Changes of radiomic features over time in longitudinal images, delta radiomics, can potentially be used as a biomarker to predict treatment response. This study aims to develop a delta-radiomic process based on machine learning by (1) acquiring and registering longitudinal images, (2) segmenting and populating regions of interest (ROIs), (3) extracting radiomic features and calculating their changes (delta-radiomic features, DRFs), (4) reducing feature space and determining candidate DRFs showing treatment-induced changes, and (5) creating outcome prediction models using machine learning. This process was demonstrated by retrospectively analyzing daily non-contrast CTs acquired during routine CT-guided-chemoradiation therapy for 90 pancreatic cancer patients. A total of 2520 CT sets (28-daily-fractions-per-patient) along with their pathological response were analyzed. Over 1300 radiomic features were extracted from the segmented ROIs. Highly correlated DRFs were ruled out using Spearman correlations. Correlation between the selected DRFs and pathological response was established using linear-regression-models. T test and linear-mixed-effects-models were used to determine which DRFs changed significantly compared with first fraction. A Bayesian-regularization-neural-network was used to build a response prediction model. The model was trained using 50 patients and leave-one-out-cross-validation. Performance was judged using the area-under-ROC-curve. External independent validation was done using data from the remaining 40 patients. The results show that 13 DRFs passed the tests and demonstrated significant changes following 2–4 weeks of treatment. The best performing combination differentiating good versus bad responders (CV-AUC = 0.94) was obtained using normalized-entropy-to-standard-deviation-difference-(NESTD), kurtosis, and coarseness. With further studies using larger data sets, delta radiomics may develop into a biomarker for early prediction of treatment response.

PurposeIn an effort for early assessment of treatment response, we investigate radiation induced changes in quantitative CT features of tumor during the delivery of chemoradiation therapy (CRT) for pancreatic cancer. MethodsDiagnostic-quality CT data acquired daily during routine CT-guided CRT using a CT-onrails for 20 pancreatic head cancer patients were analyzed. On each daily CT, the pancreatic head, the spinal cord and the aorta were delineated and the histograms of CT number (CTN) in these contours were extracted. Eight histogram-based radiomic metrics including the mean CTN (MCTN), peak position, volume, standard deviation (SD), skewness, kurtosis, energy and entropy were calculated for each fraction. Paired t-test was used to check the significance of the change of specific metric at specific time. GEE model was used to test the association between changes of metrics over time for different pathology responses. ResultsIn general, CTN histogram in the pancreatic head (but not in spinal cord) changed during the CRT delivery. Changes from the 1 st to the 26 th fraction in MCTN ranged from -15.8 to 3.9HU with an average of -4.7 HU (p<0.001). Meanwhile the volume decreased, the skewness increased (less skewed), and the kurtosis decreased (less peaked). Data Availability Statement: All relevant data are within the paper. Funding:The authors received no specific funding for this work. Competing interests:The authors have declared that no competing interests exist.pathological response is associated with the changes of MCTN, SD, and skewness. In cases of good response, patients tend to have large reductions in MCTN and skewness, and large increases in SD and kurtosis. ConclusionsSignificant changes in CT radiomic features, such as the MCTN, skewness, and kurtosis in tumor were observed during the course of CRT for pancreas cancer based on quantitative analysis of daily CTs. These changes may be potentially used for early assessment of treatment response and stratification for therapeutic intensification.

No abstract

Purpose We present the advantages of using dual-energy CT (DECT) for radiation therapy (RT) planning based on our clinical experience. Methods DECT data acquired for 20 representative patients of different tumor sites and/or clinical situations with dual-source simultaneous scanning (Drive, Siemens) and single-source sequential scanning (Definition, Siemens) using 80 and 140-kVp X-ray beams were analyzed. The data were used to derive iodine maps, fat maps, and mono-energetic images (MEIs) from 40 to 190 keV to exploit the energy dependence of X-ray attenuation. The advantages of using these DECT-derived images for RT planning were investigated. Results When comparing 40 keV MEIs to conventional 120-kVp CT, soft tissue contrast between the duodenum and pancreatic head was enhanced by a factor of 2.8. For a cholangiocarcinoma patient, contrast between tumor and surrounding tissue was increased by 96 HU and contrast-to-noise ratio was increased by up to 60% for 40 keV MEIs compared to conventional CT. Simultaneous dual-source DECT also preserved spatial resolution in comparison to sequential DECT as evidenced by the identification of vasculature in a pancreas patient. Volume of artifacts for five patients with titanium implants was reduced by over 95% for 190 keV MEIs compared to 120-kVp CT images. A 367-cm 3 region of photon starvation was identified by low CT numbers in the soft tissue of a mantle patient in a conventional CT scan but was eliminated in a 190 keV MEI. Fat maps enhanced image contrast as demonstrated by a meningioma patient. Conclusion The use of DECT for RT simulation offers clinically meaningful advantages through improved simulation workflow and enhanced structure delineation for RT planning.

PURPOSE: To investigate the feasibility of using apparent diffusion coefficient (ADC) to assesspathological treatment response in pancreatic ductal adenocarcinoma (PDAC) following neoadjuvant chemoradiation (nCR). MATERIALS/METHODS: MRI and pathological data collected for 25patients with resectable and borderline resectable PDAC following nCR were retrospectively analyzed. Pre- and post-nCR mean ADC values in the tumors were compared using Wilcoxon matched pairs test. Correlation of pathological treatment response and ADC values was assessed using Pearson’s correlation coefficient test and receiver-operating-curve (ROC) analysis. RESULTS: The average mean and standard deviation (SD) of the ADC values for all the patients analyzed were significantly higher in post-nCR (1.667±0.161×10-3) compared with those prior to nCR (1.395±0.136×10-3 mm2/sec), (P<0.05). The mean ADC values after nCR were significantly correlated with the pathological responses (r=-0.5172); P=0.02. The area under the curve (AUC) of the ADC values for differentiating G1, G2 and G3 pathological responses, using ROC analysis, was found to be 0.6310 and P=0.03. CONCLUSION: Changes of pre- and post-treatment ADC values significantly correlated with pathological treatment response for PDAC patients treated with chemoradiation therapy, indicating that the ADC could be used to assesstreatment response for PDAC.