Advancing radiomics research translation through a public database

Excerpt

Radiomics as a member of the -omics family, has witnessed a remarkable surge in interest among researchers over the past decade, resulting in an exponential increase in publications [1]. Radiomics, often referred to as texture analysis, is based on the principle of extracting previously concealed medical imaging characteristics, holding great promise across a wide range of applications [2]. Despite this growing trend, the translation of radiomics into clinical practice has not yet eventuated, primarily due to widespread uncertainty regarding its clinical value. One identified significant barrier is the challenge of reproducibility when attempting to apply the same methods and datasets [3‐6]. This lack of scientific reproducibility stems from the intricate radiomic pipeline that includes a multitude of steps similar to an advanced mathematical formula that contains only a few “variables” however, with each “variable” signifying multiple complicated calculation steps. In the context of radiomic studies, these “variables” are the complex methods and data involved. Designing and conducting a robust radiomic study requires multidisciplinary expertise spanning clinical, biostatistical, and computer science domains. The subsequently published study pipeline must include, among other things, the sharing of details related to the utilised software tools, code, and especially technical imaging acquisition and reformatting parameters [3‐6]. This comprehensive sharing is essential to enable future validation and testing of the findings. …