Abstract
The past years have seen a considerable increase in cancer cases. However, a cancer diagnosis is often complex and depends on the types of images provided for analysis. It requires highly skilled practitioners but is often time-consuming and error-prone. If Machine Learning and deep learning algorithms have been widely used, a comprehensive review of the techniques used from the pre-processing steps to the final prediction is lacking. With this review, we aim to provide a comprehensive overview of the current steps required in building efficient and accurate machine learning algorithm for cancer prediction, detection and classification. To do so, we compile the results of cancer related study using AI over the past years. We include various cancers that encompass different types of images, and therefore different related techniques. We show that tremendous improvements have been made in the early detection of cancerous tumors and tissues. The techniques used are various and often problem-tailored and our findings is confirmed through the study of a large number of research. Moreover, we investigate the approaches best suited for different types of images such as histology, dermoscopic, MRI, etc. With this work, we summarize the main finding over the past years in cancer detection using deep learning techniques. We discuss the challenges of cancer research related to the large discrepancies in the images, and we provide some notable results in the field for lung, breast, and skin cancers.
Abstract (translated)
URL
https://arxiv.org/abs/2210.11943
