Abstracts

RETURN TO ABSTRACT LISTING


2069. Application of Whole Organ Radiomics for Assessing and Predicting Post-Treatment Changes in Metastatic Liver Disease
Authors * Denotes Presenting Author
  1. Leila Mostafavi *; Harvard Medical School; Massachusetts General Hospital
  2. Fatemeh Homayounieh; Harvard Medical School; Massachusetts General Hospital
  3. Sanjai Saini; Harvard Medical School; Massachusetts General Hospital
  4. Gina Basinsky ; Harvard Medical School; Massachusetts General Hospital
  5. Annick Van Den Abbeele; Harvard Medical School; Massachusetts General Hospital
  6. Gordon Harris; Harvard Medical School; Massachusetts General Hospital
Objective:
To assess if whole liver radiomics can predict and assess post-treatment changes in patients with metastatic liver disease from breast cancer.

Materials and Methods:
Our IRB approved study included 105 women (mean age 65±18 years) with metastatic liver disease from breast cancer. All patients underwent contrast abdomen-pelvis CT in the portal venous phase at two timepoints - baseline (pre-treatment) and follow-up (between 3-12 months following treatment). Based on RECIST 1.1 criteria (Response Evaluation Criteria in Solid Tumors version 1.1), patients were subcategorized into three subgroups: 35 with stable disease (SD), 35 with partial response (PR), and 35 with progressive disease (PD) on the follow up CT. Deidentified baseline and follow-up CT exams were processed on Radiomics prototype (Siemens Healthineers) for semiautomatic segmentation of the entire liver volume. First, second, shape, and higher order radiomics (n= 1630 features) were obtained over the whole liver volume. Multivariable logistic regression analyses with the area under the curve (AUC) and p-value as outputs were performed with in the Radiomics prototype.

Results:
Original shape, GLCM, and higher order (GLSZM) radiomics of whole liver on baseline CT can predict partial response (AUC 0.82, p= 0.03) and progressive disease (AUC 0.77, p= 0.016) from stable disease. Higher order radiomics had an AUC of 0.86 for differentiating treatment response versus disease progression from baseline CT. There was no difference between the radiomics on baseline and follow-up CT in SD (AUC 0.6, p> 0.05). Higher-order radiomics from baseline and follow-up CT also enabled better differentiation of patients with PR (AUC 0.82) compared to those with PD (AUC 0.73).

Conclusion:
Semiautomatic whole liver radiomics can predict treatment response from baseline CT and can differentiate stable disease from partial response and progressive metastatic liver disease from breast cancer. Compared to dominant lesion-based RECIST 1.1 criteria, semiautomatic whole organ-based radiomics can enable holistic assessment of treatment response in patients with hepatic metastases.