Predicting Mucosal Healing in Crohn’s Disease: A Deep Learning Approach Using Intestinal Ultrasound
Predicting Mucosal Healing in Crohn’s Disease: A Deep Learning Approach Using Intestinal Ultrasound
Predicting treatment response in Crohn’s disease (CD) remains a significant challenge, hindering the optimization of therapeutic regimens. This study presents the development and validation of a novel deep learning model designed to predict mucosal healing in CD patients based on pretreatment intestinal ultrasound (IUS) images and clinical data. The inherent variability in patient response to medication necessitates more precise predictive tools, and this research offers a potential advancement in personalized medicine for CD.
The retrospective study utilized a cohort of 190 CD patients (68.9% male, mean age 32.3 ± 14.1 years) from a tertiary hospital. A total of 1548 IUS images, encompassing both longitudinal diseased bowel segments exhibiting mucosal healing and those without, were analyzed. These images were divided into training and test cohorts to develop and validate a convolutional neural network (CNN) model designed to predict mucosal healing after one year of standardized treatment. Rigorous internal validation was performed using five-fold cross-validation.
The performance of the deep learning model was evaluated using several key metrics. In the test cohort, the model achieved a mean area under the curve (AUC) of 0.73 (95% CI: 0.68-0.78). Sensitivity was 68.1% (95% CI: 60.5-77.4%), specificity was 69.5% (95% CI: 60.1-77.2%), positive predictive value was 80.0% (95% CI: 74.5-84.9%), and negative predictive value was 54.8% (95% CI: 48.0-63.7%). Analysis of heat maps generated by the model revealed that the decision-making process primarily considered information derived from the bowel wall, serous surface, and surrounding mesentery, highlighting the model’s ability to extract subtle, clinically relevant features from the IUS images.
This study demonstrates the potential of deep learning in predicting mucosal healing in CD using IUS imaging. The model’s notable accuracy, although requiring further validation and improvement, represents a significant step towards personalized treatment strategies. Future research focusing on multi-center studies and the incorporation of larger, real-world datasets will be crucial in refining the model’s predictive capabilities and enhancing its clinical applicability. The ability to accurately predict treatment response holds immense promise for optimizing patient care and improving outcomes in Crohn’s disease management.
Disclaimer: This content is aggregated from public sources online. Please verify information independently. If you believe your rights have been infringed, contact us for removal.