Assessing Lymph Node Count and Volume on Magnetic Resonance Imaging to Predict Prognosis Before Rectal Cancer Surgery: A Study to Develop an Artificial Intelligence Tool for Lymph Node Detection
Every year, thousands of rectal cancer patients undergo gruelling chemoradiotherapy before surgery—only to discover later that their tumours were low-risk and the treatment was unnecessary. Current guidelines cannot reliably distinguish low-risk from high-risk rectal cancers before treatment. This means low-risk patients receive toxic chemoradiotherapy or radiotherapy that causes permanent side effects—bladder and bowel damage, sexual dysfunction—with no survival benefit. Doctors lack a simple, pre-treatment marker to identify which patients will do well without aggressive therapy. This study tests a new marker: the number and total volume of lymph nodes visible on a pre-operative MRI. In pathology, more lymph nodes correlate with a stronger immune response and better survival. The researchers will analyse 1,400 international MRI scans to see if higher lymph node counts and volumes on imaging also predict better outcomes. They will then check whether these MRI findings correspond to actual immune cell activity in 300 tumour samples. Finally, they will build an AI tool to automatically count and measure lymph nodes on MRI, since manual counting is too labour-intensive for routine use. If successful, this could give surgeons a simple, non-invasive way to identify low-risk patients who can safely skip pre-surgical chemoradiotherapy—sparing them irreversible harm, reducing NHS costs, and enabling truly personalised rectal cancer care.
View original technical description
Background Currently, national guidelines recommend neoadjuvant treatment (before surgery) for high-risk rectal cancers. However, they do not accurately classify low- and high-risk rectal cancers. Therefore, genuinely low-risk patients are subjected to unwarranted neoadjuvant chemoradiotherapy and/or radiotherapy with significant irreversible side effects with a detrimental impact on quality of life. Currently, there is no consensus on the prognostic markers for rectal cancer before treatment. Lymph node (LN) size and count on pathology are described as surrogate markers for a superior anti-tumour immune response and enhanced survival. The study will use LN count and volume on the staging (pre-operative) Magnetic Resonance Imaging (MRI) to improve risk stratification and identify patients with a better prognosis before rectal cancer surgery. Since counting LNs on MRIs is labour-intensive, using artificial intelligence (AI) will enhance efficiency. Aims Demonstrate that a higher LN count and volume detected on the staging MRI is associated with improved survival. Demonstrate that a higher LN count and volume on the staging MRI is associated with a more significant immune cell infiltration. Develop an AI tool to obtain LN count and volume on the staging MRI. Methods A diverse group of previously treated rectal cancer patients of varying ages, sexes, and ethnicities was established to identify research priorities and advise on the study design. The study consists of three work programmes (WP): WP-1: Adult patients with rectal adenocarcinoma who have undergone surgery with and without neoadjuvant treatment will be selected from an international and diverse database of MRIs (n=1400) performed for rectal cancer staging. LNs on the MRI will be segmented to obtain the count and overall volume. Their impact on the 5-year overall and disease-free survival will be evaluated using a multivariable Cox regression model to adjust for existing prognostic clinical variables available before surgery. WP-2: Pathology samples from patients (n=300) who have previously undergone rectal cancer surgery without neoadjuvant treatment will be included. Immunohistochemistry will be conducted to obtain the density of CD3+ and CD8+ lymphocytes. The corresponding staging MRIs will be reviewed to ascertain the LN count and volume. Linear regressions will be performed to detect the correlation between high lymphocyte density and LN count and volume, the latter being the dependent variables. WP-3: Segmented MRIs from WP1 and WP2 (n=1700) will be used to develop an AI tool for automated detection of LNs to obtain the LN count and volume. The diagnostic accuracy will be assessed using an un-seen test cohort. Anticipated impact and dissemination A lead patient collaborator and advisory group will identify clinical applicability and develop lay summaries for dissemination. The evidence-based risk stratification model and AI tool will be shared with NHS policymakers, the NIHR imaging group and stakeholders. The study will directly impact patients and contribute to national guidelines for rectal cancer management to enable personalised care through enhanced risk stratification, improve quality of life by avoiding overtreatment of low-risk patients, and reduce NHS costs. The AI tool will enhance the efficiency of currently under-resourced radiology departments and align with the NHS long-term plan.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know