Abstract Background: With the increasing use of immunotherapy in advanced non-small cell lung cancer (NSCLC), managing adverse events, particularly checkpoint inhibitor-associated pneumonitis (CIP), poses a significant challenge. CIP often necessitates discontinuation of immunotherapy, even in cases of tumor response. Currently there are no effective models in predicting CIP. This study leverages artificial intelligence (AI) algorithms and the use of harmonization models to analyze radiomic features from chest CT scans, aiming to predict the occurrence of CIP in patients with NSCLC. Methods: This retrospective study examined data from 152 stage III-IV NSCLC patients undergoing immunotherapy. Tumor segmentation was performed using LIFEx software (IMIV/CEA, Orsay, France). 3D-radiomic features were extracted from both the tumor and the surrounding 1 cm thick peritumoral regions. The Random Forest (RF) algorithm was employed to develop a classification model to differentiate between CIP and non-CIP. A harmonization model was deployed to account for the scanner-related differences by using spleen and normal lung signals. The dataset was divided into a training set (75%) and a test set (25%). Bootstrapping with 1,000 iterations was performed to estimate the model's performance by calculating the median and 95% confidence interval (CI) estimate. The accuracy of the model's predictions was evaluated by creating a confusion matrix. The model's performance was assessed by calculating the sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), and area under the ROC curve (AUC) for CIP prediction. Results: Among 152 patients, 87 (57.23%) are female and 65 (42.76%) are male. The median age was 59 years. Histology types included were Adenocarcinoma 111 (73%), Squamous cell carcinoma 27 (17.8%), and others 14 (9.2%). CIP was seen in 30 (19.73%) patients. Among the 30 patients with CIP, 17 (56.7%) patients had grade 1 pneumonitis, 11 (36.7%) patients had grade 2 pneumonitis, 2 (6.6%) patients had grade 3 pneumonitis, and none had grade 4 or grade 5 pneumonitis. The model achieved a sensitivity of 0.57, a specificity of 0.67, a PPV of 0.48, and NPV of 0.38 for predicting CIP. The AUC of 0.59 indicates that the model may predict checkpoint inhibitor-associated pneumonitis with an accuracy of 59%. Conclusions: The study provides insights into the potential of radiomic analysis coupled with AI algorithms and the use of harmonization model in predicting CIP among NSCLC patients treated with immunotherapy. Larger studies are needed to validate our findings and the utility of harmonization models in predicting CIP. Citation Format: Monica Yadav, Jeeyeon Lee, Peter Haseok Kim, Seyoung Lee, Taegyu Um, Salie Lee, Maria Jose Chuchuca, Trie Arni Djunadi, Liam Il-Young Chung, Jisang Yu, Darren Rodrigues, Nicolo Gennaro, Leeseul Kim, Myungwoo Nam, Youjin Oh, Sungmi Yoon, Zunairah Shah, Yuchan Kim, Ilene Hong, Jessica Jang, Grace Kang, Amy Cho, Soowon Lee, Timothy Hong, Cecilia Nam, Yury S Velichko, Young Kwang Chae. Harmonization radiomics model to predict immune checkpoint inhibitor-related pneumonitis (CIP) in non small cell lung cancer (NSCLC) in patients treated with immunotherapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7529.
Read full abstract