The Role Of Pre-Treatment MRI And Biopsy Digital Pathology Images In Predicting Prognosis In Locally Advanced Rectal Cancer

Part Ⅰ The Role of Pre-Treatment MRI in Predicting Prognosis in Locally Advanced Rectal CancerObjective:Accurate assessment of recurrence risk in patients with locally advanced rectal cancer before treatment is crucial for formulating treatment strategies.This study aimed to use magnetic resonance imaging(MRI)technology to predict the prognosis of patients with locally advanced rectal cancer using MRI biomarkers,develop an MRI-based scoring system,provide supplementary information to the TNM staging system,and assist clinical decision-makers and patients in making informed riskbenefit discussions regarding treatment strategies.Materials and methods:Between September 2009 and June 2020,a total of 1287 patients with locally advanced rectal cancer from two cohorts,including Cohort 1 from five centers and Cohort 2,a retrospective collection of prospective data from NCT04271657,were included in this study.MRI images were obtained using a 1.5T or 3.0T MRI scanner at participating hospitals,and rectal cancer MRI biomarkers,including mrN staging,mrT staging,tumor deposits(mrTDs),extramural vascular invasion(mrEMVI),and circumferential resection margin(mrCRM),were manually evaluated.Our scoring system,mrTEC,included mrTDs,mrEMVI,and mrCRM.Each high-risk MRI biomarker was assigned one point,and the total score ranged from 0 to 3.KaplanMeier curves were plotted for 3-year disease-free survival(DFS)and 5-year overall survival(OS)to show differences in survival rates among groups,and P values were calculated using the log-rank test.Results:(1)The 3-year DFS and 5-year OS of patients with mrNl and mrTD-positive were significantly lower than those with mrNl and mrTD-negative(P<0.0001),and the 3-year DFS and 5-year OS of patients with mrN2 and mrTD-positive were significantly lower than those with mrN2 and mrTD-negative(P<0.0001).(2)The C-index of mrT staging for predicting DFS was 0.55(95%CI 0.52-0.58)and for OS was 0.57(95%CI 0.53-0.66).The C-index of mrCRM status for predicting DFS was 0.60(95%CI 0.57-0.64)and for OS was 0.64(95%CI 0.590.68).(3)The 3-year DFS of all patients in Cohort 1 was 79.6%(95%CI 77.2%82.1%),and the 5-year OS was 84.8%(95%CI 82.0%-87.6%).The 3-year DFS of patients with mrTEC total scores of 0 to 3 were 91.0%,79.5%,65.5%,and 44.0%,and the 5-year OS was 92.9%,87.1%,74.8%,and 44.5%,respectively.(4)The mrTEC score was correlated with neoadjuvant treatment efficacy in Cohort 1,and patients with TRG of 0 were often evaluated with lower marker scores and had marginal statistical significance(P=0.059).Similar trends were observed in patients in Cohort 2(P>0.05).Conclusion:(1)Ignoring the impact of tumor deposits on patients once lymph node metastasis is present,as recommended by the 8th edition TNM staging system,is completely unreasonable.We recommend that the presence of tumor deposits should be separately reported during the assessment of rectal cancer using magnetic resonance imaging.(2)Compared to mrT staging,mrCRM status has a higher prognostic value for locally advanced rectal cancer patients.Considering both mrT staging and mrCRM status is important in predicting the prognosis of locally advanced rectal cancer patients.(3)We have developed the mrTEC scoring system,which is an MRI-based scoring system used to identify high-risk and low-risk locally advanced rectal cancer patients.(4)The mrTEC scoring system has good predictive performance for the prognosis of locally advanced rectal cancer before treatment.It also has some predictive ability for neoadjuvant therapy efficacy and exhibits excellent prognostic ability for patients with the same tumor regression grade after surgery.Part Ⅱ The Role of Biopsy Digital Pathology Images in Predicting Prognosis in Locally Advanced Rectal CancerObjective:The aim of this study is to utilize artificial intelligence neural network technology and whole-slide digital pathology to automatically identify and quantify biopsy pathological indicators,for accurate prognosis prediction of locally advanced rectal cancer before treatment.Materials and methods:Patients with locally advanced rectal cancer were recruited from the clinical database of Sun Yat-sen University Cancer Center(SYSUCC)for this study,with a total of 473 patients included.Before treatment,H&E-stained biopsy slides were scanned using a digital whole-slide scanner(Leica,Aperio-AT2,USA).Artificial intelligence networks were used for tissue and cell segmentation of the pathological slides,and prognostic biomarkers were extracted at both the tissue level and immune cell level.At the tissue level,tumor mucin ratio,tumor necrosis ratio,and tumor stroma ratio were proposed.At the immune cell level,the ratio of neutrophil and lymphocyte infiltration in the tumor,as well as the ratio of neutrophil and lymphocyte infiltration in the tumor microenvironment were quantified.Kaplan-Meier curves were drawn for 3-year disease-free survival(DFS)and 5-year overall survival(OS)to demonstrate the differences in patient survival rates among the groups,and P values were calculated using the log-rank test.Cox multivariate analysis was used to calculate hazard ratios and 95%confidence intervals(CIs)to confirm the prognostic value of different variables.Results:(1)The tumor stroma ratio could predict 3-year DFS in patients.We used quartile grouping and chose the optimal threshold of TSR to be 0.609,dividing patients into high and low tumor stroma ratio groups.The DFS prognostic curve showed that high tumor stroma ratio was unfavorable for tumor prognosis(P=0.031).