Classification of HEp-2 Staining Pattern Images Using Adapted Multilayer Perceptron Neural Network-Based Intra-Class...

Area of Science:

• Medical Image Analysis
• Computational Pathology
• Machine Learning in Histopathology

Background:

• Automating the classification of HEp-2 stained cells in histopathological images is of significant clinical interest.
• Existing methods face challenges including cell density variations, feature overfitting, and large datasets.
• Accurate classification is crucial for disease diagnosis and research.

Purpose of the Study:

• To develop an automated method for classifying HEp-2 stained cells into six classes.
• To address challenges such as overfitting, data volume, and staining variations in histopathological images.
• To improve the accuracy and efficiency of HEp-2 cell classification.

Main Methods:

• An adapted multi-class multilayer perceptron with an added hidden layer was utilized.
• Higher-order spectra features (mean, scale, kurtosis, skewness) of cell shape information were calculated.
• A Softmax activation function was employed for joint training and probability calculation.

Main Results:

• The proposed method achieved 90.3% accuracy with data augmentation, outperforming 87.5% without.
• The technique was trained and tested on ICPR-2014 and ICPR-2016 competition datasets (Task-1).
• The developed framework demonstrated superior performance compared to existing methods and competition benchmarks.

Conclusions:

• The adapted multilayer perceptron effectively classifies HEp-2 stained cells, overcoming common challenges.
• The proposed method shows significant potential for improving automated histopathological image analysis.
• The results confirm the method's effectiveness and superiority over current approaches.