[Radiomics in Practice and Its Basic Theory for Neurosurgeons].

Abstract

Medical images, including magnetic resonance imaging scans, are composed of numerical data, making them well-suited for machine learning and statistical approaches such as deep learning and radiomics. While qualitative analysis of neurological images may have been sufficient for research a decade ago, current standards increasingly demand some level of quantitative analysis. Although the term "radiomics" may imply complex mathematical processing or advanced programming, its foundational concepts are surprisingly accessible, with origins tracing back to 1973. The mathematical formulas used in radiomic feature are generally within the scope of high school-level mathematics. This paper provides a framework for individuals keen on integrating radiomics into their analytical methodologies, structured in the following manner: In Section II a detailed, methodical example of the procedures involved in conducting radiomic analysis is provided. Section III provides a brief overview of the historical development of radiomics. Sections IV and V explore the two image feature concepts that underpin radiomics: the gray level co-occurrence matrix and the gray level run length matrix, providing readers a deeper understanding of the significance of the calculated image features.