OBJECTIVE: This study aimed to evaluate the prediction roles of clinical plus imaging features and multidetector computed tomography (MDCT) texture analysis in preoperative risk grade classification of small bowel (SB) gastrointestinal stromal tumors (GISTs). METHODS: This study included 213 SB GIST patients. Clinical features and MDCT imaging findings were reviewed. Tumor risk stratifications were determined according to modified National Institutes of Health criteria. Random forest models were performed to evaluate the correlation of risk stratification. RESULTS: The model of clinical plus imaging findings showed an area under receiver operating characteristic curve (AUC) of 92.0%. The AUC of texture analysis based on MDCT portal phase was 93.3%, without statistical difference from that of clinical plus imaging model (P = 0.378). The AUC of the model combined clinical plus imaging features and MDCT texture analysis was 94.3%, which was significantly higher than the AUC of clinical imaging model (P = 0.042). CONCLUSION: Texture analysis may become an important comprehensive tool for preoperative risk stratification of SB GISTs.
OBJECTIVE: This study aimed to evaluate the prediction roles of clinical plus imaging features and multidetector computed tomography (MDCT) texture analysis in preoperative risk grade classification of small bowel (SB) gastrointestinal stromal tumors (GISTs). METHODS: This study included 213 SB GISTpatients. Clinical features and MDCT imaging findings were reviewed. Tumor risk stratifications were determined according to modified National Institutes of Health criteria. Random forest models were performed to evaluate the correlation of risk stratification. RESULTS: The model of clinical plus imaging findings showed an area under receiver operating characteristic curve (AUC) of 92.0%. The AUC of texture analysis based on MDCT portal phase was 93.3%, without statistical difference from that of clinical plus imaging model (P = 0.378). The AUC of the model combined clinical plus imaging features and MDCT texture analysis was 94.3%, which was significantly higher than the AUC of clinical imaging model (P = 0.042). CONCLUSION: Texture analysis may become an important comprehensive tool for preoperative risk stratification of SB GISTs.