Early Detection of Parkinson’s Disease Dementia Using Dual-Sided Multi-Scale Convolutional Neural Networks (DSMS-CNN)

Abstract

Detecting the potential for Parkinson’s Disease Dementia (PDD) as early as possible is crucial to ensure that quality of life can be maintained. However, the full origins of this condition are unknown and analysing potential causes such as the influence of the Cholinergic Basal Forebrain (cBF) can be challenging due to variation in brain tissue as well as low scan resolution. Additionally, the structure and function of the cBF can span both brain hemispheres, and therefore prove difficult to analyse using a singular deep learning method. In this paper, we propose a multi-scale, dual-sided approach to analysis of regions with low surface area such as the cBF. Initially, images are parsed using super-resolution to increase resolution and contrast. Then, a dual sided multi-scale convolutional neural network (DSMS-CNN) model is proposed to classify subjects as either normal cognition or PDD based on both hemispheres of the cBF together. Ablation studies and comparison experiments with state-of-the-art CNN models show that DSMS-CNN can achieve promising and superior performance.

Citation

Recommended citation: Altham, C. et al. (2023). Early Detection of Parkinson’s Disease Dementia Using Dual-Sided Multi-scale Convolutional Neural Networks (DSMS-CNN). In: Su, R., Zhang, Y., Liu, H., F Frangi, A. (eds) Medical Imaging and Computer-Aided Diagnosis. MICAD 2022. Lecture Notes in Electrical Engineering, vol 810. Springer, Singapore. https://doi.org/10.1007/978-981-16-6775-6_17 https://link.springer.com/chapter/10.1007/978-981-16-6775-6_17