サイトロゴ

Non-Invasive AI Diagnostic Model and Test Optimization System for Central Diabetes Insipidus

A diagnostic AI technology that analyzes routine clinical data to accurately predict the necessity of hypertonic saline load testing, reducing patient burden and healthcare costs.

Advantages

- Avoidance of Invasive Testing: Achieves highly accurate predictions using only routine clinical laboratory data, enabling approximately 50% of patients to forgo burdensome load testing.
- Improvement in Patient Safety and Comfort: Optimizes the indication for invasive testing associated with discomfort and adverse risks.
- Highly Versatile Biomarker Selection: Built on standard clinical indicators measurable with widely available analytical equipment and existing diagnostic platforms.

Current Stage and Key Data

Completed Proof of Concept (PoC) phase, currently preparing for multi-center validation and commercial software development.
- Predictive Accuracy (AUC): Demonstrated high diagnostic predictive performance with an AUC of 0.879 using machine learning models built on clinical data.
- Test Avoidance Rate: Achieved a 52% test avoidance rate (95% CI: 39–64%) when using diagnostic thresholds maintaining positive/negative predictive values above 90%.
- High Diagnostic Sensitivity: Achieved exceptional screening performance with 89.3% sensitivity in target patient subgroup evaluations.

Partnaring Model

Joint development of medical AI software, technology licensing, and strategic partnerships for clinical validation.
- Potential partners: Clinical laboratory service providers, diagnostic reagent and equipment manufacturers, medical AI solution developers, and SaMD vendors.

Background and Technology

Although stimulation tests such as hypertonic saline load testing are widely used to differentiate central diabetes insipidus, they place a significant burden on patients and present safety risks for individuals with heart failure or advanced age. Furthermore, these tests require multiple blood draws and substantial healthcare resources, driving a strong clinical demand for simpler, less invasive screening methods. This technology predicts load test outcomes with high precision using routine laboratory parameters, streamlining clinical workflows and improving patient quality of life by preventing unnecessary invasive procedures.

Principal Investigator

Satoshi Naito (Nagoya University Hospital, Tokai National Higher Education and Research System)

Patents and Publications

- Patent pending

Project No:bk-05265