Feasibility of Using ¹⁸F-FDG PET/CT Radiomics and Machine Learning to Detect Drug-Induced Interstitial Lung Disease

Area of Science:

• Oncology
• Radiology
• Pulmonology

Background:

• Bleomycin is crucial for cancer treatment but can cause interstitial lung disease (DIILD).
• Current methods for detecting DIILD are limited.
• Early detection of DIILD is vital for patient management.

Purpose of the Study:

• To assess the feasibility of ¹⁸F-FDG PET/CT and radiomics for distinguishing DIILD.
• To explore the potential of machine learning in identifying DIILD.
• To evaluate novel biomarkers for bleomycin-induced lung injury.

Main Methods:

• Retrospective analysis of 18 Hodgkin's lymphoma (HL) patients, 10 with DIILD, and 5 DLBCL controls.
• ¹⁸F-FDG PET/CT scans were performed at baseline and during treatment.
• Analysis included standard SUV metrics, 504 radiomics features, and a Random Forest classifier.

Main Results:

• HL patients with DIILD showed significant increases in SUVmean, texture strength, and zone distance entropy.
• These increases were not observed in HL patients without DIILD or DLBCL controls.
• The Random Forest classifier achieved 72.2% accuracy in identifying DIILD.

Conclusions:

• ¹⁸F-FDG PET radiomics features can identify DIILD in HL patients treated with bleomycin.
• Longitudinal increases in specific radiomics features indicate potential as DIILD biomarkers.
• ¹⁸F-FDG PET radiomics, with or without machine learning, may offer new avenues for DIILD detection.