Optimized Cnn–Aco–Lstm Hybrid Networks For Early And Accurate Lung Cancer Classification

Abstract

Introduction: Lung cancer remains one of the leading causes of cancer-related mortality worldwide, primarily due to delayed diagnosis and limitations in current diagnostic approaches. Deep learning techniques such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) models have shown remarkable performance in medical image classification tasks. However, optimizing these architectures for complex and high-dimensional medical data remains a challenge.

Aim: This study aims to develop an optimized hybrid deep learning model that combines CNN, Ant Colony Optimization (ACO), and LSTM networks to enhance early and accurate lung cancer classification from computed tomography (CT) images.

Methods: The proposed CNN–ACO–LSTM framework integrates three stages: (1) CNN extracts spatial and hierarchical image features; (2) ACO performs feature selection and hyperparameter optimization to enhance model generalization; and (3) LSTM captures temporal and sequential dependencies for improved classification accuracy. The model was trained and validated on benchmark lung cancer image datasets, with performance evaluated using metrics such as accuracy, precision, recall, F1-score, and area under the ROC curve (AUC). Comparative experiments were conducted against CNN, CNN–LSTM, and CNN–SVM models.

Results and Conclusion: The proposed CNN–ACO–LSTM hybrid model demonstrated superior performance with a classification accuracy of 97.8%, outperforming conventional models by a significant margin. The ROC curve analysis revealed improved sensitivity and specificity, indicating robust diagnostic capability. These results suggest that the optimized CNN–ACO–LSTM framework can effectively support early lung cancer detection and assist clinicians in decision-making, ultimately improving patient outcomes. Future work will explore real-time implementation and extension to other medical imaging modalities.