Medicine & AI

Abstract

There are more than 7,000 rare diseases, some affecting 3,500 or fewer patients in the US. Due to clinicians’ limited experience with such diseases and the heterogeneity of clinical presentations, approximately 70% of individuals seeking a diagnosis today remain undiagnosed. Deep learning has demonstrated success in aiding the diagnosis of common diseases. However, existing approaches require labeled datasets with thousands of diagnosed patients per disease. Here, we present SHEPHERD, a few shot learning approach for multi-faceted rare disease diagnosis. SHEPHERD performs deep learning over a biomedical knowledge graph enriched with rare disease information to perform phenotype-driven diagnosis. Once trained, we show that SHEPHERD can provide clinical insights about real-world patients. We evaluate SHEPHERD on a cohort of N = 465 patients representing 299 diseases (79% of genes and 83% of diseases are represented in only a single patient) in the Undiagnosed Diseases Network. SHEPHERD excels at several diagnostic facets: performing causal gene discovery (causal genes are predicted at rank = 3.56 on average), retrieving “patients-like-me” with the same causal gene or disease, and providing interpretable characterizations of novel disease presentations. We additionally examine SHEPHERD on two other real-world cohorts, MyGene2 (N = 146) and Deciphering Developmental Disorders Study (N = 1,431). SHEPHERD demonstrates the potential of deep learning to accelerate rare disease diagnosis and has implications for using deep learning on medical datasets with very few labels.

Corporate Initiatives

Aiming to reduce the pain caused by illness, we provide three services to support personalized medicine: 1) Proposing optimal treatment for each patient through genetic testing; 2) Utilizing medical data to improve diagnostic accuracy and accelerate drug development; 3) Providing clinical support, including clinical trial information, through AI apps.

The TMINI® Miniature Robotic System developed by THINK Surgical is a surgical robot that promotes personalized medicine in the field of orthopedics. TMINI creates CT-based 3D surgical plans based on each patient's anatomical structure and needs. During surgery, bone pins can be placed with extremely high precision while correcting the surgeon's movements in real time. This allows for bone resection and implant placement that are optimized for each patient. Another notable feature is the existence of TMINI's unique database, "ID-HUB." This system integrates and manages various implant information from multiple manufacturers, allowing doctors to flexibly select the implant that is most suitable for the patient's condition and bone structure. This makes it possible to create the optimal treatment plan for each patient without relying on standardized surgical procedures.