Augmented cell-graphs for automated cancer diagnosis

Area of Science:

• Computational pathology
• Medical image analysis
• Graph theory in medicine

Background:

• Malignant glioma diagnosis relies on histopathological analysis of brain biopsies.
• Accurate and efficient diagnostic tools are crucial for timely treatment of brain tumors.
• Computational methods offer potential for objective and reproducible diagnostic assessments.

Purpose of the Study:

• To introduce a novel computational method, augmented cell-graphs (ACG), for the mathematical diagnosis of malignant glioma.
• To evaluate the diagnostic performance of the ACG approach using a dataset of human brain biopsy samples.

Main Methods:

• Constructing augmented cell-graphs (ACG) from low-magnification tissue images, where nodes represent cell clusters and edges represent relationships.
• Assigning weights to nodes and edges to capture tissue topology.
• Applying the ACG method to a dataset of 646 human brain biopsy samples from 60 patients.

Main Results:

• The ACG approach achieved a sensitivity of 97.53% in glioma diagnosis.
• Specificities of 93.33% for inflamed tissue and 98.15% for healthy tissue were obtained.
• The method demonstrated high accuracy at the tissue level for brain cancer diagnosis.

Conclusions:

• Augmented cell-graphs (ACG) provide a robust computational framework for the mathematical diagnosis of malignant glioma.
• The ACG method shows significant potential for improving the accuracy and efficiency of brain cancer diagnostics.
• This novel approach contributes to the advancement of computational pathology in neuro-oncology.