Glioma Grading Using Cell Nuclei Morphologic Features in Digital Pathology Images

Area of Science:

• Computational pathology
• Medical image analysis
• Oncology

Background:

• Accurate grading of brain gliomas is crucial for effective treatment planning.
• Existing methods for glioma grading from histopathology images have limitations.
• Computational analysis of cell nuclei morphology offers a promising avenue for objective assessment.

Purpose of the Study:

• To develop a computationally efficient technique for cell nuclei morphologic feature analysis in brain gliomas.
• To optimize cell nuclei segmentation and extract representative morphologic features.
• To classify brain tumors into glioblastoma multiforme (GBM) and low-grade glioma (LGG) using machine learning.

Main Methods:

• An optimized cell nuclei segmentation method was developed by evaluating existing techniques.
• K-mean clustering was employed to extract representative nuclei morphologic features (area, perimeter, eccentricity, major axis length).
• A multilayer perceptron (MLP) classifier was used to differentiate between GBM and LGG based on extracted features.

Main Results:

• The proposed method demonstrated high accuracy in classifying brain gliomas.
• Quantitative evaluation using precision, recall, and accuracy was performed on TCGA dataset images.
• An average accuracy of approximately 94% was achieved through 10-fold cross-validation, confirming method efficacy.

Conclusions:

• The developed technique offers an efficient and accurate approach for brain glioma characterization.
• K-mean clustering for feature extraction overcomes limitations of previous methods.
• The findings support the use of computational morphologic analysis for objective glioma grading in pathology.