Virtual Library
Start Your Search
Prantesh Jain
Author of
-
+
MA13 - Tumor Biology: Focus on EGFR Mutation, DNA Repair and Tumor Microenvironment (ID 214)
- Event: WCLC 2020
- Type: Mini Oral
- Track: Tumor Biology and Systems Biology - Basic and Translational Science
- Presentations: 1
- Moderators:
- Coordinates: 1/31/2021, 16:45 - 17:45, Scientific Program Auditorium
-
+
MA13.02 - Novel Non-Invasive Radiomic Signatures Extracted from Radiographic Images can Predict Response to Systemic Treatment in Small Cell Lung Cancer (ID 3151)
16:45 - 17:45 | Presenting Author(s): Prantesh Jain
- Abstract
Introduction
SCLC is an aggressive malignancy characterized by inevitable resistance to chemotherapy. There are no predictive biomarkers that can accurately guide use of systemic therapy in SCLC patients. We hypothesized that quantitative radiomic features (i.e. computer extracted imaging) from pretreatment CT scans can prognosticate survival and predict sensitivity to chemotherapy. We sought to train a prognostic classifier and use it to predict response to chemotherapy.
Methods
180 SCLC patients who recieved platinum-based chemotherapy were selected. 27 patients were excluded, with no measurable disease. Remaining 153 patients were randomly divided into training (n=77) and validation set (n=76). Lung tumors were contoured on 3D-Slicer® software by an expert reader. 1542 radiomic features (textural and shape) were extracted from intra and peritumoral regions. Primary endpoints of this study were overall survival (OS) and objective response to chemotherapy per RECIST. Patients with complete or partial response were defined as responders (R) and those with stable or progression of disease as non-responders (NR). Radiomic risk score (RRS) was generated by using least absolute shrinkage and selection operator, and Cox regression model was used to predict OS. Kaplan Meier and log-rank tests were performed to assess discriminative ability of the features. Features prognostic of OS were used to train a machine learning classifier to predict response to chemotherapy. A linear discriminant analysis (LDA) classifier was trained and used to predict response. Area under receiver operating characteristic curve (AUC) was calculated for response to chemotherapy.
Results
153 SCLC patients were included, median age 66 years, 72.8% men and median OS of 9.37 months. 75% had ES and 35% LS disease. Multivariate Cox regression analysis indicated that RRS was significantly associated with OS in the training set [HR: 1.53; 95% CI, 1.1–2.2; P=0.021; C-index=0.72) and validation set [HR: 1.4; 95% CI, 1.1–1.82; P=0.0127; C-index=0.69). Chemotherapy response was achieved in 71 (66%); labeled responders (R) and the rest 36 (34%) labeled as non-responders (NR). LDA classifier trained with prognostic features was able to predict response with AUC of 0.76 ± 0.03 within the training set and corresponding AUC of 0.72 within the validation set. Multivariate Cox regression analysis with radiomic features and clinical biomarkers identified the RRS and cancer stage (LS or ES) as two risk factors in OS for patients in the training set (RRS: HR, 2.1, 95 % CI: 1.53, 2.85, P = 0.0076; clinical stage: HR, 1.66, 95 % CI: 1.01, 2.7, P = 0.048; and age: HR, 1.04, 95 % CI: 0.99, 1.09, P = 0.071; C-index = 0.75) and corresponding validation set (RRS: HR, 1.9, 95 % CI: 1.23, 2.2, P = 0.0012; clinical stage: HR, 1.61, 95 % CI: 1.2, 2.17, P = 0.041 and age: HR, 1.01, 95 % CI: 0.99, 1.03, P = 0.22; C-index = 0.71).
Conclusion
Texture features extracted from within and around the lung tumor from pretreatment CT images were both prognostic of OS and predictive of response in SCLC patients. Pretreatment radiomic features may permit early assessment of benefit and expedite alternative treatment options. Additional independent validation of these image-based biomarkers is warranted.
-
+
P68 - Tumor Biology and Systems Biology - Basic and Translational Science - Radiomics (ID 207)
- Event: WCLC 2020
- Type: Posters
- Track: Tumor Biology and Systems Biology - Basic and Translational Science
- Presentations: 1
- Moderators:
- Coordinates: 1/28/2021, 00:00 - 00:00, ePoster Hall
-
+
P68.02 - Computer Extracted Morphology Features of Tumor Nuclei Predict Response to Chemotherapy and Prognostic of OS in Small Cell Lung Cancer (ID 3574)
00:00 - 00:00 | Presenting Author(s): Prantesh Jain
- Abstract
Introduction
Small cell lung cancer (SCLC) is an aggressive malignancy that accounts for ~15% of all lung cancers and is characterized by inevitable chemotherapy resistance and rapid progression. To date, there are no consistent predictive biomarkers that can accurately guide the use of systemic therapy in patients with SCLC. We hypothesize that computer extracted morphology features of cancer nuclei from digitized whole slide images (WSI) of pre-treatment H&E biopsy specimens are prognostic of overall survival (OS) and also predict sensitivity to platinum-based chemotherapy.
Methods
106 patients with extensive and limited-stage SCLC who received platinum-doublet chemotherapy were selected for this study. WSIs of SCLC tumor tissues were retrospectively digitized at 40x magnification. Tumor regions were manually annotated by an expert pathologist. An automated machine learning model was employed to automatically identify cancer nuclei. A set of 100 features related to the morphological phenotype and functional characteristics (shape, size, intensity, cellular texture) of the cancer nuclei were extracted from each case. Primary endpoints of the study were overall survival (OS) and the best objective response to chemotherapy (RECIST criteria). The predictive and prognostic capabilities of the features were assessed by a Naive Bayes classifier and cross-validation scheme with a Cox regression model respectively. The median risk score was used as a threshold for labeling the patient as either having a low or high risk of death. Kaplan-Meier survival analysis was used to evaluate the procedure. The statistical significance of the cross-validated Kaplan-Meier curves was evaluated by using a permutation test.
Results
The median age was 66 years, 68% had extensive stage disease and 32% had limited stage, median follow-up was 9 months. Based on the approach, median OS for patients identified by our model as being at high risk of death was 11 months vs. 14 months for low-risk patients. No statistically significant difference between extensive stage and limited stage disease was found (p=0.3916). On univariable survival analysis, high-risk patients had a hazard ratio of 1.66 (95% CI: 1.03-2.68, p=0.0367), with a statistical significance of 4% after 500 permutations. The model predicted responders from non-responders with an accuracy of 0.63 and a precision-recall of 0.83.
Conclusion
A computerized image analysis model based on the morphological features of cancer nuclei on pretreatment H&E biopsy slide images were found to be predictive of response and prognostic for OS in SCLC patients. Future work will entail additional independent multi-site validation of the signature.
