A classification system based on a new wrapper feature selection algorithm for the diagnosis of primary and secondary...

Area of Science:

• Hematology
• Medical Informatics
• Computational Biology

Background:

• Primary and Secondary Polycythemia are bone marrow diseases impacting blood composition and donor eligibility.
• Early and accurate diagnosis of these potentially fatal conditions is crucial for patient management.

Purpose of the Study:

• To propose a novel classification system for diagnosing Primary Polycythemia (PP) and Secondary Polycythemia (SP).
• To introduce and evaluate a new wrapper feature selection algorithm, LM-FM, for optimizing classifier performance.

Main Methods:

• A two-level binary classification system: Healthy/non-Healthy and PP/SP.
• Implementation of a novel LM-FM feature selection algorithm combining Local Maximization and Floating Maximization stages.
• Utilizing the Support Vector Machine (SVM) classifier for data discrimination.

Main Results:

• The proposed LM-FM algorithm with SVM achieved high accuracy: up to 98.9% at the first level and 96.6% at the second level.
• Demonstrated excellent robustness of the classification system across different feature subset sizes.
• Outperformed established methods like Sequential Floating Forward Selection (SFFS) and Maximum Output Information (MOI).

Conclusions:

• The developed classification system, featuring the LM-FM algorithm and SVM, offers a highly accurate and robust method for diagnosing polycythemia.
• This advancement supports earlier disease detection, potentially improving patient prognosis and blood donation screening.