Zhenchao Tang1, Xiao-Yan Zhang2, Zhenyu Liu3, Xiao-Ting Li2, Yan-Jie Shi2, Shou Wang4, Mengjie Fang4, Chen Shen4, Enqing Dong5, Ying-Shi Sun6, Jie Tian7. 1. School of Mechanical, Electrical & Information Engineering, Shandong University, Weihai, China. 2. Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Department of Radiology, Peking University Cancer Hospital & Institute, Beijing, China. 3. CAS Key Laboratory of Molecular Imaging, Institute of Automation, Beijing, China; Beijing Key Laboratory of Molecular Imaging, Beijing, China. 4. CAS Key Laboratory of Molecular Imaging, Institute of Automation, Beijing, China. 5. School of Mechanical, Electrical & Information Engineering, Shandong University, Weihai, China. Electronic address: enqdong@sdu.edu.cn. 6. Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Department of Radiology, Peking University Cancer Hospital & Institute, Beijing, China. Electronic address: sys27@163.com. 7. CAS Key Laboratory of Molecular Imaging, Institute of Automation, Beijing, China; University of Chinese Academy of Sciences, Beijing, China; CAS Center for Excellence in Brain Science and Intelligence Technology, China. Electronic address: jie.tian@ia.ac.cn.
Abstract
BACKGROUND AND PURPOSE: Locally advanced rectal cancer (LARC) patients showing pathological good response (pGR) of down-staging to ypT0-1N0 after neoadjuvant chemoradiotherapy (nCRT) may receive organ-preserving treatment instead of total mesorectal excision (TME). In the current study, quantitative analysis of diffusion weighted imaging (DWI) is conducted to predict pGR patients in order to provide decision support for organ-preserving strategies. MATERIALS AND METHODS: 222 LARC patients receiving nCRT and TME are enrolled from Beijing Cancer Hospital and allocated into training (152) and validation (70) set. Three pGR prediction models are constructed in the training set, including DWI prediction model based on quantitative DWI features, clinical prediction model based on clinical characteristics, and combined prediction model integrating DWI and clinical predictors. Prediction performances are assessed by area under receiver operating characteristic curve (AUC), classification accuracy (ACC), positive and negative predictive values (PPV and NPV). RESULTS: The DWI (AUC = 0.866, ACC = 91.43%) and combined (AUC = 0.890, ACC = 90%) prediction model obtains good prediction performance in the independent validation set. Nevertheless, the clinical prediction model performs worse than the other two models (AUC = 0.631, ACC = 75.71% in validation set). Calibration analysis indicates that the pGR probability predicted by the combined prediction model is close to perfect prediction. Decision curve analysis reveals that the LARC patients will acquire clinical benefit if receiving organ-preserving strategy according to combined prediction model. CONCLUSION: Combination of quantitative DWI analysis and clinical characteristics holds great potential in identifying the pGR patients and providing decision support for organ-preserving strategies after nCRT treatment.
BACKGROUND AND PURPOSE: Locally advanced rectal cancer (LARC) patients showing pathological good response (pGR) of down-staging to ypT0-1N0 after neoadjuvant chemoradiotherapy (nCRT) may receive organ-preserving treatment instead of total mesorectal excision (TME). In the current study, quantitative analysis of diffusion weighted imaging (DWI) is conducted to predict pGR patients in order to provide decision support for organ-preserving strategies. MATERIALS AND METHODS: 222 LARC patients receiving nCRT and TME are enrolled from Beijing Cancer Hospital and allocated into training (152) and validation (70) set. Three pGR prediction models are constructed in the training set, including DWI prediction model based on quantitative DWI features, clinical prediction model based on clinical characteristics, and combined prediction model integrating DWI and clinical predictors. Prediction performances are assessed by area under receiver operating characteristic curve (AUC), classification accuracy (ACC), positive and negative predictive values (PPV and NPV). RESULTS: The DWI (AUC = 0.866, ACC = 91.43%) and combined (AUC = 0.890, ACC = 90%) prediction model obtains good prediction performance in the independent validation set. Nevertheless, the clinical prediction model performs worse than the other two models (AUC = 0.631, ACC = 75.71% in validation set). Calibration analysis indicates that the pGR probability predicted by the combined prediction model is close to perfect prediction. Decision curve analysis reveals that the LARC patients will acquire clinical benefit if receiving organ-preserving strategy according to combined prediction model. CONCLUSION: Combination of quantitative DWI analysis and clinical characteristics holds great potential in identifying the pGR patients and providing decision support for organ-preserving strategies after nCRT treatment.