Homology-Based Image Processing for Automatic Classification of Histopathological Images of Lung Tissue

Area of Science:

• Medical image analysis
• Computational pathology
• Artificial intelligence in healthcare

Background:

• Accurate classification of histopathological lung tissues is crucial for diagnosis.
• Existing methods may lack efficiency or accuracy in complex cases.
• Development of automated systems can aid pathologists.

Purpose of the Study:

• To develop and validate a computer-aided diagnosis (CAD) system for automatic classification of lung histopathological images.
• To compare the effectiveness of two image feature extraction methods: texture analysis (TA) and homology-based image processing (HI).
• To evaluate the performance of machine learning algorithms in classifying lung tissue images.

Main Methods:

• Utilized two datasets: a private set (94 images, 5 categories) and a public set (15,000 images, 3 categories).
• Employed machine learning algorithms with multiscale analysis for image feature extraction.
• Compared conventional texture analysis (TA) against homology-based image processing (HI) for feature extraction.

Main Results:

• The CAD system achieved accurate classification on both private and public datasets.
• Homology-based image processing (HI) demonstrated superior performance compared to texture analysis (TA).
• The developed CAD system proved effective for lung tissue classification.

Conclusions:

• An accurate computer-aided diagnosis (CAD) system for lung tissues was successfully developed.
• Homology-based image processing (HI) is a more effective feature extraction method for lung tissue CAD systems than texture analysis (TA).
• Automated classification of histopathological lung images is feasible and beneficial.