Virtual Library
Start Your Search
jose Arimateia Batista Araujo Filho
Author of
-
+
P47 - Small Cell Lung Cancer/NET - Biology / Translational (ID 234)
- Event: WCLC 2020
- Type: Posters
- Track: Small Cell Lung Cancer/NET
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P47.06 - Delta-Radiomics Features for Assessment of Individualized Therapeutic Response in Small Cell Lung Cancer – A Pilot Study (ID 3340)
00:00 - 00:00 | Presenting Author(s): jose Arimateia Batista Araujo Filho
- Abstract
Introduction
The difficulty of assessment of treatment response by current RECIST (Response Evaluation Criteria in Solid Tumors) criteria, limited by interobserver agreement and tumoral heterogeneity, may hinder personalized-medicine strategies in small cell lung cancer (SCLC). Our purpose is to evaluate computed tomography (CT) radiomic features before and after platinum-based chemotherapy in SCLC patients and to investigate how therapy-induced changes in these features, called delta-radiomics (DR) features, are related to treatment response as assessed by RECIST v 1.1.
Methods
18 patients with limited (2/18) or extensive (16/18) stage SCLC were retrospectively enrolled and CT scans immediately before and after platinum-based chemotherapy were analyzed. Patients were dichotomized into a group of responders [R] (partial response) and nonresponders [NR] groups (meaning stable or progressive disease) according RECIST v 1.1 criteria. Six patients were excluded for lack of standardization (acquisition protocol, intravenous contrast or image quality) between scans. Tumors were manually segmented (ITK SNAP v 3.8) and 101 radiomics features were extracted from both pre- and post-treatment images. A delta-radiomics signature was constructed using LASSO regression, a support vector machine and 5-fold cross-validation algorithms. ROC analysis to evaluate diagnostic accuracy for R vs NR differentiation was performed with a DeLong confidence interval for Area Under Curve (AUC).
Results
The delta-radiomics signature showed higher AUC than single-CT based radiomics signatures. Significant differences were noted for 12 radiomics features, between R and NR groups (P < 0.05). After LASSO regression 3 parameters were advanced to modelling. The accuracy of the final delta-radiomics model was 92%, sensitivity 86%, specificity 100.0%, positive predictive value 100%, negative predictive value 83.3% and area under ROC curve of 0.89.
Conclusion
Preliminary results showed a potential role of a delta-radiomics CT model in the individualized assessment of platinum-based chemotherapy response in SCLC. Larger sample studies are required to validate the reliability and reproducibility of our proposed radiomics model.