RPS 601a - Advances in rectal cancer imaging

Purpose:

To compare MR imaging morphology of rectal cancer with clinicopathologic features.

Methods and materials:

50 rectal MRI reports with the synoptic term “polypoid” tumour were reviewed retrospectively and 90 reports containing “circumferential” and “partially circumferential” tumours were randomly selected to serve as controls. Preoperative tumour (T) and lymph node (N) stages of the polypoid tumours were determined by two readers and compared to reported T-stages for the controls. To assess the difference in degrees of attachment to wall circumference (WC) (≤¼ or >¼) and the presence of a pedicle, 25 randomly selected polypoid tumours and 25 randomly selected controls were reviewed by two additional readers. Inter-reader agreement was assessed using the kappa statistics for binary measures, weighted kappa for ordinal measures, and nonparametric interclass correlation coefficient (ICC) for continuous measures. Features were compared between cases and controls using Fisher’s exact test for categorical variables and Wilcoxon rank-sum test for continuous variables.

Results:

Polypoid shape, pedicle presence, and degree of attachment showed moderate agreement (k=0.41-0.76). N size showed good agreement (ICC=0.83-0.91). Distinguishing characteristics of polypoid tumours were an attachment to < ¼ WC (p<0.001, 0.002) and presence of a pedicle (p<0.001, both) in significantly more cases compared to controls. Lower T and N stages, smaller largest node, and fewer number of nodes > 3mm were significantly associated with polypoid morphology compared to controls (p<0.001 to 0.008).

Conclusion:

Primary “polypoid” tumours at rectal MRI represent a distinct phenotype with lower pathologic T and N stages. If the above definition of “polypoid” is validated, this synoptic term may be more than descriptive and could potentially inform patient prognostication.

Limitations:

Retrosepctive study.

Ethics committee approval

IRB waiver of patient consent.

Funding:

NCI P30 CA008748 Cancer Center Support Grant.

Purpose:

To evaluate development of peritoneal carcinomatosis or other adverse short- or long-term outcomes in patients with MRI-based T4a rectal cancer.

Methods and materials:

In the 5 year period from 2013-2018, the clinico-pathologic records of all patients with pretreatment rectal MRI with reported clinical (c) T3c, T3d, T4a, and T4b primary rectal adenocarcinoma were retrospectively reviewed by one radiology fellow. All MRI reported features such as nodal stage and tumour height were collected. Pathologic T and N stages were summarised among patients undergoing curative resection. Recurrence-free survival (RFS) and overall survival (OS) were estimated using the Kaplan-Meier method. Differences in survival were compared using the log-rank test. Descriptive statistics were used to describe distant metastases.

Results:

256 patients were included with 28 clinical T4a, 135 T3cd, and 93 T4b patients reported at MRI. Median follow-up among survivors was 27 months (range 0.36-70 months). All patients underwent neoadjuvant treatment and, at surgical pathology, only 3 cases were pT4a. 9 patients developed peritoneal carcinomatosis (9/256 3.5%) who had been clinical T3cd (n=5/135, 3%), T4a (n=2/28, 7%), and T4b (n=2/93, 2%). 5 of the 9 patients underwent surgery and all had pT3 disease. RFS (p=0.065) and OS (p=0.1694) did not differ between cT3cd and cT4a nor between cT4a and cT4b. RFS was worse in cT4b compared with cT3cd (p=0.03).

Conclusion:

Peritoneal carcinomatosis occurs rarely in rectal cancer, even in cases diagnosed as cT4a on baseline MRI. Overall and recurrence-free survival in patients with MRI-based T4a was not worse than those with T3cd disease.

Limitations:

Retrospective study.

Ethics committee approval

IRB waived the need for consent.

Funding:

This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

Purpose:

To determine the accuracy of MRI radiomics to distinguish between post-treatment residual cellular and acellular mucinous rectal neoplasms.

Methods and materials:

At 5 institutions from the Rectal and Anal Disease Focused Panel of the Society of Abdominal Radiology, a retrospective study of all patients with rectal adenocarcinoma containing mucin on post-neoadjuvant CRT MRI and prior to surgery between 2003 and 2017 (n= 362) was performed. Patients with <75% mucin on MRI were excluded (n=308). Manual volumetric segmentation of mucinous masses was performed on Terra Recon software (Aquarius iNtuition Viewer v 4.4.13.P4) by one experienced radiologist. Radiomics and machine learning models were performed using CERR within MatLab (2017b) to generate 22 first order and 80 second order parameters. Statistical analysis was performed with SPSS (v.25) using ROC curves to determine significant individual parameters (AUC > 0.70).

Results:

54 patients, 30 males, mean age 58yr (39-75yr), and 24 females, mean age 58yr (32-87yr), had stage 0-2 and 3 disease (n= 41, 13, respectively). 54 patients underwent neoadjuvant CRT. Histopathology showed 10 acellular, mucinous, non-neoplastic masses and 44 mucinous adenocarcinomas. In univariate testing, 4/102 parameters were of significance, with the best performing radiomic model for neoplasm classification showing an AUC of 0.63 using a cubic support vector machine with 10-fold cross-validation.

Conclusion:

An MRI radiomic model performed modestly in distinguishing post-treatment acellular, non-neoplastic masses from residual mucinous rectal adenocarcinoma with an AUC 0.63. While not ready for clinical dissemination, such modeling may be a step forward for informing treatment decisions in these aggressive tumours.

Limitations:

Retrospective, a small number of cases.

Ethics committee approval

Waiver of consent from IRB.

Funding:

No funding was received for this work.