Submission ID 92704
| Poster Code | HR-P-19 |
|---|---|
| Title of Abstract | A Multi-Modality Radiomics-Based Model for Recurrence Risk Stratification in Non-Small Cell Lung Cancer |
| Abstract Submission | Background/Purpose: To develop a model integrating multi-modal quantitative imaging features (radiomics) from tumour and non-tumour volumes of interest, qualitative features, and clinical data to improve risk-stratification of patients with non-small cell lung cancer (NSCLC). Methods: A dataset of 135 patients with early-stage NSCLC who underwent surgical resection was analyzed. The tumour and peri-tumoural volumes on both pre-treatment computed tomography (CT) and positron emission tomography (PET) were segmented, while the bone marrow (L3-L5 vertebral bodies) was segmented on PET. Radiomic features were extracted from the segmented volumes. Combined with the clinical and qualitative CT features, feature selection was performed on 5092 features to select the top features to predict time to recurrence in the training cohort (n=101). A Random Survival Forest model was built in the training cohort and evaluated on the testing cohort (n=34). Model performance was assessed using the concordance index and compared to a baseline clinical model with cancer staging. Patients were stratified into high- and low-risks of recurrence using Kaplan-Meier analysis. Results: A total of seven features were selected to predict time to recurrence. These consisted of cancer stage, three texture features (one PET peri-tumoural feature and two CT tumour features) and three intensity features (one CT peri-tumoural, one PET tumour and one bone marrow feature). The radiomics model achieved concordances of 0.78 and 0.76, significantly outperforming the baseline stage-only model of 0.67 (p<0.005) and 0.60 (p=0.008) in the training and testing cohorts, respectively. Patients were significantly stratified into high- and low-risks of recurrence in both the training (p<0.005) and the testing (p=0.03) cohorts. Discussion: Our radiomic model, consisting of stage, tumour, peri-tumoural and bone marrow features from CT and PET significantly stratified patients into low- and high-risk of recurrence. Radiomics has the potential to help clinicians provide high-risk patients with more aggressive or personalized treatment options. |
| Please indicate who nominated you | Thomas Drysdale; Schulich School of Medicine & Dentistry, Western University |
| What Canadian Institutes of Health Research (CIHR) institute is your research most closely aligned? | Cancer Research |
| What Canadian Institutes of Health Research (CIHR) pillar of health research does your research fall under? | Biomedical |
| PDF of abstract | No file |
| Presenter and Author(s) | Jaryd Christie Jaryd Christie Omar Daher Mohamed Abdelrazek Viswam Nair Sarah Mattonen |
