Radiopathomics strategy combining multiparametric MRI with whole-slide image for pretreatment prediction of tumor regression grade to neoadjuvant chemoradiotherapy in rectal cancer
Lizhi Shao,Zhenyu Liu,Lili Feng,Xiaoying Lou,Zhenhui Li,Xiao-Yan Zhang,Xuezhi Zhou,Kai Sun,Da-Fu Zhang,Lin Wu,Guanyu Yang,Ying-Shi Sun,Ruihua Xu,Xiangbo Wan,Xinjuan Fan,Jie Tian
DOI: https://doi.org/10.1158/1538-7445.AM2020-2014
IF: 11.2
2020-01-01
Cancer Research
Abstract:Abstract Backgrounds: Tumor regression grade (TRG) reflects the chemoradiosensitivity of rectal cancer patients undergone neoadjuvant treatment, and relates to distinct probabilities of cancer recurrence and survival. However, in clinical practice, it is significantly challenging to rely solely on radiographic or clinical diagnostic information to obtain a patient's pathological response pre-treatment for treatment optimization. The aim was combining both the tumor information of macroscopic radiological and microscopic pathological images to develop and validated a more accurate signature for pretreatment prediction of TRG to neoadjuvant chemoradiotherapy (nCRT) in patients with locally advanced rectal cancer (LARC) on a multicenter dataset. Materials and Methods: We prospectively enrolled 981 patients (303 in the primary cohort, 678 in three external validation cohort) with clinicopathologically confirmed LARC treated with nCRT followed by surgery between Aug 2007 and Nov 2017, and collected their pretreatment multi-parametric MRI (mp-MRI), whole-slide image (WSI) of biopsy specimen, and pathologic response according to AJCC TRG system as well as clinical outcomes. Briefly, an artificial intelligence model integrating quantitative imaging features of mp-MRI and WSI was proposed to predict each nCRT treated patient into a particular 4-category TRG. The signature from the model (hereafter Rp-Grade) was further assessed in three independent validation cohort. Results: Rp-Grade yielded an overall ACC of 87.76% and revealed significant improvement than signature generating from model with mp-MRI or WSI features alone in all validation cohorts (P<0.0.5) (Table). Performance of radiopathomics signature for predicting 4-category AJCC/CAP TRGMetrics (%) [95% CI]Validation Cohort1 (N=480)Validation Cohort2 (N=150)Validation Cohort3 (N=48)Accuracy87.76 [84.86-90.66]79.71 [72.52-84.9]81.24 [70.15-92.34]PPV (TRG=0)94.58 [90.48-98.69]93.52 [84.68-100.0]83.09 [61.19-100.0]PPV (TRG=1)92.02 [86.19-97.85]70.26 [56.42-84.1]70.26 [56.42-84.1]PPV (TRG=2)83.9 [79.6-88.19]74.24 [63.57-84.92]71.49 [52.04-90.94]PPV (TRG=3)-92.54 [78.21-100.0]-Sensitivity (TRG=0)96.49 [93.03-99.95]96.62 [89.92-100.0]91.3 [74.73-100.0]Sensitivity (TRG=1)96.49 [93.03-99.95]75.57 [61.75-89.4]63.39 [42.96-83.82]Sensitivity (TRG=2)97.16 [95.02-99.29]100.0 [100.0-100.0]100.0 [100.0 100.0]Sensitivity (TRG=3)-35.55 [19.99-51.12]-Note: The evaluation results were represented by ‘-' when the number of categories was insufficient in the independent cohort test. Conclusions: Combining the macroscopic radiological information of the whole tumor and microscopic pathological information of local lesions from biopsy, radiopathomics was a potential strategy for the pretreatment prediction of the TRG in patients with LARC who underwent nCRT. Citation Format: Lizhi Shao, Zhenyu Liu, Lili Feng, Xiaoying Lou, Zhenhui Li, Xiao-Yan Zhang, Xuezhi Zhou, Kai Sun, Da-Fu Zhang, Lin Wu, Guanyu Yang, Ying-Shi Sun, Ruihua Xu, Xiangbo Wan, Xinjuan Fan, Jie Tian. Radiopathomics strategy combining multiparametric MRI with whole-slide image for pretreatment prediction of tumor regression grade to neoadjuvant chemoradiotherapy in rectal cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2014.