Acute lymphoblastic leukemia diagnosis using machine learning techniques based on selected features

• Published in: Scientific reports Vol. 15; no. 1; pp. 28056 - 12
• Main Author: El Houby, Enas M.F.
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 01.08.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 631/114; 631/67; 639/166; Acute lymphoblastic leukemia; Algorithms; Artificial intelligence; Automation; Bayes Theorem; Blood; Cancer; Child; Computer-aided diagnosis; Datasets; Decision making; Diagnosis; Diagnosis, Computer-Assisted - methods; Humanities and Social Sciences; Humans; Leukemia; Lymphatic leukemia; Machine Learning; Machine learning techniques; multidisciplinary; Precursor Cell Lymphoblastic Leukemia-Lymphoma - diagnosis; Science; Science (multidisciplinary); Support Vector Machine; Acute lymphoblastic leukemia; Machine learning techniques; Computer-aided diagnosis
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-12361-4

Cancer is considered one of the deadliest diseases worldwide. Early detection of cancer can significantly improve patient survival rates. In recent years, computer-aided diagnosis (CAD) systems have been increasingly employed in cancer diagnosis through various medical image modalities. These systems play a critical role in enhancing diagnostic accuracy, reducing physician workload, providing consistent second opinions, and contributing to the efficiency of the medical industry. Acute lymphoblastic leukemia (ALL) is a fast-progressing blood cancer that primarily affects children but can also occur in adults. Early and accurate diagnosis of ALL is crucial for effective treatment and improved outcomes, making it a vital area for CAD system development. In this research, a CAD system for ALL diagnosis has been developed. It contains four phases which are preprocessing, segmentation, feature extraction and selection phase, and classification of suspicious regions as normal or abnormal. The proposed system was applied to microscopic blood images to classify each case as ALL or normal. Three classifiers which are Naïve Bayes (NB), Support Vector Machine (SVM) and K-nearest Neighbor (K-NN) were utilized to classify the images based on selected features. Ant Colony Optimization (ACO) was combined with the classifiers as a feature selection method to identify the optimal subset of features among the extracted features from segmented cell parts that yield the highest classification accuracy. The NB classifier achieved the best performance, with accuracy, sensitivity, and specificity of 96.15%, 97.56, and 94.59%, respectively.