Lung cancer remains the leading cause of cancer-related mortality worldwide. Early detection of pulmonary nodules is crucial for timely diagnosis and effective treatment. Conventional computer-aided detection systems have shown limitations, including high false-positive rates and low sensitivity. Recent advances in deep learning, particularly convolutional neural networks (CNNs), have shown great potential in improving the accuracy and reliability of nodule detection and classification. This study aimed to develop and evaluate an automatic method for lung nodule detection and classification using a CNN-based architecture applied to computed tomography images from the publicly available LIDC-IDRI database.
Methods
This retrospective study was conducted on 82 patients (10,496 computed tomography slices) selected from the LIDC-IDRI database. The proposed method consists of five main steps: image preprocessing, lung parenchyma segmentation using Otsu's thresholding and morphological operations, detection of nodule candidates, feature extraction, and classification using a CNN model. The CNN architecture includes two convolutional layers (20 and 30 filters, 3×3 kernel), ReLU activation, max-pooling layers, and a Softmax output layer. The network was trained with a mini-batch size of 32 for 50 epochs using the Stochastic Gradient Descent with Momentum optimizer (learning rate = 0.001, momentum = 0.9). Model performance was evaluated in terms of sensitivity, specificity, precision, and accuracy.
Results
The proposed CNN model successfully detected pulmonary nodules and achieved accurate classification between benign and malignant nodules. On the LIDC-IDRI dataset, the model achieved a sensitivity of 98.7%, specificity of 97.5%, precision of 97.9%, and accuracy of 98.4%. Comparative analysis with recent studies, including hybrid CNN-long short-term memory and ResNet-based models, demonstrated that the proposed method provides competitive performance while maintaining lower computational complexity. The classification of nodule subtypes (solid, partially frosted, totally frosted) showed satisfactory discrimination results.
Conclusions
The proposed CNN-based system demonstrates the feasibility and robustness of deep learning for automatic lung nodule detection and classification. Despite strong results, the study acknowledges limitations such as single-database validation and a relatively small training size. Future work will focus on validating the model across other datasets (e.g., ELCAP, NELSON) and optimizing multi-class classification performance to enhance generalizability and clinical applicability.
Full text:
https://www.xiahepublishing.com/2472-0712/ERHM-2025-00032
The study was recently published in the Exploratory Research and Hypothesis in Medicine .
Exploratory Research and Hypothesis in Medicine (ERHM) publishes original exploratory research articles and state-of-the-art reviews that focus on novel findings and the most recent scientific advances that support new hypotheses in medicine. The journal accepts a wide range of topics, including innovative diagnostic and therapeutic modalities as well as insightful theories related to the practice of medicine. The exploratory research published in ERHM does not necessarily need to be comprehensive and conclusive, but the study design must be solid, the methodologies must be reliable, the results must be true, and the hypothesis must be rational and justifiable with evidence.
