Whole-volume apparent diffusion coefficient histogram analysis for prediction of programmed cell death ligand 1...

To assess the possibility of employing whole-volume ADC histogram analysis for predicting programmed cell death ligand 1 (PD-L1) expression in periampullary carcinomas (PCs).

We retrospectively evaluated imaging records of 65 patients with PC who received pancreaticoduodenectomy in our hospital. PD-L1 expression was systematically categorized as positive or negative based on the tumor proportion score (TPS), immune cell score (ICS), and the combined positive score (CPS), with an immunohistochemistry assay. Univariate analysis was conducted to assess differences in parameters between PD-L1-positive and PD-L1-negative groups. Spearman's correlation analysis was utilized to explore associations between variables and PD-L1 expression. Receiver operating characteristic (ROC) analysis was performed to evaluate the differential diagnostic performance of parameters in distinguishing two groups.

Several ADC histogram parameters were significantly different between PD-L1-positive and PD-L1-negative group, and showed significant correlations with PD-L1 expression, most notably the 5th and 10th percentiles. In TPS grouping, the 5th percentile demonstrated the highest area under the curve (AUC) of 0.690, which was improved to 0.740 when combined with tumor size and carbohydrate antigen 19-9. In ICS grouping, the 10th percentile showed the highest AUC of 0.690, which was improved to 0.772 when integrated with the degree of differentiation. In CPS grouping, the 5th percentile demonstrated the highest AUC of 0.694, which was improved to 0.752 when combined with tumor size and carcinoembryonic antigen.

Whole-volume ADC histogram parameters of primary tumors hold great potential in predicting PD-L1 expression in PCs.