Study on Contribution of Biological Interpretable and Computer-Aided Features Towards the Classification of Childhood Medulloblastoma Cells

Abstract

Diagnosis and Prognosis of brain tumour in children is always a critical case. Medulloblastoma is that subtype of brain tumour which occurs most frequently amongst children. Post-operation, the classification of its subtype is most vital for further clinical management. In this paper a novel approach of pathological subtype classification using biological interpretable and computer-aided textural features is forwarded. The classifier for accurate features prediction is built purely on the feature set obtained by segmentation of the ground truth cells from the original histological tissue images, marked by an experienced pathologist. The work is divided into five stages: marking of ground truth, segmentation of ground truth images, feature extraction, feature reduction and finally classification. Kmeans colour segmentation is used to segment out the ground truth cells from histological images. For feature extraction we used morphological, colour and textural features of the cells followed by feature reduction using Principal Component Analysis. Finally both binary and multiclass classification is done using Support Vector Method (SVM). The classification was compared using six different classifiers and performance was evaluated employing five-fold cross-validation technique. The accuracy achieved for binary and multiclass classification before applying PCA were 95.4 and 62.1% and after applying PCA were 100 and 84.9% respectively. The run-time analysis are also shown. Results reveal that this technique of cell level classification can be successfully adopted as architectural view can be confusing. Moreover it conforms substantially to the pathologist's point of view regarding morphological and colour features, with the addition of computer assisted texture feature.

CoMB-Deep: Composite Deep Learning-Based Pipeline for Classifying Childhood Medulloblastoma and Its Classes

“…Table 7 displays this comparison. Das et al (2018b) extracted textural features, comprising GLCM, HOG, Tamura, LBP, and GLCM. They fused all these features and used PCA to reduce them and SVM classifier to classify pediatric MB classes.…”

“…This means that investigating a different combination of feature sets and selecting the most influential fused feature set can improve the accuracy of the classifier. In the same manner, CoMB-Deep pipeline explored different combinations of features extracted from several CNNs and fused using DWT to select the most fused feature sets that impact the performance of the classifier As mentioned before DL techniques are more favorable than the traditional feature extraction methods (Das et al, 2018b(Das et al, , 2020a. CoMB-Deep merges DL and textural analysis benefits, as it first extracts spatial features from 10 CNNs.…”

“…Few research articles have investigated this multi-class classification problem. To the best of our knowledge, the first and only research group that investigated the classification of the four MB subtypes from histopathological images using machine and deep learning (DL) techniques were studies by Das et al (2018b) who presented a diagnostic framework to diagnose the subclasses of such pediatric tumor. The authors initially segmented images using k-means clustering.…”

“…They found out that the MANOVA method increased the classification accuracy to 65.2%, which is greater than that of 56.5% without MANOVA. In the same year, Das et al (2020b) decided that instead of using individual sets of features (Das et al, 2018b), they produced various groups of fused features to study the influence of feature fusion and selected the best mixture of fused feature sets. They employed PCA and SVM and achieved an accuracy of 96.7% utilizing the four combined features.…”

“…They employed PCA and SVM and achieved an accuracy of 96.7% utilizing the four combined features. Later, in the same year, Das et al (2020c) These previous methods have shown several drawbacks that can be summarized as follows: the approaches by Das et al (2018bDas et al ( , 2020a utilized traditional individual feature extraction methods that depend on textural analysis, color, or morphological operations. Drawbacks of conventional methods are discussed earlier.…”