Early detection of lung cancer is crucial for improving patient outcomes. This study introduces a novel hybrid model combining Convolutional Neural Networks (CNNs) and DenseNet-121 for enhanced lung cancer classification using clinical data and medical imaging. CNNs are employed to extract key features from histopathological images, followed by preprocessing for improved image quality. DenseNet-121’s densely connected layers refine these features, enhancing learning and feature reuse. An ensemble learning approach with XGBoost integrates clinical data, such as patient demographics and medical history, further boosting performance. To address class imbalance, SMOTE (Synthetic Minority Over-sampling Technique) is applied to the clinical data, and GANs (Generative Adversarial Networks) are used to generate synthetic histopathological images. The model achieves remarkable results, with 98.65% accuracy in binary classification and 98.12% in multi-class classification. Tested on two datasets, it reaches 98.50% accuracy on the image dataset and 97.80% on the clinical dataset. This hybrid approach efficiently distinguishes between benign and malignant cases, identifying multiple cancer subtypes. The key contribution of this work lies in the integration of clinical and imaging data, the effective handling of class imbalance, and the optimization of model performance, offering a scalable solution for early stage lung cancer detection, improving diagnostic precision, and patient care.
Lilhore et al. (Sat,) studied this question.