Deep learning based automation of mean linear intercept quantification in COPD research

Area of Science:

• Pulmonary Medicine
• Computational Pathology
• Biomedical Imaging

Background:

• Chronic obstructive pulmonary disease (COPD) is a leading cause of death worldwide, requiring new treatments.
• Emphysema, a key component of COPD, involves lung airspace enlargement.
• Manual quantification of emphysema via mean linear intercept (MLI) is subjective and labor-intensive.

Purpose of the Study:

• To develop and validate a deep learning-based automated approach for determining MLI in histological lung sections.
• To assess the performance of the AIxCell software for automated MLI measurement.
• To compare automated MLI values with manual assessments and evaluate its utility in distinguishing between smoke-exposed and control mouse lungs.

Main Methods:

• A deep learning pipeline using the AIxCell AutoML software was developed for semantic segmentation of histological lung sections.
• The pipeline was trained and evaluated on stained histological images from C57BL/6 mice exposed to cigarette smoke and a control group.
• Image processing involved quantifying the mean linear intercept (MLI) to assess airspace enlargement characteristic of emphysema.

Main Results:

• The AIxCell segmentation algorithm achieved high performance with Intersection over Union (IoU) scores consistently above 90%.
• The automated method yielded consistently higher MLI values than manual measurements, indicating reliable quantification.
• The automated MLI determination demonstrated statistical significance in differentiating between smoke-exposed and control mouse lungs.

Conclusions:

• The AIxCell-based automated approach provides a reliable, objective, and efficient method for quantifying MLI in emphysema research.
• This automated method preserves comparability with existing manual assessments while overcoming their limitations.
• The tool is effective for distinguishing lung pathology in COPD models, supporting the development of novel therapies.