2022 World Conference on Lung Cancer (Posters)-P1.15-01. Differential Diagnosis of Pneumonitis in Metastatic NSCLC (Non-Small Cell Lung Cancer) Patients Receiving Immunotherapy With Radiomics

P1.15-01. Differential Diagnosis of Pneumonitis in Metastatic NSCLC (Non-Small Cell Lung Cancer) Patients Receiving Immunotherapy With Radiomics

Back to course

Pdf Summary

The objective of this study was to develop an AI-based model to improve the diagnosis of immune-induced pneumonitis (IIP) in patients with metastatic non-small cell lung cancer (NSCLC) receiving immunotherapy. The researchers reviewed 556 NSCLC patients from a clinical trial who were treated with PD-(L)1 inhibitors in multiple centers. They identified cause-specific pneumonitis events and examined seven regions of interest (ROIs) in the chest CT images to extract radiomic features.<br /><br />The best radiomic model achieved a ROC AUC of 0.76 and an NPV of 84% using a 75mm spheroidal ROI. Specific radiomic features, including wavelet first-order minimum with high-pass filtering and original glszm zone entropy, were included in the radiomic signature. The radiomic and combined models provided correct diagnoses in 67-75% of patients with non-conclusive radiological judgments. They also correctly classified 82% of patients who were considered non-conclusive cases by medical oncologists.<br /><br />The study concludes that radiomic biomarkers applied to CT imaging can support clinicians in making the differential diagnosis of pneumonitis in NSCLC patients receiving immunotherapy, particularly when the radiologic assessment is inconclusive. The AI-companion model developed in this study has the potential to improve the accuracy of diagnosing immune-induced pneumonitis.<br /><br />The study acknowledges the investigators from the centers and the company Bristol Myers Squibb for their support. The clinical rationale and study design of the decision support system are presented in Figure 1. The results of the study, including ROC curves and calibration plots, are shown in Figure 3. Figure 4 compares the model predictions with radiological assessments.

Asset Subtitle

Alberto Traverso, Netherlands

Meta Tag

Speaker Alberto Traverso, Netherlands

Topic Metastatic Non-small Cell Lung Cancer – Immunotherapy

Keywords

AI-based model

immune-induced pneumonitis

metastatic non-small cell lung cancer

immunotherapy

PD-(L)1 inhibitors

chest CT images

radiomic features

ROC AUC

NPV

radiomic signature