Migraine ranks third among the worldwide causes of disability, and current diagnosis is primarily symptom-based [1]. With rapid advances of medical imaging technologies, it is now possible to acquire multiple modalities of imaging data for the same patient. However, research has not been transformed into a clinical decision support system due to the lack of three important traits: (1) Flexibility (ability to incorporate features defined at different aggregation levels, e.g., ROI-wise, voxel-wise, etc.), (2) Sufficient accuracy (utilizing state-of-the-art feature selection methods, ex., evolutionary computation), and (3) Interpretability (employing white-box approaches for feature processing and model-building, enabling clinical interpretation).
A multi-modality imaging based diagnostic decision support system (MMI-DDS) is proposed to address the three aforementioned limitations. MMI-DDS includes three interconnected components: (1) a modality-wise principal component analysis (PCA) that reduces data dimensionality and meanwhile provides the flexibility for opting out tedious and error-prone co-registration for multi-modality images; (2) a novel constrained particle swarm optimization (cPSO) based classifier that is built upon the joint set of the principal components (PCs) from all the imaging modalities and achieves nearly-optimal diagnostic accuracy; (3) a clinical utility engine that employs inverse operations to identify contributing imaging features (a.k.a. biomarkers) in diagnosing the disease.
MMI-DDS is applied to multimodality magnetic resonance images (MRIs) to generate accurate, interpretable models to predict migraine patient diagnosis. The model introduces more objectivity in migraine patient diagnosis, identifies biomarkers that are consistent with findings in the migraine literature, and discovers new biomarkers that (with further research) may also serve as indicators of migraine.
Olesen, J. (2018). International classification of headache disorders. The Lancet Neurology, 17(5), 396-397.
Gaw, N., Schwedt, T. J., Chong, C. D., Wu, T., & Li, J. (2018). A clinical decision support system using multi-modality imaging data for disease diagnosis. IISE Transactions on Healthcare Systems Engineering, 8(1), 36-46. https://doi.org/10.1080/24725579.2017.1403520.
