Detection and localization of radiation-induced pneumonitis using T-2 -mapping magnetic resonance imaging

Klaar, Rabea; Begaj, Kaltra; Rabe, Moritz; Corradini, Stefanie; Eze, Chukwuka; Belka, Claus; Sabel, Bastian; Landry, Guillaume; Kurz, Christopher; Dinkel, Julien (2025): Detection and localization of radiation-induced pneumonitis using T-2 -mapping magnetic resonance imaging. Physics and Imaging in Radiation Oncology, 36: 100878. ISSN 24056316

Abstract

Background and purpose:
Radiation-induced pneumonitis (RP), a complication of lung radiotherapy, occurs at the earliest 6–12 weeks post-treatment. To assess RP, repeated computed tomography (CT)-scans post-radiotherapy are standard-of-care, but increase the patients’ dose burden and secondary cancer risk. We propose a pipeline based on magnetic resonance imaging (MRI) T2-mapping acquired 2–3 months post-radiotherapy that provides an automated patient stratification and initial segmentation of the RP lung volume.

Materials and methods:

In total, 24 lung tumor patients received MRI-guided radiotherapy at a 0.35T MR-Linac. MRI T2-maps were retrieved from T2-weighted images acquired at a diagnostic 1.5T MRI-scanner 8–20 weeks post-radiotherapy. Mean baseline-corrected T2-values were calculated in the planning target volume and the lung volume receiving ≥ 20Gy excluding the gross tumor volume. Their stratification potential (endpoint RP grade ≥ 1) was assessed in a univariate receiver operating characteristic curve–area under the curve (ROC–AUC) analysis using bootstrapping. Significant differences were probed (Mann–Whitney U test, α Stats =0.05). Thresholding using the maximal Youden index was utilized for the T2-based RP segmentation. The Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), sensitivity, precision and segmentation AUC (SegAUC) were used for the comparison with the ground-truth CT-based RP segmentation.

Results:
RP grade ≥ 1 was diagnosed in 15/24 patients. The T2-values in both regions achieved significant separation of distributions (median 13.8/2.9ms and 5.0/-2.6ms RP/non-RP) with p-values < 0.05 and AUC ≥ 0.76. Moderate quantitative agreement was found between
T2-based and ground-truth segmentation (DSC=0.32, HD95=20.1mm and SegAUC=0.76).

Conclusion:
MRI T2-values allow an automated RP patient stratification and initial RP lung volume estimation.

• Dokumententyp: Artikel (Klinikum der LMU)
• Organisationseinheit (Fakultäten): 07 Medizin > Klinikum der LMU München > Klinik und Poliklinik für Radiologie; 07 Medizin > Klinikum der LMU München > Klinik und Poliklinik für Strahlentherapie und Radioonkologie
• DFG-Fachsystematik der Wissenschaftsbereiche: Lebenswissenschaften
• DFG: Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) - 491502892