Radiomic features extracted from segmented tumor regions have shown great power in gene mutation prediction, while deep learning–based (DL-based) segmentation helps to address the inherent limitations of manual segmentation. We therefore investigated whether deep learning–based segmentation is feasible in predicting KRAS/NRAS/BRAF mutations of rectal cancer using MR-based radiomics. In this study, we proposed DL-based segmentation models with 3D V-net architecture. One hundred and eight patients’ images (T2WI and DWI) were collected for training, and another 94 patients’ images were collected for validation. We evaluated the DL-based segmentation manner and compared it with the manual-based segmentation manner through comparing the gene prediction performance of six radiomics-based models on the test set. The performance of the DL-based segmentation was evaluated by Dice coefficients, which are 0.878 ± 0.214 and 0.955 ± 0.055 for T2WI and DWI, respectively. The performance of the radiomics-based model in gene prediction based on DL-segmented VOI was evaluated by AUCs (0.714 for T2WI, 0.816 for DWI, and 0.887 for T2WI+DWI), which were comparable to that of corresponding manual-based VOI (0.637 for T2WI, P=0.188; 0.872 for DWI, P=0.181; and 0.906 for T2WI+DWI, P=0.676). The results showed that 3D V-Net architecture could conduct reliable rectal cancer segmentation on T2WI and DWI images. All-relevant radiomics-based models presented similar performances in KRAS/NRAS/BRAF prediction between the two segmentation manners.

Background: The prognosis prediction of locally advanced rectal cancer (LARC) was important to individualized treatment, we aimed to investigate the performance of ultra-high b-value DWI (UHBV-DWI) in progression risk prediction of LARC and compare with routine DWI. Methods: This retrospective study collected patients with rectal cancer from 2016 to 2019. Routine DWI (b = 0, 1000 s/mm²) and UHBV-DWI (b = 0, 1700 ~ 3500 s/mm²) were processed with mono-exponential model to generate ADC and ADCuh, respectively. The performance of the ADCuh was compared with ADC in 3-year progression free survival (PFS) assessment using time-dependent ROC and Kaplan-Meier curve. Prognosis model was constructed with ADCuh, ADC and clinicopathologic factors using multivariate COX proportional hazard regression analysis. The prognosis model was assessed with time-dependent ROC, decision curve analysis (DCA) and calibration curve. Results: A total of 112 patients with LARC (TNM-stage II-III) were evaluated. ADCuh performed better than ADC for 3-year PFS assessment (AUC = 0.754 and 0.586, respectively). Multivariate COX analysis showed that ADCuh and ADC were independent factors for 3-year PFS (P < 0.05). Prognostic model 3 (TNM-stage + extramural venous invasion (EMVI) + ADCuh) was superior than model 2 (TNM-stage + EMVI + ADC) and model 1 (TNM-stage + EMVI) for 3-year PFS prediction (AUC = 0.805, 0.719 and 0.688, respectively). DCA showed that model 3 had higher net benefit than model 2 and model 1. Calibration curve demonstrated better agreement of model 1 than model 2 and model 1. Conclusions: ADCuh from UHBV-DWI performed better than ADC from routine DWI in predicting prognosis of LARC. The model based on combination of ADCuh, TNM-stage and EMVI could help to indicate progression risk before treatment.

Background: Prognostic biomarkers in colorectal carcinoma (CRC) have an important role in therapeutic strategy. Studies have shown that high expression of Aquaporin (AQP) is associated with poor prognosis in a variety of human tumors. AQP is involved in the initiation and development of CRC. The present study aimed to investigate the correlation between the expression of AQP1, 3 and 5 and clinicopathological features or prognosis in CRC.Methods: The AQP1, 3 and 5 expressions were analyzed based on the immunohistochemical staining of tissue microarray specimens including 112 patients with CRC between June 2006 and November 2008. The expression score of AQP (Allred_score and H_score) was digitally obtained with Qupath software. Patients were divided into high or low expression subgroups based on the optimal cut-off values. The relationship between expression of AQP and clinicopathological characteristics were evaluated using chi-square test, t-test, or one-way ANOVA, when appropriate. Survival analysis of 5-year progression free survival (PFS) and overall survival (OS) was performed with time-dependent ROC, Kaplan-Meier curves, univariate and multivariate COX analysis.Results: The AQP1, 3 and 5 expressions were associated with regional lymph node metastasis, histological grading, and tumor location in CRC, respectively (p < 0.05). Kaplan-Meier curves showed that patients with high AQP1 expression had worse 5-year PFS than those with low AQP1 expression (Allred_score: 47% vs. 72%, p = 0.015; H_score: 52% vs. 78% p = 0.006), as well as 5-year OS (Allred_score: 51% vs. 75%, p = 0.005; H_score: 56% vs. 80%, p = 0.002). Multivariate Cox regression analysis indicated that AQP1 expression was an independent risk prognostic factor (p = 0.033, HR = 2.274, HR95% CI: 1.069–4.836). There was no significant correlation between the expression of AQP3 and 5 and the prognosis.Conclusion: The AQP1, 3 and 5 expressions correlate with different clinicopathological characteristics and the AQP1 expression may be a potential biomarker of prognosis in CRC.