HEP-2 Specimen Image Segmentation and Classification Using GLCM and DCT Based Feature Extraction with CNN Classifier

Abstract

Effective Human Epithelial-2 (HEp-2) cell image classification can encourage the determination of numerous autoimmune system diseases Different computerized image processing methods can be successfully connected to perform HEp-2 segmentation and classification process. This work, a programmed framework for HEp-2 segmentation and classification utilizing image processing ideas is utilized. HEP-2 cell image classification (CIC) utilizing convolutional neural network (CNN) with Gray-Level Co-Occurrence Matrix (GLCM) and Discrete Cosine Transform (DCT) feature extraction (HEP-CIC-CNN-DCT-GLCM) is proposed in this work to improve the presentation of classification precision. Adaptive Gamma Correction (ADC) technique is used as preprocessing technique to improve the differentiation of the cell image for the segmentation process. Median Filtering technique method is utilized to expel noise from the image if any kind of distortion or cracks occurred while acquisition or transmission. Data augmentation is utilized in the training stage to improve the viability of training process. Different arrangement of data is created by pivoting pictures in various points to get more sample images to perform great training. Proposed technique can classify six classes such as Homogeneous, Speckled, Nucleolar, Centromere, Nuclear Membrane, and Golgi. Mean Class Accuracy (MCA) of about 96.56%, which is each a lot higher contrasted with past work related to accessible ICPR 2014 dataset.