Technical Note: Identification of CT Texture Features Robust to Tumor Size Variations for Normal Lung Texture...

Area of Science:

• Radiology
• Medical Imaging
• Computational Pathology

Background:

• Normal lung CT texture features are explored for predicting radiation-induced lung disease (RILD).
• Clinical utility of these features necessitates robustness against tumor size variations and independence from normal lung volume (peri-tumoral region - PTR).

Purpose of the Study:

• To evaluate the robustness of normal lung CT texture features against variations in gross tumor volume (GTV) size.
• To assess the correlation of these features with the volume of the peri-tumoral region (PTR).

Main Methods:

• CT images from 14 lung cancer patients were analyzed.
• Simulated GTVs of varying sizes were placed in the contralateral lung.
• 27 texture features (intensity histogram, GLCM, GLRM) were extracted from the PTR.
• Bland-Altman analysis was used to determine the normalized range of agreement (nRoA) for feature robustness.

Main Results:

• 16 out of 27 texture features were identified as robust (nRoA < 100%).
• No robust features showed correlation with PTR volume.
• No statistically significant differences in features were observed based on GTV location.

Conclusions:

• 16 robust normal lung CT texture features were identified.
• These features demonstrate potential for clinical application in predicting RILD.
• Further examination of these robust features is warranted for RILD prediction.