Abstract: Cancer is a common disease that has caused fatalities among all age groups worldwide causing thousands of deaths each year. It is, therefore, necessary to diagnose cancer at an early stage. Deep learning has been proven pivotal for the early detection of cancer. This project uses deep convolutional neural networks to classify cancer in medical images belonging to four common cancers. This project is an effort to apply deep learning for cancer detection using both custom made DCNN models and pre-trained models. Images are analyzed using various edge detection algorithms. Data augmentation has been employed on the images. The four cancers image datasets used for this project are breast cancer histopathological images, brain MRI images, lungs CT scan images and skin lesion images. The performance of the proposed models is compared based on classification accuracy, precision, recall and f-score. After a comparison of the performances of the model, the model with the best performance will be deployed using flask REST API. This is an attempt to make the use of deep learning practical for a medical professional to diagnose cancer. This project is an attempt to bolster the previous research and development in the field of cancer diagnosis using deep learning.

Keywords: BrecanNet, BrainNet, Classification Accuracy, Deep Convolution Neural Network, LungNet, MelNet

PDF | DOI: 10.17148/IJARCCE.2022.11314

Open chat
Chat with IJARCCE