A machine learning-based unenhanced computed tomography (CT) texture analysis may be a useful tool for predicting BRCA1-associated protein I (BAP1) mutations status in patients with clear cell renal cell carcinoma (ccRCC), according to a recent study.
Previous research has shown that a BAP1 mutation is an unfavorable factor for survival in patients with clear cell RCC; however, existing literature on BAP1 mutation lacks studies that consider the reliability of texture features in their workflow.
This study attempted to conduct texture analysis on samples of RCC. Texture analysis is a quantitative image processing method that identifies repetitive patterns that may not be perceptible with the human eye.
“Recognizing molecular genetics of ccRCCs holds promise to classify patients more accurately, offering a better prediction of patient prognosis and personalized treatment strategies,” the researchers wrote.
They used texture features with a high interobserver agreement to develop and validate a machine learning-based radiomic model to predict BAP1 mutations status. They analyzed 65 patient samples.
Out of 744 textures features identified, 468 had an excellent interobserver agreement. Using selected features, a random forest plot correctly classified 84.6% of the labelled slices for BAP1 mutation status; the area under the receiver operating characteristic curve was 0.897.
For predicting clear cell RCCs with BAP1 mutation, sensitivity was 90.4%, specificity was 78.8%, and precision was 81%. For predicting clear cell RCCs without BAP1 mutation, the sensitivity was 78.8%, specificity 90.4%, and precision 89.1%.
The researchers acknowledged the small number of patients in the study and its retrospective design as potential limitations.
Kocak B, Durmaz ES, Kaya OK, Kilickesmez O. Machine learning-based unenhanced CT texture analysis for predicting BAP I mutation status of clear cell renal cell carcinomas [published online October 21, 2019]. Acta Radiol. doi: 10.1177/0284185119881742