ConvPath: A software tool for lung adenocarcinoma digital pathological image analysis aided by a convolutional neural...

Area of Science:

• Computational pathology
• Cancer biology
• Bioinformatics

Background:

• Cellular spatial distribution is crucial for understanding cancer hallmarks.
• Manual analysis of pathology slides is time-consuming and prone to errors.

Purpose of the Study:

• To develop an automated pipeline (ConvPath) for cell type classification in lung cancer pathology images.
• To extract tumor microenvironment features and develop a prognostic model.

Main Methods:

• Nuclei segmentation and convolutional neural network-based classification of tumor cells, stromal cells, and lymphocytes.
• Extraction of tumor microenvironment features from spatial cell maps.
• Development and validation of an image feature-based prognostic model.

Main Results:

• ConvPath achieved high classification accuracy (92.9% training, 90.1% testing).
• The pipeline generates spatial maps of cell types.
• The prognostic model, based on extracted features, independently predicted patient risk.

Conclusions:

• ConvPath automates the creation of cell-type spatial maps from pathology images.
• This facilitates comprehensive analysis of cell organization in tumor progression and metastasis.
• The approach empowers cancer research and prognostic modeling.