Advancing Blood Cell Image Classification: Hybrid CNN-LSTM vs. Traditional CNN Approaches

The classification of microscopic blood cell images is a vital component in medical diagnostics, as accurately identifying different cell types is crucial for diagnosing a range of hematological conditions. Traditional Convolutional Neural Networks (CNNs) are used for this goal due to their effectiveness in extracting spatial features. However, these models often encounter challenges in capturing the sequential patterns present in imaging data, leading to limitations in classification accuracy. In this study, we propose and evaluate a hybrid CNN-LSTM model that leverages the strengths of CNNs for feature extraction combined with Long Short-Term Memory (LSTM) networks for managing sequential dependencies. The dataset used in this study includes 17,092 high-quality microscopic images of peripheral blood cells, classified into 8 categories: neutrophils, eosinophils, basophils, lymphocytes, monocytes, immature granulocytes, erythroblasts, and platelets, annotated by pathologists to ensure the dataset's reliability for model training and evaluation. Our experimental findings reveal a notable enhancement in classification accuracy using the hybrid CNN-LSTM model, which achieved an impressive accuracy rate of 98%. This is a substantial improvement compared to the 63% accuracy reached by the traditional CNN approach. The hybrid model's superior performance underscores its capability to effectively capture both spatial and sequential features, which are critical for the accurate classification of blood cell images. This study not only highlights the potential of hybrid architectures in advancing medical image classification but also establishes a new benchmark for future research in the field. The results suggest that integrating sequential learning mechanisms with conventional CNN frameworks could significantly improve classification accuracy, making it a promising approach for developing robust automatic recognition systems for blood cell analysis.