article · Insights into Imaging
Artificial intelligence models can effectively classify pulmonary nodules identified in computed tomography (CT) scans to support lung cancer diagnosis. An evaluation using 1,007 nodules from 551 patients in the LIDC-IDRI dataset compared traditional statistical machine learning against deep learning architectures. For machine learning, texture and local binary pattern features were filtered to four components, achieving a top accuracy of 81.9 percent with support vector machines and an area under the receiver operating characteristic curve of 0.885 using random forest. Deep learning models applying transfer learning demonstrated superior results, with DenseNet-121 attaining 90.39 percent accuracy and 93.65 percent specificity, while a simple convolutional neural network achieved an area under the curve of 96.0 percent. Overall, deep learning reduced training effort and improved prediction performance, though further improvements remain possible through three-dimensional lesion representation and larger training datasets.
Lung cancer detection relies heavily on identifying and classifying abnormal tissue nodules in medical imaging. Using automated artificial intelligence models to evaluate these scans can assist clinicians in diagnostic workflows. Demonstrating that pre-trained deep learning models achieve high accuracy while saving training effort provides clear guidance for developing reliable computer-aided diagnostic systems.
The tested algorithms could form the foundation for clinical decision-support software used by radiologists and oncologists interpreting chest CT scans. This work represents applied research tested on a retrospective benchmark dataset of two-dimensional nodule cut-outs. Transitioning toward real-world commercial diagnostic tools would require training on larger datasets and expanding the models to process full three-dimensional volumetric lesion data.
AI-generated from the published abstract. Always read the original work before citing.
OBJECTIVES: This study aimed to explore and develop artificial intelligence approaches for efficient classification of pulmonary nodules based on CT scans. MATERIALS AND METHODS: A number of 1007 nodules were obtained from 551 patients of LIDC-IDRI dataset. All nodules were cropped into 64 × 64 PNG images , and preprocessing was carried out to clean the image from surrounding non-nodular structure. In machine learning method, texture Haralick and local binary pattern features were extracted. Four features were selected using principal component analysis (PCA) algorithm before running classifiers. In deep learning, a simple CNN model was constructed and transfer learning was applied using VGG-16 and VGG-19, DenseNet-121 and DenseNet-169 and ResNet as pre-trained models with fine tuning. RESULTS: In statistical machine learning method, the optimal AUROC was 0.885 ± 0.024 with random forest classifier and the best accuracy was 0.819 ± 0.016 with support vector machine. In deep learning, the best accuracy reached 90.39% with DenseNet-121 model and the best AUROC was 96.0%, 95.39% and 95.69% with simple CNN, VGG-16 and VGG-19, respectively. The best sensitivity reached 90.32% using DenseNet-169 and the best specificity attained was 93.65% when applying the DenseNet-121 and ResNet-152V2. CONCLUSION: Deep learning methods with transfer learning showed several benefits over statistical learning in terms of nodule prediction performance and saving efforts and time in training large datasets. SVM and DenseNet-121 showed the best performance when compared with their counterparts. There is still more room for improvement, especially when more data can be trained and lesion volume is represented in 3D. CLINICAL RELEVANCE STATEMENT: Machine learning methods offer unique opportunities and open new venues in clinical diagnosis of lung cancer. The deep learning approach has been more accurate than statistical learning methods. SVM and DenseNet-121 showed superior performance in pulmonary nodule classification.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1186/s13244-023-01441-6
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
