Multiple Sclerosis Identification by 14-Layer Convolutional Neural Network With Batch Normalization, Dropout, and...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurology

Background:

• Multiple sclerosis (MS) is a debilitating central nervous system disease with diverse symptoms.
• Early diagnosis and treatment are crucial for managing MS progression and improving patient outcomes.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate and early detection of multiple sclerosis.
• To assess the efficacy of stochastic pooling in enhancing CNN performance for MS diagnosis.

Main Methods:

• A 14-layer convolutional neural network (CNN) was designed, incorporating batch normalization, dropout, and stochastic pooling.
• Data augmentation techniques were employed to expand the training dataset.
• The model underwent 10 independent runs using a randomly allocated hold-out set for validation.

Main Results:

• The proposed CNN achieved high diagnostic performance, with a sensitivity of 98.77 ± 0.35%, specificity of 98.76 ± 0.58%, and accuracy of 98.77 ± 0.39%.
• Stochastic pooling demonstrated superior performance compared to maximum and average pooling methods.
• The developed CNN model outperformed six state-of-the-art methods, including traditional AI and deep learning approaches.

Conclusions:

• Stochastic pooling is an effective technique for improving CNN performance in medical image analysis.
• The proposed deep learning model offers a promising, superior alternative for the early diagnosis of multiple sclerosis.