Adaptive Hybrid Deep Learning For Multi-Cancer Identification: Synergizing Cnns With Aco-Enhanced Lstm Networks

Abstract

Accurate and timely identification of multiple cancer types is critical for enhancing patient prognosis and enabling precision medicine. Traditional diagnostic approaches often rely on manual interpretation of imaging data, which can be time-consuming and prone to human error. To address these challenges, this study proposes an adaptive hybrid deep learning framework that integrates Convolutional Neural Networks (CNNs) with Ant Colony Optimization (ACO)-enhanced Long Short-Term Memory (LSTM) networks for multi-cancer classification. In this framework, CNNs are employed to automatically extract hierarchical spatial features from histopathological and radiological images, capturing intricate patterns associated with different cancer types. Subsequently, ACO is utilized to optimize LSTM hyperparameters and select the most informative features, ensuring efficient sequential learning and reducing overfitting. The optimized LSTM module then performs multi-class classification, effectively capturing temporal dependencies and complex inter-class relationships. Experimental evaluations were conducted on benchmark multi-cancer datasets, including lung, breast, and colorectal cancer images, with training, validation, and testing splits carefully designed to simulate real-world diagnostic scenarios. Results demonstrate that the CNN–ACO–LSTM hybrid framework significantly outperforms conventional CNN, LSTM, and hybrid CNN–LSTM models, achieving superior accuracy, precision, recall, F1-score, and AUC metrics. The integration of spatial feature extraction, adaptive optimization, and temporal modeling makes the proposed approach robust, scalable, and interpretable. This framework not only facilitates early detection and precise identification of multiple cancer types but also provides a powerful computational tool for clinical decision support, potentially reducing diagnostic errors and enabling personalized treatment planning. The proposed method highlights the potential of hybrid deep learning architectures in transforming multi-cancer diagnostics and advancing the field of AI-assisted oncology.