Differentiating peritoneal tuberculosis and peritoneal carcinomatosis based on a machine learning model with CT: a multicentre study.

Purpose: It is still a challenge to make early differentiation of peritoneal tuberculosis (PTB) and peritoneal carcinomatosis (PC) clinically as well as on imaging and laboratory tests. We aimed to develop a model to differentiate PTB from PC based on clinical characteristics and primary CT signs.

Methods: This retrospective study included 88 PTB patients and 90 PC patients (training cohort: 68 PTB patients and 69 PC patients from Beijing Chest Hospital; testing cohort: 20 PTB patients and 21 PC patients from Beijing Shijitan Hospital). The images were analyzed for omental thickening, peritoneal thickening and enhancement, small bowel mesentery thickening, the volume and density of ascites, and enlarged lymph nodes (LN). Meaningful clinical characteristics and primary CT signs comprised the model. ROC curve was used to validate the capability of the model in the training and testing cohorts.

Results: There were significant differences in the following aspects between the two groups: (1) age; (2) fever; (3) night sweat; (4) cake-like thickening of the omentum and omental rim (OR) sign; (5) irregular thickening of the peritoneum, peritoneal nodules, and scalloping sign; (6) large ascites; and (7) calcified and ring enhancement of LN. The AUC and F1 score of the model were 0.971 and 0.923 in the training cohort and 0.914 and 0.867 in the testing cohort.

Conclusion: The model has the potential to distinguish PTB from PC and thus has the potential to be a diagnostic tool.