Early detection of lung diseases is integral to improving patient outcomes, yet traditional diagnostic methods often fall short. A new study published on March 13, 2025, showcases significant advancements made using machine learning (ML) and deep learning (DL) techniques to classify lung X-ray images, helping medical practitioners identify whether the cases are benign or malignant alongside specific diseases, including atelectasis, pneumonia, and nodules.
Lung diseases present substantial global health challenges, with the World Health Organization estimating approximately 334 million people suffer from asthma, and others leading to about 1.4 million deaths annually from tuberculosis and 1.6 million from lung cancer. The pressing need for effective early diagnosis of these conditions has never been more evident, particularly as the COVID-19 pandemic also illuminated the burden these diseases impose on healthcare systems.
The researchers utilitized the National Institutes of Health (NIH) Chest X-ray dataset, which contains over 112,000 images formatted with labels for various lung diseases. From this extensive dataset, the study filtered 8,000 images—3,000 benign and 5,000 malignant—for training and testing their classification algorithms, partitioning the data at 80% for training and 20% for testing.
The study employed several machine learning classifiers, including k-nearest neighbors (kNN), support vector machines (SVM), decision trees, and Naïve Bayes, as well as the deep learning model called _inception v3_. Remarkably, the deep learning model achieved accuracy levels of 97.05%, outperforming traditional ML approaches by as much as 11.8%.
Among the classifiers tested, the SVM algorithm with the Radial Basis Function (RBF) kernel emerged as the top performer, achieving 85% accuracy with the malignant images. Such results illuminate how deep learning, particularly through convolutional neural networks (CNNs), can revolutionize the analysis of complex medical images, particularly for life-threatening conditions like lung cancer.
Machine learning algorithms have transformed the way data is processed, enabling them to recognize patterns and make predictions with impressive accuracy. Utilizing advanced methods such as the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), the researchers effectively ranked the machine learning classifiers and established quantifiable metrics for comparison.
By merging various attributes through statistical and textural features attained from the images, the SVM classifier was able to generate classifiers with higher efficiency. Notably, the classification accuracy rose significantly when using combined GLCM and LBP features against individual feature systems, showcasing how broad structural information and finer texture elements work coherently to improve model capabilities.
The findings offer promising avenues for enhancing early detection systems for lung diseases, utilizing AI and imaging technologies as invaluable tools. With the ability to process substantial volumes of data and identify subtle signs of lung conditions, these methods represent the next evolutionary step toward more effective diagnostic and treatment options.
Overall, the study reflects the growing integration of artificial intelligence within healthcare and the transformative potential it holds for the future of medical diagnostics. Continuing to refine these technologies and methodologies positions the modern medical community closer to combating the fierce battle against lung diseases—one classified X-ray at a time.