Related Experiment Video
Updated: Jul 15, 2026
DIPLOMA Approach for Standardized Pathology Assessment of Distal Pancreatectomy Specimens
Published on: February 1, 2020
CT texture analysis and delta radiomics in predicting R0 resection after preoperative chemotherapy in resectable
Riccardo De Robertis1, Nicolò Cardobi1, Flavio Spoto2
1Department of Diagnostics and Public Health, Radiology Unit, University of Verona, G.B. Rossi Hospital, AOUI Verona, Verona, Italy.
Purpose:
To evaluate the predictive performance of CT-based radiomic delta-features for predicting R0 resection in resectable pancreatic ductal adenocarcinoma (PDAC) following NALIRIFOX neoadjuvant chemotherapy.
Materials And Methods:
This prospective study included 107 consecutive patients with resectable PDAC treated with NALIRIFOX. After applying strict selection criteria for imaging quality and timing, 55 patients underwent comprehensive radiomic analysis. Radiomic features were systematically extracted from pre-treatment and post-treatment CT scans, with delta-features calculated as percentage variations between timepoints. Features demonstrating inter-observer intraclass correlation coefficient greater than 0.90 in both arterial and venous phases were selected. Following correlation analysis to remove redundant features (absolute correlation greater than 0.80), three predictive models were developed based on pre-treatment, post-treatment, and delta-features. Model validation was performed using 1000-replication bootstrap methodology.
Results:
Surgical resectability was achieved in 80.0% of cases, with R0 resection margins in 56.8% of operated patients. Seven radiomic features met reproducibility criteria: VolumeNum, Sphericity, Entropy, Contrast, Energy, SmallAreaEmphasis, and RunLengthNonUniformity. Energy was excluded due to high correlation with Contrast, resulting in six features. The delta-features model demonstrated superior performance (AUC=0.85, 95% CI: 0.78-0.91) compared to pre-treatment (AUC=0.76, p = 0.008) and post-treatment models (AUC=0.82, p = 0.042) for predicting R0 margins. Δ-VolumeNum emerged as the strongest independent predictor (OR=3.14, 95% CI: 1.89-5.21, p < 0.001), followed by Δ-Entropy (OR=2.45, 95% CI: 1.56-3.85, p < 0.001) and Δ-Contrast (OR=1.92, 95% CI: 1.24-2.97, p = 0.003).
Conclusion:
Radiomic delta-features analysis provides a robust and non-invasive tool for predicting R0 resection following NALIRIFOX therapy, with dynamic measurements significantly outperforming static assessments in resectable PDAC.
