Expert knowledge for the recognition of leukemic cells

Area of Science:

• * Medical image analysis
• * Hematology
• * Computational pathology

Background:

• * Visual analysis of microscopic images is crucial for diagnosing hematological disorders.
• * Current image recognition techniques are essential for accurate blood tissue identification.
• * Expert knowledge integration can enhance diagnostic accuracy in pathology.

Purpose of the Study:

• * To develop a method for extracting expert knowledge from blood cell images.
• * To classify healthy versus leukemic cells using extracted expert knowledge.
• * To compare the proposed method against deep learning approaches.

Main Methods:

• * Employed Gaussian mixtures, evolutionary computing, and image processing for knowledge extraction.
• * Utilized support vector machines and multilayer perceptrons for cell classification.
• * Implemented convolutional neural networks as a benchmark for performance comparison.

Main Results:

• * Achieved 99.63% accuracy in blood cell recognition using the proposed method.
• * Convolutional neural networks achieved an average accuracy of 97.74%.
• * The developed approach demonstrated minimal computational effort and explainability.

Conclusions:

• * Expert knowledge-driven pattern recognition is highly effective for leukemic cell identification.
• * Traditional methods, when enhanced with expert knowledge, can match or exceed deep learning performance.
• * The proposed methodology offers an efficient and interpretable alternative for hematological image analysis.