Year: 2026 | Month: January | Volume: 16 | Issue: 1 | Pages: 333-348
DOI: https://doi.org/10.52403/ijhsr.20260138
Accuracy of Magnetic Resonance Imaging in Staging of Rectal Cancer: A Systematic Literature Review
Mutaz Alkhammash
Department of Radiology, Al-Hada Armed Forces Hospital, Taif, Saudi Arabia.
Corresponding Author: Mutaz Alkhammash
ABSTRACT
Background: Accurate preoperative staging of rectal cancer is essential for selecting appropriate treatment pathways, including primary surgery, neoadjuvant therapy, and organ-preservation strategies. Magnetic resonance imaging (MRI) is widely used for local staging; however, its diagnostic performance varies across staging domains and clinical settings. Objective: This systematic literature review evaluated the diagnostic performance of MRI for rectal cancer staging and related prognostic assessments, with emphasis on baseline staging, post-neoadjuvant restaging, circumferential resection margin (CRM) assessment, and emerging radiomics/artificial intelligence (AI) approaches.
Methods: A systematic review of the literature was conducted using predefined eligibility criteria. Studies assessing MRI performance in rectal cancer staging were included, encompassing baseline T/N staging, restaging after neoadjuvant therapy (ypT/ypN), CRM/mesorectal fascia (MRF) involvement, extramural venous invasion (EMVI), and response assessment (mrTRG). Surgical histopathology or node-by-node histopathology served as reference standards where applicable; one narrative review was included for contextual synthesis. Risk of bias was assessed using QUADAS-2.
Results: Ten studies were included (predominantly retrospective), with sample sizes ranging from 70 to 5,539. Baseline T-staging performance was variable, with accuracy ranging from 55% to 92.2%, and evidence of overstaging in early disease. Baseline N-staging was consistently less reliable than T-staging (accuracy 65%–88.7%), with modest discrimination using routine MRI interpretation in a T3 cohort. Post-neoadjuvant restaging demonstrated moderate performance (ypT accuracy 74.4%; ypN accuracy 60.1%). MRI showed clinical utility for CRM/MRF assessment, particularly for excluding threatened margins (high negative predictive value), while AI models achieved high discrimination for pretreatment CRM prediction (AUC up to 0.953). Radiomics and AI approaches improved performance for selected tasks, including lymph node staging (AUC 0.876) and dichotomized T-staging (AUC 0.82). QUADAS-2 highlighted generally acceptable reference standards, with higher risk of bias concentrated in the index test domain for AI/radiomics studies.
Conclusion: MRI provides substantial value in rectal cancer staging, particularly for local assessment and CRM planning, but limitations persist in early T-staging, nodal evaluation, and post-treatment response. Radiomics and AI are promising but require external validation and standardized implementation.
Key words: rectal cancer; magnetic resonance imaging; staging; diagnostic accuracy; radiomics; artificial intelligence