(2)Tumor-infiltrating neutrophils could predict 3-year DFS in patients.The optimal threshold was determined using "maxstat" in R software and was 93.304 cells/mm2.Patients were divided into high and low tumor-infiltrating neutrophil groups,and the DFS prognostic curve showed that high neutrophil infiltration was unfavorable for tumor prognosis(P=0.0098).(3)A Pathology Comprehensive Scoring System(Path-M)was constructed,with high tumor stroma ratio assigned 1 point and high tumor-infiltrating lymphocytes assigned 1 point,with a total score of 0-3 points.In multivariate analysis,age and Path-M pathology comprehensive scoring system were independent prognostic factors for DFS.As the Path-M score increased,patients had a trend of worse prognosis(P=0.041).Conclusion:(1)Tumor stromal ratio(TSR)in rectal cancer biopsy pathology is an independent prognostic factor for three-year disease-free survival at the tumor tissue level.Patients with a high TSR have a lower three-year disease-free survival rate.(2)Tumor-infiltrating neutrophils(TINs)in rectal cancer biopsy pathology are independent prognostic factors for three-year disease-free survival at the tumor immune cell level.Patients with a high TINs have a lower three-year disease-free survival rate.(3)This study used a fully automated intelligent method to quantitatively evaluate TSR and TINs and constructed a pathology comprehensive scoring system(Path-M).The scoring system assigns 1 point for high TSR and 1 point for high TINs,with a total score range of 0-3.Patients with higher scores have a lower three-year disease-free survival rate,especially those with both high TSR and high TINs.Part Ⅲ Predicting Prognosis in Locally Advanced Rectal Cancer Using mrTEC and Path-M Scoring SystemObjective:This study aimed to compare the prognostic performance of mrTEC scoring system based on magnetic resonance imaging(MRI)and Path-M pathology comprehensive scoring system based on biopsy pathology sections for predicting the prognosis of locally advanced rectal cancer.We also attempted to combine the two modalities to explore whether it could improve the prognostic performance of the model.Materials and methods:We recruited 473 patients with locally advanced rectal cancer from the clinical database of Sun Yat-sen University Cancer Center(SYSUCC).MRI images were acquired using a 1.5T or 3.0T MRI scanner at participating hospitals.The mrTEC scoring system,which included mrTDs,mrEMVI,and mrCRM,was used to score each patient by radiologists.Each high-risk MRI marker was assigned 1 point,and the total score ranged from 0 to 3 points.H&E-stained biopsy sections were scanned using a digital whole-slide scanner(Leica,Aperio-AT2,USA).AI network was used to segment the tissue and cell from the pathological sections and extract the prognostic markers from the tissue level and immune cell level.The tumor-stroma ratio and tumor-infiltrating neutrophils were extracted and evaluated using the PathM pathology comprehensive scoring system.High tumor-stroma ratio and high tumor-infiltrating neutrophils were assigned 1 point each,and the total score ranged from 0 to 3 points.Kaplan-Meier curves were plotted for 3-year DFS and 5-year OS to show the difference in patient survival rates among groups,and P values were calculated using the log-rank test.Cox multivariate analysis was used to calculate hazard ratios and 95%confidence intervals(Cls)to confirm the prognostic value of different variables.Results:(1)The C-index of the mrTEC scoring system for predicting DFS in patients was 0.64(95%CI 0.58-0.71),and that of the Path-M pathology comprehensive scoring system was 0.57(95%CI 0.51-0.64).The C-index of the combined mrTEC and PathM scoring systems for predicting DFS in locally advanced rectal cancer was 0.70(95%CI 0.64-0.76).(2)In multivariate analysis,age,mrTEC scoring system,and Path-M pathology comprehensive scoring system were independent prognostic factors for DFS.As the Path-M score increased,the prognosis of patients tended to worsen(P=0.041).Conclusion:(1)The mrTEC scoring system based on MRI had better prognostic performance than the Path-M pathology comprehensive scoring system based on biopsy pathology sections.(2)Both the mrTEC scoring system and Path-M pathology comprehensive scoring system constructed in this study were independent prognostic factors for locally advanced rectal cancer patients.They could predict the prognosis of patients more accurately from the aspects of magnetic resonance imaging and biopsy pathology,respectively,and provide better guidance for treatment decisions.