Predictive impact of T2-MRI radiomics model on initial diagnosis of bone metastasis in prostate cancer patients

The purpose of this study was to examine the potential predictive impact of the T2-MRI radiomics model on the initial diagnosis of bone metastasis in patients with prostate cancer (PCa).

We retrospectively analyzed a total of 141 patients with confirmed PCa from clinical pathology records. Among them, 52 cases had bone metastasis and 89 cases did not. By employing a computer, the patients were randomly assigned to either a training group or a test group. Using ITK-SNAP software, we manually outlined T2WI images for all patients and performed radiomic analysis using Analysis Kit (AK) software. A total of 396 tumor texture features were extracted. In the training group, a single-variable t-test was conducted to identify features strongly associated with PCa bone metastasis. Statistical significance was defined as P < 0.05. After dimensionality reduction, the Lasso model was employed to select the best subset, and a random forest model was established. To evaluate the performance of the radiomics model in predicting PCa bone metastasis in the test group, receiver operating characteristic (ROC) curves and confusion matrices were utilized.

The selected imaging features exhibited a significant correlation with the differential diagnosis of prostate cancer presence or absence of metastasis. The radiomic model demonstrated high predictive efficiency for PCa bone metastasis, achieving accuracy rates of 0.81% and 0.85% in the training and test groups, respectively. The sensitivities were 92% and 93%, and the specificities were 85% and 81%. The area under the curve values were 0.88 and 0.80 for the training and test groups, respectively.

The MRI radiomics method based onT2WI images shows promise in accurately predicting PCa bone metastasis and can serve as a valuable tool for developing clinical treatment plans.