Brain Cancer Diagnostics in Images Based on Deep Learning

Abstract

Traditional methods for brain cancer diagnosis rely on manual interpretation of medical images, which have limitations such as subjectivity and human error. In recent years, the use of deep learning (DL) in brain cancer diagnosis has shown promising results, especially in detecting and classifying various types of brain tumors. This research aims to develop an accurate and effective method for brain cancer diagnosis in medical images using six deep learning techniques. The algorithms were trained on two large datasets from the global Kaggle website. The results of the diagnosis accuracy were mixed. When tested on the first dataset, CNN and CNN-GRU achieved 99%, while VGG19 achieved 94%, CNN-LSTM achieved 92%, CNN-SVM was somewhat convincing with 87% accuracy, and DNN lagged very low compared to the rest of the models with 61% accuracy. When applied to the second dataset, the results showed that CNN maintained its efficiency with 99% accuracy, and CNN-GRU failed when compared to the results achieved on the first dataset with 60% accuracy. CNN-LSTM, VGG19, and CNN-SVM models achieved 97%, 96%, and 95% results, respectively. In addition, DNN also did not achieve good results with an accuracy of 62%.