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

Comprehensive Summary

Kim et al. evaluated the economic value of an AI-based MRI triage strategy for the early detection of Parkinson’s Disease (PD) using a cost-benefit analysis conducted in South Korea and the United States. A simulation model was constructed to represent diagnostic and cost pathways of individuals within a theoretical national cohort. Based on national demographics and reported incidence rates, the cohort comprised 48,888 adults aged 65 years and older with clinically suspected PD in South Korea and 90,000 in the United States. The AI model utilized was Heuron IPD (Idiopathic Parkinson’s Disease), a commercially available software designed for automated nigrosome-1 evaluation, which generated bilateral nigral hyperintensity maps, standardized Z-scores, and volumetric measures, providing quantitative support for IPD assessment. Two diagnostic strategies were compared: conventional Positron Emission Tomography (PET) imaging for diagnostic confirmation versus an AI-based MRI model used to determine the need for PET imaging. This approach facilitated early detection in high-confidence AI-MRI cases while reducing unnecessary PET procedures. Each cohort member was assigned to one of 24 unique patient types defined by binary variables: 1) presence or absence of PD, 2) presence or absence of economic barriers to PET access, 3) MRI detection outcome, 4) PET detection outcome, and 5) AI utilization. Cost-benefit analyses for each patient type were simulated along their respective diagnostic pathways, accounting for early diagnosis, misdiagnosis, delayed diagnosis, and PET avoidance. In South Korea, the net benefit was estimated at $9.29 million at 30% AI adoption, rising to $30.97 million at full adoption, with a cost-benefit ratio of 1.48. In the United States, the net benefit reached $75.96 million at 30% adoption and $253.19 million at full adoption, with a cost-benefit ratio of 1.36. The break-even AI unit cost was approximately $226 in South Korea and $1,506 in the United States, suggesting economic viability despite high implementation costs. Long-term projections based on PD prevalence trends and population aging in South Korea estimated an annual net benefit of $2.1 million in 2025, projected to grow to $264.7 million by 2050.

Outcomes and Implications

The model focused on the clinically relevant cases in which PET would be considered, instead of assuming universal PET use across all patients. For the South Korea cohort, both medical and non-medical costs were included in the economic analysis. For the United States cohort, only medical costs were considered. Adoption rates of 30%, 65%, and 100% were modeled to reflect gradual acceptance of technology. The PET unaffordability rate varied from 0% to 100%, but did not change the trend of the results. The AI determined PET was unnecessary for over 31% of patients, which enabled 13,000 PD patients to receive imaging who otherwise would have been excluded due to financial barriers. Important limitations included reliance on existing literature for key parameters such as the cost of delayed diagnosis and the proportion of patients who forgo PET imaging due to financial burden, use of controlled trial settings for AI performance metrics, and a simplified linear trajectory for AI adoption.