Using MRI Radiomics to Predict the Efficacy of Immunotherapy for Brain Metastasis in Patients with Small Cell Lung Cancer

Abstract

Immune-checkpoint inhibitors (ICIs) combined with chemotherapy has been widely used in the first-line treatment of small cell lung cancer (SCLC) patients. However, the efficacy of ICIs for patients with brain metastases (BMs) of SCLC is limited. There are no effective factors to predict the efficacy. We developed and validate a model to predict intracranial efficacy for ICIs in patients with BMs from SCLC by using MRI radiomics. In this study, we collected 101 SCLC patients with BMs treated with ICIs. They clinical characteristics and pre-treatment Magnetic Resonance Imaging (MRI) were collected. Seventy cases collected from our hospital as training cohort and 31 cases collected from another hospital as an independent validation cohort. Brain metastatic lesions were contoured on ITK-SNAP software and 3748 radiomic features capturing both intra- and peritumoral texture patterns were extracted. The primary endpoint of this study was intracranial overall response rate (ORR). Intraclass correlation coefficient (ICC) and random forest (RF) model were used to select radiomic features. The top 10 selected radiomic features were adopted to build prognostic models by using logistic regression. Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) was chosen as the metric to assess model performances. We also performed a nomogram based on a multivariate logistic regression model includes radiomic and clinical features to predict ICIs intracranial efficacy. The clinical characteristics including number of treatment lines and concurrent brain radiotherapy, ten intratumoral signatures and one peritumoral signatures were found significantly associated with ICIs intracranial efficacy. Predictors contained in the individualized prediction nomogram included the radiomics signature, and clinical features. The model showed favorable discrimination with a C-index of 0.861[95% CI: 0.779-0.943] and AUC of 0.861[95% CI: 0.778- 0.944] in training cohort and with a C-index of 0.812[95% CI: 0.647-0.978] and AUC of 0.812[95% CI:0.646-0.979] in the test cohort. Radiomic features from pre-treatment MRI images were predictive for intracranial efficacy in SCLC patients with BMs. Pre-treatment radiomics may allow early prediction of benefit and expedite more aggressive treatment for high-risk patients and it has to be further explored as predictor of outcome in a larger series of patients.