Leukemia detection and classification using computer-aided diagnosis system with falcon optimization algorithm and...

Area of Science:

• Medical Diagnostics
• Computational Biology
• Oncology

Background:

• Leukemia, a blood cancer, involves abnormal white blood cell (WBC) growth in bone marrow.
• Early and accurate leukemia detection is challenging due to limitations in current diagnostic tools like flow cytometry.
• Existing computer-aided diagnostic (CAD) and machine learning (ML) methods aim to improve leukemia analysis.

Purpose of the Study:

• To propose a novel deep learning technique for accurate leukemia detection and classification.
• To address the limitations of time-consuming and less accurate traditional diagnostic methods.
• To enhance the early diagnosis of leukemia through advanced computational approaches.

Main Methods:

• A deep convolutional neural network integrated with a Falcon optimization algorithm (FOADCNN-LDC) was developed.
• Median filtering (MF) was used for initial noise reduction in medical images.
• ShuffleNetv2 was employed for efficient feature extraction, followed by a convolutional denoising autoencoder (CDAE) for classification.
• The Falcon Optimization Algorithm (FOA) optimized the hyperparameters of the CDAE model.

Main Results:

• The FOADCNN-LDC technique demonstrated high performance in leukemia detection and classification.
• The proposed method achieved a superior accuracy of 99.62% on a benchmark medical dataset.
• Comparative analysis indicated that FOADCNN-LDC outperforms existing techniques in accuracy.

Conclusions:

• The FOADCNN-LDC technique offers a highly accurate and efficient approach for leukemia diagnosis.
• This deep learning model shows significant potential for improving early cancer detection in clinical settings.
• The study highlights the effectiveness of integrating optimization algorithms with deep learning for medical image analysis.