Optimized Classification of Blood Cancer Using Recurrent Neural Networks

Abstract

Blood cancer remains one of the most fatal diseases, underscoring the need for early detection to improve survival rates. Traditional convolutional neural network (CNN)-based models, while effective in feature extraction, often suffer from overfitting when applied to small or imbalanced datasets, thereby reducing their predictive accuracy. To overcome this limitation, this study proposes a hybrid deep learning model that integrates bidirectional long short-term memory (BiLSTM) with MobileNet. MobileNet efficiently extracts spatial features from medical images, while BiLSTM captures long-term dependencies within the data, enhancing model robustness and generalization. This combination improves the system’s ability to recognize complex patterns, resulting in more accurate blood cancer classification. By leveraging the strengths of both architectures, the proposed model mitigates overfitting, enhances predictive performance, and ensures reliable diagnosis. Experimental results demonstrate that the hybrid approach significantly outperforms conventional CNN-based models in terms of accuracy and generalization. This study highlights the potential of deep learning in medical diagnostics and offers an effective solution for improving blood cancer prediction, ultimately contributing to better clinical decision-making and improved patient outcomes.