Research Area: Machine Learning
Cancer is a group of diseases caused by the abnormal and disorderly growth of cells, representing the second leading cause of deaths worldwide. The number of cancer cases is growing yearly, medical systems are an essential tool to speed up the diagnosis process and increase patient survival probabilities. Electronic health record systems store the patient’s health data, which can be of structured and unstructured types. Physicians use all the information available in these systems during a cancer diagnostic, regardless of its sort and modality. Deep learning is a machine learning sub-field that has algorithms able to process data end-to-end using deep architectures, often inspired by the brain’s synaptic model. Using more than one data source in these architectures is known as multimodal deep learning. Computer-aided detection and diagnosis systems developed using multimodal deep learning algorithms have been achieving promising results in diagnostic performance, which are often comparable to human specialists. E-health and telemedicine applications can be boosted with these high-performance detection and diagnosis systems, which can provide real-time analysis capabilities to the healthcare staff and improve the quality of the medical services. This chapter explores the theory and the applications of multimodal deep learning techniques to develop computer-aided detection and diagnosis systems for cancer, describing examples of the state-of-the-art in this area, looking for essential and innovative aspects of these systems. Heterogeneous and hybrid fusion strategies combining high-quality imaging, clinical attributes and genetic data in deep multimodal architectures are resulting in computer-aided systems for cancer with promising diagnosis performance.
Keywords:
Author(s) Name: Alan Baronio Menegotto & Silvio Cesar Cazella
Journal name: Enhanced Telemedicine and e-Health
Conferrence name:
Publisher name: Springer
DOI: 10.1007/978-3-030-70111-6_13
Volume Information: pp 267–287
