Objective: To use an artificial intelligence (AI) based framework to detect Parkinson’s disease (PD) in Magnetic Resonance Imaging (MRI) brain scans.

Background: Candidate neuroprotective treatments for PD are highlighting the need for early diagnostic tests, as timely intervention would require early detection. A number of exploratory imaging techniques have suggested that early pathological brain changes may be possible to detect using dedicated experimental MRI sequences. We explored whether machine learning (ML) might be employed to detect such brain changes on routine MRI scans, as routine brain imaging would provide a cost-effective means of screening for PD. A subset of ML known as deep learning (DL) has shown great promise in diagnostic medical imaging, sometimes matching or even exceeding the diagnostic performance of radiologists. DL offers the potential of automated diagnosis by detecting patterns that might be invisible to the human eye. DL methods have sometimes been criticised for being “black boxes”, but newly emerging explainability methods in the field of AI are allowing the decisions made by DL models to be better interpreted by practitioners.

Method: We trained a convolutional neural network to classify 138 PD and 60 control brain MRI images acquired from the Parkinson’s Progression Marker Initiative (PPMI) database. Models were assessed using 5-fold cross-validation. We used Deep SHapley Additive exPlanations (DeepSHAP) to calculate and visualise the contribution of individual pixels to the model’s prediction.

Results: A model was developed using a combined dataset of axial T2 and proton density MRI scans, which classified images with 79% accuracy and a Receiver Operating Characteristic area under the curve (AUC) of 0.86. Another model was trained on just T2 scans, and classified images with 81% accuracy and AUC of 0.83. A further model was trained on just proton density scans, and classified images with 84% accuracy and AUC of 0.88. The heatmaps generated using Shapley values demonstrated predominant contribution to the prediction in the midbrain slices.

Conclusion: Our models exhibited good diagnostic performance. The use of explainable AI highlighted regions of interest consistent with the known neuropathology of PD, providing a focus for future work. We will validate these models in a large dataset of routinely collected NHS MRI scans, many of which precede onset of motor symptoms.

« Back to 2022 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/artificial-intelligence-based-detection-of-parkinsons-disease-in-magnetic-resonance-imaging-brain-scans/