MRI Radiomics: Tumor Identification and Reverse Image Search Using Convolutional Neural Network

This was originally published in February 2021

Abstract

Magnetic Resonance Imaging (MRI) is commonly used in the diagnosis of brain tumors. Traditionally, physicians are trained to use their eyes when interpreting the scans; as the old saying goes, "a radiologist looking for a ruler is a radiologist in trouble." The recent emergence of the field of radiomics – which quantifies tumor features in MRI scans to create mineable high-dimensional data – is changing the game and creating new clinical decision support tools for identifying cases with tumor similarities. This paper leverages deep learning to match MRI tumor images to scans with similar features with the goal of providing clinicians a tool to increase accuracy and speed to diagnosis. The methodology presented herein involves combining multiple artificial neural networks in order to identify tumor regions and then match the tumor in question to those with the highest degree of similarities.