Artificial Intelligence for Improved Hepatosplenomegaly Diagnosis

This study developed convolutional neural networks to automatically segment the liver and spleen from CT scans, achieving near-perfect agreement with manual measurements. The models produced highly accurate volumetric estimates, revealing significant differences in organ sizes between males and females, which required sex-specific thresholds for defining enlargement. Compared to traditional radiologist assessments, which had moderate sensitivity but high specificity, the CNN-derived measurements show promise for enhancing the accuracy of diagnosing hepatomegaly and splenomegaly.

Access full article here