A review of texture-centric diagnostic models for thyroid cancer using convolutional neural networks and visualized texture patterns

Abstract

Structural features observed in the microscopic domain (such as anatomical appearances in histopathology images) can serve as accurate indicators of many biological activities and medical conditions. However, the assessment of those features usually requires invasive procedures, such as biopsy. Developing predictive models for studying new biomarkers and features in noninvasive acquired medical imaging can help in early prediction and early intervention of diseases such as cancer. In this chapter, we focus on thyroid cancer and investigate the possibility to develop texture-centric noninvasive diagnostic models. We explore the feasibility to extract texture patterns for accurate cancer diagnosis. By doing this, we can develop accurate noninvasive computer-aided diagnosis systems to bridge the gap between radiomics and microscopic domains. © 2023 Elsevier Inc. All rights reserved.

Keywords Convolutional neural networks, Microscopic domains, Radiomics, Texture, Thyroid cancer