Radiomics-Based Prediction of Surgery in Ulcerative Colitis Resistant to Drugs
Overview
This study utilized radiomics features extracted from admission CT images combined with clinical data to predict the need for surgical intervention in patients hospitalized with ulcerative colitis (UC) relapse. Machine learning models demonstrated promising accuracy in distinguishing patients requiring surgery from those managed medically.
Background
Ulcerative colitis is a chronic inflammatory bowel disease characterized by recurrent colon inflammation and relapses despite pharmacological treatment. Although advances in drug therapies and endoscopic monitoring have improved management, 8–24% of patients still require surgery, often due to refractory disease or complications. Determining the optimal timing for surgery during relapse remains challenging, as decisions rely on clinical judgment and laboratory data. Radiomics, an emerging imaging analysis technique extracting high-dimensional features from medical images, has shown potential in predicting treatment outcomes in various diseases but its utility in UC remains under investigation.
Data Highlights
Parameter
Details
Patient Cohort
157 CT scans from UC relapse hospitalizations (2015-2022)
Imaging
Noncontrast pelvic CT, rectal wall ROI segmentation
Radiomics Features
93 features extracted (excluding shape features)
Modeling
LASSO with tenfold cross-validation; 80% training, 20% validation
Outcome
Surgery during same hospitalization vs medical treatment alone
Key Findings
Radiomics features from admission CT images can predict surgical intervention in UC relapse patients.
Integration of radiomics scores with clinical factors improved predictive accuracy.
The study used a robust machine learning approach (LASSO) for feature selection and model building.
Manual segmentation of the rectal wall ROI was performed by experienced gastrointestinal surgeons to ensure data quality.
The predictive model showed good performance metrics (AUC, sensitivity, specificity) in both training and validation cohorts.
Clinical Implications
Radiomics analysis of CT images at admission offers a noninvasive tool to aid clinicians in identifying UC patients at high risk for requiring surgery during relapse. This approach may support earlier surgical decision-making and personalized treatment planning, potentially improving patient outcomes. Incorporating radiomics with clinical data enhances prediction beyond conventional assessments.
Conclusion
Radiomics combined with clinical parameters provides a promising method to predict surgical needs in UC patients resistant to pharmacological therapy, facilitating more informed and timely treatment decisions.
References
Keio University School of Medicine 2022 -- Predicting Surgical Intervention in Ulcerative Colitis Resistant to Pharmacological Treatment Using Radiomics
