features across 40 types of biomarkers in fundus photography.

✓ Anatomical Features
✓ Vascular Features
✓ Neural Features

MODULE / MODEL

CoMoE

(Cooperative Mixture-of-Experts)

Combines 40 biomarkers extracted via OASIS to extract optimized features for assessing risks of ocular diseases, brain disorders, and ADHD.

Risk Assessment Performance
for Ocular, Neurological, & ADHD Conditions

Ocular, Brain, & ADHD Risk Analysis

87%

EXPLAINABLE AI

REMEDIS

Technology that highlights key focus areas analyzed by AI within fundus images using green markers.

✓ Visual Explanation
✓ Influential Region Heatmap

Multi-Label Retinal Diagnosis

Saliency-Guided Sparse Expert Routing

Disentangling Co-occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing

A Saliency-Guided Sparse MoE framework designed to disentangle and independently diagnose co-occurring retinal pathologies in fundus images.

Lesion-centric expert routing minimizes cross-disease interference, achieving 0.912 mAUC and 0.653 mF1 on ODIR-2019 , with an inference speed of 5.23 ms per image .

0.912 mAUC

0.653 mF1

5.23 ms Inference / Image

Disentangling Co-occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing

Nagur Shareef Shaik, Jeongwoo Park, Yeong-Jin Kim, Jaeuk Jung, Hyunjung Oh, Dong Hye Ye

2026 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)

RESEARCH & DATA

Oculomics

World-leading Oculomics research conducted in collaboration with Seoul National University Hospital.

• Fundus Images: 1 Million
• MRI & CT Scans: 400,000

GLOBAL PARTNERSHIP

Cooperation

Piehealthcare collaborates with leading domestic and international institutions to advance technology development and research.

• Piehealthcare: Development of AI-based fundus & disease analysis technology
• Seoul National University Hospital: Joint clinical data research & validation
• University of Utah CMI: FDA collaboration with University of Utah CMI
• Georgia State University: Joint development of X-DEMENTIA-M

Piehealthcare

Seoul National University Hospital

University of Utah CMI

Georgia State University

TECHNOLOGY PLATFORM

PIEGENT

Integrated Medical AI Platform Delivering Safe and Precise Analysis

Biomarker Analysis Images

PLATFORM MODULE

OASIS

(Ophthalmic Analysis Segmentation Image System)

Advanced biomarker analysis technology for fundus photography, covering anatomical structures, blood vessels, and neural features.
Identifies and analyzes 41 distinct biomarkers in fundus images.

Technology that combines 41 biomarkers analyzed by OASIS to identify the optimal combination for cerebral aneurysm analysis.

Anatomical Features
Vascular Features
Neural Features

MODULE / MODEL

CoMoE

(Cooperative Mixture-of-Experts)

Technology that combines 41 biomarkers analyzed by OASIS to determine the optimal combination for analyzing ocular diseases, brain disorders, and ADHD.

Risk Analysis Performance

Ocular, Brain, & ADHD Risk Analysis

87%

EXPLAINABLE AI

REMEDIS

A technology that visualizes which regions of the fundus photograph the AI primarily focused on for analysis using green markers.

Visual Explanation
Influential Region Heatmap

Retinal Disease Diagnosis

REMEDIS

A Clinical AI Framework for Retinal Disease Diagnosis with Explainable Fundus Image Analysis

This paper introduces a system that diagnoses 4 major retinal diseases through real-time fundus image analysis in a SaaS-based cloud environment
and presents visual diagnostic rationale for AI using Grad-CAM technology.

REMEDIS: A Clinical AI Framework for Retinal Disease Diagnosis with Explainable Fundus Image Analysis

Youngkyu Lee¹, Jinho Lee¹, Youngdae Jo¹, Jeongwoo Park¹,²

Multi-Label Retinal Diagnosis

Saliency-Guided Sparse Expert Routing

Disentangling Co-occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing

A Saliency-Guided Sparse MoE framework designed to disentangle and independently diagnose co-occurring retinal pathologies in fundus images. Lesion-centric expert routing minimizes cross-disease interference, achieving 0.912 mAUC and 0.653 mF1 on ODIR-2019, with an inference speed of 5.23 ms per image.

0.912 mAUC

0.653 mF1

5.23 ms Inference / Image

Disentangling Co-occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing

Nagur Shareef Shaik, Jeongwoo Park, Yeong-Jin Kim, Jaeuk Jung, Hyunjung Oh, Dong Hye Ye

2026 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)

RESEARCH & DATA

Oculomics

This is the world's largest Oculomics research initiative, conducted in collaboration with Seoul National University Hospital.

Fundus Images: 1 Million
MRI, MRA, CT, US: 400,000

GLOBAL PARTNERSHIP

Cooperation

Piehealthcare collaborates with leading domestic and international institutions on technology development and research.

Piehealthcare: Development of AI-based fundus and disease analysis technology
Seoul National University Hospital: Joint clinical data research and validation
University of Utah CMI: Collaboration with FDA and University of Utah CMI
Georgia State University: Joint development of X-DEMENTIA-M

Piehealthcare

Seoul National University Hospital

University of Utah CMI

Georgia State University