Whole-tumor Multiparametric MRI Histogram Analysis for Predicting Intracranial Solitary Fibrous Tumor Grades and Ki-67 Expression

PURPOSE: Our aim was to explore the noninvasive prediction of intracranial solitary fibrous tumors (ISFTs) with World Health Organization (WHO) grade and Ki-67 expression level based on semantic and whole-tumor histogram features from multiparametric Magnetic Resonance Imaging (MRI). METHODS: We retrospectively evaluated thirty-nine ISFTs with histologically proved WHO grade 1 (n=25), WHO grade 2-3 (n=14), low Ki-67 (n=31), and high Ki-67 (n=8). Clinical data, MR semantic, and whole-tumor histogram features were collected. The values between the two groups were compared with Mann-Whitney U test or Fisher's exact test. Logistic regression analysis, receiver operating characteristic curve, and integrated discrimination improvement (IDI) were applied to identify the diagnostic performance. RESULTS: Hemorrhage and tumor-brain interface showed significant differences between WHO grade 1 and 2-3 groups (all P＜0.05). WHO grade 1 group showed significantly higher CE-T1WI_mean (contrast-enhanced T1-weighted imaging), CE-T1WI_median, and CE-T1WI_P90 values than grade 2-3 group (all P＜0.05). The combined model integrating tumor-brain interface and CE-T1WI_median achieved the highest area under the receiver operating characteristic curves (AUC) of 0.84, which was superior to each single model with improved IDI (IDI=0.15-0.21). Cystic degeneration and midline shift showed significant differences between high and low Ki-67 groups (all P＜0.05). High Ki-67 group was showed lower ADC_mean (apparent diffusion coefficient), ADC_median, ADC_P90, and T2WI_skewness (T2-weighted imaging) values than low Ki-67 group (all P＜0.05). The combined model integrating ADC_median and T2WI_skewness achieved the highest AUC of 0.94 and outperformed each single model with improved IDI (IDI=0.34-0.43). CONCLUSION: The combination of MR semantic and whole-tumor histogram features can predict ISFT grades and Ki-67 expression with favorable predictive performance.