Multiple Sclerosis Detection via 6-layer Stochastic Pooling Convolutional Neural Network and Multiple-way Data Augmentation

Abstract

Multiple sclerosis is one of most widespread autoimmune neuroinflammatory diseases which mainly damages body function such as movement, sensation, and vision. Despite of conventional clinical presentation, brain magnetic resonance imaging of white matter lesions is often applied to diagnose multiple sclerosis at the early stage. In this article, we proposed a 6-layer stochastic pooling convolutional neural network with multiple-way data augmentation for multiple sclerosis detection in brain MRI images. Our approach does not demand hand-crafted features unlike those traditional machine learning methods. Via application of stochastic pooling and multiple-way data augmentation, our 6-layer CNN achieved equivalent performance against those deep learning methods which consist of so many layers and parameters that ordinarily bring difficulty to training. The results showed that this 6-layer CNN obtained a sensitivity of 95.98±0.46%, a specificity of 95.67±0.92%, and an accuracy of 95.82±0.58%. According to comparison experiments, our results are better than state-of-the-art approaches. Further, we also conducted ablation experiments to examine the contribution of stochastic pooling and multiple-way data augmentation to the original CNN model. The contrast experiments revealed that our scheme of stochastic pooling and multiple-way data augmentation enhanced the original 6-layer CNN model compared to those using maximum pooling or average pooling and inadequate data augmentation.

Multiple Sclerosis Diagnosis Using Machine Learning and Deep Learning: Challenges and Opportunities

“…Similarly, in most of the previous studies that used MRI for the diagnosis of MS, the proposed models classified the patient sample with MS versus healthy controls [ 49 , 51 , 52 , 55 , 82 , 83 , 84 , 85 , 91 , 97 , 98 ], and the discrimination between these two classes is relatively simple. However, there is a need to devise a model that discriminates among MS and other diseases that are similar on MRI scan like brain tumors.…”

“…Wang and Lima [ 82 ] used multiple augmentation techniques to better train the model. However, due to extensive augmentation, the model might have suffered from overfitting; augmentation techniques generate synthetic data.…”

“…Wang et al [ 82 ] introduced a six-layer stochastic pooling CNN to detect MS with multiple-way data augmentation. The MRIs dataset utilized was collected from 38 MS patients acquired from the Laboratory of eHealth of the University of Cyprus and 26 healthy controls acquired from a private source.…”

Exemplar MobileNetV2-Based Artificial Intelligence for Robust and Accurate Diagnosis of Multiple Sclerosis

“…These images helped the model gain a better understanding of MS and distinguish it, thus enhancing the overall performance of the model (see Table 3). [34] employed a six-layer CNN and stochastic pooling. It was tested on 676 MRI data and achieved high accuracy, sensitivity, and F1-scores.…”

AI-based model for automatic identification of multiple sclerosis based on enhanced sea-horse optimizer and MRI scans

“…In this experiment, we compared our proposed technique with other MS identification methods from machine and deep learning studies, such as Haar Wavelet Transform, principal component analysis, and Logistic Regression (HWT + PCA + LR) 61 , Wavelet Entropy and Feedforward Neural Network Trained by Adaptive Genetic Algorithm (WE + FNN + AGA) 44 , Hu Moment Invariant and Feedforward Neural Network Trained by Particle Swarm Optimization (HMI + FNN + PSO) 46 , gray level co-occurrence matrix and Feedforward Neural Network (GLCM + FNN) 45 , multiscale amplitude-modulation frequency-modulation (MAMFM) 62 , gray level www.nature.com/scientificreports/ co-occurrence matrix and ensemble Learning along with LogitBoost algorithm (GLCM + ensemble + Logit-Boost) 41 , gray level co-occurrence matrix and the gray level run length matrix (GLCM + GLRLM) 63 , 6-layer stochastic pooling convolutional neural network (6l-CNN) 64 , discrete wavelet transform, principal component analysis, and least squares support vector machine (DWT + PCA + LS-SVM) 65 , Wavelet Entropy and Hybridization of Biogeography-Based Optimization and Particle Swarm Optimization (WE + HBP) 66 , discrete wavelet transform and probabilistic principal component analysis with random forests (DWT + PPCA + RF) 67 , fractional Fourier entropy, multilayer perceptron, and Self-adaptive Three-segment-encoded Jaya algorithm (FRFE + MLP + ST-Jaya) 16 , Biorthogonal Wavelet Transform, RBF Kernel Principal Component Analysis, and Logistic Regression (BWT + RKPCA + LR) 38 , Minkowski-Bouligand dimension, single hidden layer neural network, and three-segment representation biogeography-based optimization (MBD + SHLNN + TSR-BBO) 68 , stationary wavelet entropy and k-nearest neighbors (SWE + KNN) 37 , and biorthogonal wavelet features and fitness-scaled adaptive genetic algorithm (BWF + FAGA) 69 . All the techniques were tested on the same open-access dataset as ours.…”