In digital image processing field there are various segmentation techniques. This paper is based on watershed segmentation, one of the powerful segmentation method which helps the analysist to achieve the desired region from the image. This technique is successfully used in various domains for different applications. Here cancer affected thyroid images are subjected watershed algorithm which segments the image using morphological operations. This helps to achieve the region of interest from where its textural features such as mean, variance, entropy, standard deviation can be extracted.
In recent years, deep learning models have shown remarkable achievements in the analysis of various medical images, including the classification of brain tumors from MRI scans. Brain tumor detection and classification is very important in the medical field as it is one of the deadliest diseases. With the advent of the Vision Transformer (ViT) architecture, there arises an opportunity to leverage its capabilities for medical image analysis. This study presents a novel approach for brain tumor classification utilizing hybrid ViT models. The proposed hybrid model uses the efficient classifier ResNet152 as the backbone. The hybrid model with ViT-Huge (ViT-H) uses the patches of size 16×16 generated from the feature maps of ResNet152 for training, validating and testing purposes. The proposed model ResNet152-ViT-H/16 is compared with the other 4 existing models of ViT. The overall results show that the model is outperforming the state-of-the-art models with an accuracy of 97.33 %.
Breast cancer is an life threatening disease in USA and UK. It is also one of the major diseases that has greater death rate. Cancer is the erratic growth of cells that originate in the blood tissue and Tumours may be malignant or benign. Early detection increases the chances of survival and reduces the death rate. This paper compares the approach to classify the mammogram based on the features extracted using local binary pattern (LBP) and local gradient pattern (LGP) with their histograms and the results were compared. Local binary pattern and Local gradient pattern are the techniques that are generally used for textural pattern analysis. The generated pattern is used to classify tumours using support vector machine (SVM) classifier which classifies the breast cancer.
This paper discuses about the automatic Electroencephalogram (EEG) artifact removal using Independent Component Analysis (ICA) and Image Processing Algorithms. ICA is used for obtaining the Independent Components (IC's) that are linear in nature from the EEG signal. These IC's also known as topoplots are used for feature extraction and classification. Local binary pattern (LBP) is being utilized to obtain the features. These features are used for training the classifiers and helps in achieving automatic artifact elimination. Classifiers such as Linear Discriminant Analysis (LDA), K-Nearest Neighbour (KNN), and Support Vector Machine (SVM) are used for identifying the artifacts and thereby to remove such signals. The remaining IC components are used for reconstructing the de-noised signal. Of the existing methods mentioned, with respect to artifact elimination methods, LDA gives the best performance in artifact removal and helps in reconstruction of the signal.