Document Type

Conference Proceeding

Publication Date


Publication Title

Ann Oncol


Background: The addition of checkpoint inhibitors to chemotherapy in SCLC patients provides modest benefit, with a median survival of 12 months. Development of non-invasive imaging predictors to identify patients most likely to benefit from chemo-immunotherapy would enable personalized management of SCLC.

Methods: A cohort of 31 extensive-stage SCLC patients treated with atezolizumab, carboplatin, and etoposide from June 2020 to May 2021 were identified and pre-treatment CT scans were curated. The axial slice at the level of the carina (S1) was identified and center-cropped. 304 3D radiomic features from 5 slices surrounding S1 were extracted for analysis. After feature selection, the most discriminative radiomic feature was used to train and evaluate a random forest machine classifier for mortality prediction using leave-one-out cross-validation (LOOCV). A baseline classifier was trained using clinical variables. LOOCV mortality probabilities were recorded for each patient and used to stratify patient risk. Overall survival (OS) analysis was performed using Cox modeling.

Results: Median follow-up was 343 days. Patient data included median age of 67 (46-85), race (24 white, 7 black), 58% female, and liver metastases at diagnosis in 29%. The Haralick difference variance feature had an AUC of 0.77 (c-index: 0.70) compared to the clinical baseline AUC of 0.56 (c-index: 0.64) for mortality and OS. The radiomic classifier identified low (N=12) and high (N=19) risk cohorts with median OS of 519.5 and 194 days, respectively (p=.01). There was no significant difference in OS for low and high risk cohorts identified by clinical features (p=0.47).

Conclusions: Patient survival following chemo-immunotherapy in SCLC can be predicted using computational analysis of pre-treatment images. Our results encourage study of larger patient cohorts to further understand the relationship between imaging signatures and survival in SCLC, potentially leading to improved personalized disease management.


Legal entity responsible for the study: Stony Brook University. Funding: Has not received any funding.

Disclosure: P. Prasanna: Financial Interests, Personal, Research Grant: IBM. S.M. Gadgeel: Financial Interests, Personal, Advisory Board: AstraZeneca, Amgen, Genentech/Roche, Bristol Myers Squibb, Pfizer, Novartis, Blueprint, Daiichii; Financial Interests, Personal, Other, Data Safety Monitoring Board: AstraZeneca. All other authors have declared no conflicts of interest.



First Page




To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.