Economic value of AI-based MRI triage for Parkinson’s disease: a cost-benefit study in South Korea and the United States

Abstract

Background: Early and accurate diagnosis of Parkinson's disease (PD) remains a major clinical and economic challenge, particularly in settings where dopaminergic imaging, such as positron emission tomography (PET) scans, is limited by cost, availability, and patient access. Artificial intelligence (AI) has emerged as a promising tool to support magnetic resonance imaging (MRI)-based diagnosis of PD, but its economic value has yet to be fully evaluated. Methods: The AI model used in this study analyzes susceptibility map-weighted MRI to detect nigrosome-1 signal loss (the "swallow-tail sign"), providing objective support for early PD identification. We conducted a patient-level cost-benefit analysis (CBA) comparing current PET-based diagnostic pathways with an MRI-based AI triage strategy for PD. A total of 24 mutually exclusive diagnostic scenarios were modeled to capture variation in disease presence, AI accuracy, and PET access. The analysis was conducted from a societal perspective in South Korea and a healthcare system perspective in the United States, covering both short-term (1-year) and long-term (2025-2050) horizons. Sensitivity analyses and AI adoption rate scenarios (30, 65, 100%) were included. Results: In short-term analysis, AI-assisted diagnosis yielded net benefits of 9.3 million US dollars (USD) (South Korea) and 76.0 million USD (United States) under 30% adoption, which increased to 31.0 million USD and 253.2 million USD, respectively, under full AI adoption. Benefit-cost (B/C) ratios exceeded 1.4 in Korea and 1.3 in the U. S., and net benefit remained positive up to an AI unit cost of 226 USD in Korea and 1,506 USD in the U. S. The AI model also reduced PET use by over 31% through effective triage and enabled over 13,000 Korean PD patients to access PET who might otherwise have forgone it due to cost. Long-term projection (Korea only) indicated cumulative net savings of 2.5 billion USD by 2050 with gradually increasing AI adoption. Discussion: MRI-based AI triage for PD diagnosis is a cost-beneficial strategy with the potential to reduce unnecessary imaging and expand access among underserved populations. Particularly in health systems with limited PET availability, this approach may offer scalable economic and clinical advantages over time.

Ähnliche Arbeiten