J-K Wang1. 1. Department of Medical Equipment, Henan Provincial Hospital (Henan Children's Hospital), Zhengzhou, Henan Province, China. jkwang4@sina.com.
Abstract
OBJECTIVE: By analyzing MSCT (multi-slice computed tomography) signs with different risks in gastrointestinal stromal tumors, this paper aimed to discuss the predictive value and modeling analysis of MSCT signs in GISTs (gastrointestinal stromal tumor) to pathological risk degree. PATIENTS AND METHODS: 100 cases of primary GISTs with abdominal and pelvic MSCT scan were involved in this study. All MSCT scan findings and enhanced findings were analyzed and compared among cases with different risk degree of pathology. Then GISTs diagnostic model was established by using support vector machine (SVM) algorithm, and its diagnostic value was evaluated as well. RESULTS: All lesions were solitary, among which there were 46 low-risk cases, 24 medium-risk cases and 30 high-risk cases. For all high-risk, medium-risk and low-risk GISTs, there were statistical differences in tumor growth pattern, size, shape, fat space, with or without calcification, ulcer, enhancement method and peritumoral and intratumoral vessels (p<0.05). However, there were no statistical differences in the location of tumor and CT value at each period (plain scan, arterial phase, venous phase) (p>0.05). The apparent difference lied in plain scan, arterial phase and venous phase for each risk degree. The diagnostic accuracy of SVM diagnostic model established with 10 imaging features as indexes was 70.0%, and it was especially reliable when diagnosing GISTs of high or low risk. CONCLUSIONS: Preoperative analysis of MSCT features is clinically significant for its diagnosis of risk degree and prognosis; GISTs diagnostic model established on the basis of SVM possesses high diagnostic value.
OBJECTIVE: By analyzing MSCT (multi-slice computed tomography) signs with different risks in gastrointestinal stromal tumors, this paper aimed to discuss the predictive value and modeling analysis of MSCT signs in GISTs (gastrointestinal stromal tumor) to pathological risk degree. PATIENTS AND METHODS: 100 cases of primary GISTs with abdominal and pelvic MSCT scan were involved in this study. All MSCT scan findings and enhanced findings were analyzed and compared among cases with different risk degree of pathology. Then GISTs diagnostic model was established by using support vector machine (SVM) algorithm, and its diagnostic value was evaluated as well. RESULTS: All lesions were solitary, among which there were 46 low-risk cases, 24 medium-risk cases and 30 high-risk cases. For all high-risk, medium-risk and low-risk GISTs, there were statistical differences in tumor growth pattern, size, shape, fat space, with or without calcification, ulcer, enhancement method and peritumoral and intratumoral vessels (p<0.05). However, there were no statistical differences in the location of tumor and CT value at each period (plain scan, arterial phase, venous phase) (p>0.05). The apparent difference lied in plain scan, arterial phase and venous phase for each risk degree. The diagnostic accuracy of SVM diagnostic model established with 10 imaging features as indexes was 70.0%, and it was especially reliable when diagnosing GISTs of high or low risk. CONCLUSIONS: Preoperative analysis of MSCT features is clinically significant for its diagnosis of risk degree and prognosis; GISTs diagnostic model established on the basis of SVM possesses high diagnostic value.