Diffuse Bowel Dilation as a Novel Diagnostic Marker for Jejunoileal Atresias: A Machine Learning Approach Integrating Risk Factors

Background Jejunoileal atresias (JIA) are congenital intestinal obstructions with a complex etiology, where timely surgery is crucial but prenatal diagnosis is challenging due to overlapping ultrasound signs with other disorders. This study aimed to identify key prenatal ultrasound and clinical predictors of JIA in fetuses with bowel dilation, with a focus on novel morphological markers. Methods We performed a retrospective analysis of cases from From January 1, 2018, to November 15, 2024, with fetal bowel dilation. Demographic information and ultrasound data were collected. The Lasso-logit model was employed to screen the final variables. Moreover, the marginal propensity diagrams were utilized to visually display the change in the probability of JIA occurrence with the variation of these variables. Results Dilation type, polyhydramnios, and bowel diameter were primary explanatory variables in the model, with odds ratios of 320.86, 108.85, and 55.09 respectively. JIA probability exceeds 50% when the normalized bowel diameter reaches 1.1 (about 34.4 mm actual diameter).With a specific bowel diameter, JIA probability is highest when both diffuse dilation and polyhydramnios exist, followed by only diffuse dilation, and lowest without both. Conclusions Diffuse bowel dilation, larger bowel diameter, and polyhydramnios are critical ultrasound markers for JIA. The proposed Lasso-Logit model highlights diffuse dilation as a novel, independent diagnostic criterion, addressing gaps in prenatal specificity.