Virtual Library
Start Your Search
R.J. Gillies
Author of
-
+
P1.01 - Poster Session with Presenters Present (ID 453)
- Event: WCLC 2016
- Type: Poster Presenters Present
- Track: Epidemiology/Tobacco Control and Cessation/Prevention
- Presentations: 1
- Moderators:
- Coordinates: 12/05/2016, 14:30 - 15:45, Hall B (Poster Area)
-
+
P1.01-041 - Quantitative Imaging Features Predict Response of Immunotherapy in Non-Small Cell Lung Cancer Patients (ID 5729)
14:30 - 14:30 | Author(s): R.J. Gillies
- Abstract
Background:
Although immunotherapy has revolutionized the field of cancer treatment, response rates are only ~20% in non-small cell lung cancer (NSCLC) patients and cost of this therapy is high. Predictive biomarkers are needed to identify patients likely to benefit. Converting digital medical images into high-dimensional data (‘Radiomics’) contains information that reflects underlying pathophysiology and that can be revealed via quantitative analyses. We extracted radiomic imaging features from baseline CT scans (prior to initiation of immunotherapy) and identified features that predict response to immunotherapy in NSCLC patients. This work is an initial test of the hypothesis that radiomic data may predict who will respond favorably and who will not.
Methods:
We curated a subset of data and images from 13 different institutional immunotherapy clinical trials. Patients were stage III/IV NSCLC and received PD-1, PD-L1, or doublet checkpoint inhibitors. All target nodules were identified on the CT prior scan prior to initiation of immunotherapy. RECIST guidelines 1.1 were used to measure patient response from baseline to last follow-up scan. Based on last follow-up, 43 patients had progressive disease (PD) and 28 patients with partial response (PR) or complete response (CR). Since we focused on extreme responses, stable disease (SD) patients were not included in the current analyses. We extracted 219 radiomic features including size, shape, location, and texture information from a total of 210 target nodules (lung, lymph nodes, or other). Backward-elimination analyses were utilized to generate parsimonious radiomic models associated with objective responses (PD vs. PR/CR) and post estimation computed performance statistics.
Results:
There were no significant differences for the patient characteristics between patients with PD vs. CR/PR. Analysis of the radiomic features for all target nodules to differentiate PD patients vs. PR/CR patients resulted in a final model containing 2 features that provided an AUROC of 0.64 (95% CI 0.56–0.72). When we analyzed features for only lung target nodules, we identified a final model with 4 features that produced an AUROC of 0.79 (95% CI 0.68–0.89). When we analyzed the imaging features for lymph node target nodules, we found that a final model with 1 feature yielded an AUROC of 0.67 (95% CI 0.51–0.82).
Conclusion:
Radiomic features of lung target nodules have better performance statistics for predicting response to immune therapies compared to target nodules from other organ sites. With this model, cutoffs can be chosen to reduce non-responders with high confidence. Change feature analyses following therapy are underway.
-
+
P1.03 - Poster Session with Presenters Present (ID 455)
- Event: WCLC 2016
- Type: Poster Presenters Present
- Track: Radiology/Staging/Screening
- Presentations: 1
- Moderators:
- Coordinates: 12/05/2016, 14:30 - 15:45, Hall B (Poster Area)
-
+
P1.03-063 - Quantitative Imaging Features Predict Incidence Lung Cancer in Low-Dose Computed Tomography (LDCT) Screening (ID 5669)
14:30 - 14:30 | Author(s): R.J. Gillies
- Abstract
Background:
Although the NLST demonstrated a benefit of LDCT screening for reducing lung cancer and all-cause mortality, LDCT screening identifies large numbers of indeterminate pulmonary nodules and there are limited clinical decision tools that predict probability of cancer development. Using data and images from the NLST, we extracted quantitative imaging features from nodules of baseline positive screens (T0) and delta features from T0 to first follow-up (T1) and performed analyses to identify imaging features that predict incidence lung cancer. Analyses were stratified by nodule size since guidelines in the U.S. have increased the size threshold for positivity to 6 mm.
Methods:
We extracted 438 features from T0 nodules, and delta features from T0 to T1, including size, shape, location, and texture information. Nodules were identified for 170 cases that were diagnosed with incidence lung cancer at the first (T1) or second (T2) follow-up screen and for 328 controls that had three consecutive positive screens (T0 to T2) not diagnosed as lung cancer. The cases and controls were split into a training cohort and a testing cohort and classifier models (Decision tree-J48, Rule Based Classier-JRIP, Naive Bayes, Support Vector Machine, Random Forests) that were stratified by nodule size (< 6 mm, 6 to 16 mm, ≥ 16 mm) were used to identify the most predictive feature sets.
Results:
The training cohort consisted of 83 cases and 172 controls and a testing cohort of 77 cases and 135 controls. Within and across each cohort, there were no significant differences in demographic and clinical covariates. Training and testing was first performed using three difference nodule size groups (< 6 mm, 6 to 16 mm, and ≥ 16 mm) which revealed for < 6 mm a final model of 5 features (AUROC=0.75), 6 to 16 mm a final model with 10 features (AUROC=0.73), and ≥ 16 mm a final model with 10 features (AUROC=0.83). Finally, we combined the two larger groups and found for ≥ 6 mm a final model with 10 features with an AUROC of 0.84 (95% CI 0.8-0.87), Sensitivity of 60% (95% CI 0.50-0.71), and Specificity of 95% (95% CI 0.92-0.99).
Conclusion:
In this analysis we revealed a set of highly informative imaging features that predicts subsequent development of incidence lung cancer among individuals presenting with a ≥ 6 mm nodule. These imaging features could be scored in the clinical setting to improve nodule management of current size-based screening guidelines.
