Development of a VGG19-Batch Normalization-based Framework for Accurate Detection and Classification of Sickle Cell Anaemia

Abstract

Introduction:In Sickle Cell Anemia (SCA), a genetic blood disorder, RBC distort into a sickle shape, impairing oxygen transport and leading to serious medical complications. Effective therapy and management for patients with SCA rely on its rapid and reliable detection.
Objectives: Existing diagnostic methods are often labour-intensive, manual, and subject to variation due to personal interpretations by healthcare professionals. Automated diagnostic systems face challenges such as inefficient feature extraction and classification, inconsistent blood smear quality, and a lack of available datasets.
Methods: This paper proposes a deep learning framework based on VGG19 with Batch Normalization (VGG19-BN) to address these challenges in the identification and classification of SCA. The methodology includes pre-processing the images to enhance quality and standardize input data. BN layers are integrated into the VGG19 framework to stabilize training, reduce overfitting, and accelerate convergence. The convolutional layer’s extract features to classify RBC into normal and sickle categories. The framework was trained and validated using high-quality blood smear images. The primary objective of this work is to develop a reliable and efficient diagnostic tool capable of achieving high accuracy while remaining user-friendly and interpretable in clinical settings.
Results: The results showed that the VGG19-BN model outperformed baseline deep learning systems and traditional techniques, achieving 96.7% accuracy in classification, 95.3% sensitivity, and 97.8% precision.
Conclusions:By improving both accuracy and efficiency, incorporating this method into clinical workflows could revolutionize SCA diagnosis and improve patient outcomes.