Radiomics: The endocrinologists’ new best friend?

Abstract

We have read with great interest the study by Shi et al. investigating adipose tissue textures in patients with metabolic diseases and weight loss after metabolic surgery [1]. They used abdominal computer tomography (CT) slices to assess volume and textures of visceral and subcutaneous adipose tissue. Using machine learning and neuronal networks to identify clinical and CT-based markers (radiomics), Shi et al were able to identify patients developing metabolic disease with a high predictive value. Furthermore, a combination of different radiomic markers were able to predict weight loss after bariatric surgery. The most important radiomic parameter identified was “runentropy”, which is defined as “the uncertainty/randomness in the distribution of run lengths and gray levels”. This study investigated an important and relevant topic. Traditional parameters used for metabolic health such as weight, body mass index (BMI) and others are unreliable and do not accurately predict survival and development of metabolic diseases. Sharma et al. showed years ago that BMI is a poor predictor for survival and therefore proposed the Edmonton Obesity Staging System to identify patients at risk for detrimental outcomes due to obesity associated diseases [2,3]. Similarly, several other studies showed that obese patients can be metabolically healthy while lean patients can have a high cardiovascular risk [4,5]. Therefore, it is of paramount interest for the treatment of metabolic diseases to reliably identify the patients having the highest risk for cardiovascular events or development of microvascular complications and, consequently, benefiting the most from an early intervention. This study presents an important step in this direction. However, there are also several points that must be addressed in future studies. The patients in this study were relatively healthy, even the patients in the group with obesity and metabolic syndrome. Average HbA1c levels were normal in the obese patients and there were no separate data on patients with more severe metabolic disease. Similarly, while the authors mention that radiomics was able to