Cancer detection and segmentation using machine learning and deep learning techniques: a review

Cancer is the most fatal diseases in the world which has highest mortality rate as compared to other type’s human diseases. The most common and dangerous types of cancers are lung cancer, skin cancer, brain tumors, breast cancer, Colorectal cancer, Prostate Cancer, Blood cancer and many other. The millions of person lose their life due to these highly dangerous, fatal types of disease. Hence it is required to provide the solution using computer added automatic cancer detection technique in early stage, for the research gap analysis is required. In this paper we have studied the various cancer detection techniques based on traditional machine learning (ML) and deep learning (DL) techniques and summarize the research gap for the various cancers detection techniques. The study has been conducted based the types of technique uses, types of features utilized, dataset used and accuracy of the cancer detection achieved using best technique. In this study we have conducted the reviewed over 100 recently published research papers and focused on 7 types of most fatal cancer types such as lung cancer, breast cancer, skin cancer, brain tumor, colorectal, prostate, and Leukemia (Blood cancer). The study also used the state-of-art table to compare the previous and current study conducted on cancer detection techniques. We have visualized using separate comparison table for 7 types of cancer detection using traditional ML method and DL methods also visualized through the bar chart. The best accuracy result obtained using ML and DL methods are 100% and the most commonly used ML classifier is Support Vector Machine whereas CNN is most commonly used DL classifier. The main challenges we observed are the data imbalance issue, varieties of feature extraction techniques, small medical dataset, classifier parameters optimizations, Execution time, Adaptive classifier, and common technique for segmentation and classification. The main objective of this review is to investigate the existing methods used for the various types of cancer detection and finding the research gap, challenges, and recent advancement mainly in the use of ML and DL models which may help the researchers to find the better solutions.