Multi-scale lung tissue classification for interstitial lung diseases using learned Gabor filters

• Published in: Microsystem technologies : sensors, actuators, systems integration actuators, systems integration, 2023-04, Vol.29 (4), p.599-607
• Main Authors: Dasari, Nageshbabu, Reddy, B. V. Ramana
• Format: Article
• Language: English
• Subjects: Electronics and Microelectronics; Engineering; Instrumentation; Mechanical Engineering; Nanotechnology; Technical Paper
• ISSN: 0946-7076; 1432-1858
• DOI: 10.1007/s00542-023-05413-0

Early prediction of Interstitial Lung Diseases (ILD) is essential for the treatment and even it is a challenging problem for the trained radiologists to diagnosis the particular lung tissue pattern because they exhibit similar characteristics with other tissue patterns and dissimilar characteristics with in the same tissue patterns. With the outstanding performance of deep Convolution Neural Network (CNN) in the field of medical image analysis, we designed an automatic Computer Aided Design (CAD) system for the classification of five categories of lung tissue patterns such as Normal, Ground Glass, Emphysema, Micronodules and Fibrosis, which are major among the Interstitial Lung Diseases (ILD). We present a novel and robust Multi-scale Convolution Neural Network (M-CNN) model with input is a texture-enhanced image obtained by applying Gabor filter, instead of raw image patch. The proposed method is evaluated on the publicly available ILD dataset and experiment result shows an outstanding performance with 90.67% accuracy, 92.1% Precision, and 90.14% average F-Score compared to the state-of-art.