Lung Segmentation using Active Shape Model to Detect the Disease from Chest Radiography.

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Parametric models aid lung segmentation for disease diagnosis.
• Accurate lung segmentation is crucial for detecting abnormalities like nodules and tuberculosis.

Purpose of the Study:

• To enhance the detection of lung diseases, specifically nodules and tuberculosis, using an active shape model (ASM) on chest radiographs.
• To evaluate the efficacy of ASM in segmenting and identifying pathological findings in lung images.

Main Methods:

• Employed six grouping methods (physicians, Dice similarity, correlation coefficients, and three SVM-based methods) for classifying chest radiographs.
• Selected the most effective grouping method, validated by a radiologist, as input for ASM-based image segmentation.
• Assessed segmentation accuracy using various parameters on the Japanese Society of Radiological Technology (JSRT) and tuberculosis databases.

Main Results:

• The ASM achieved high detection rates for pulmonary nodules: 94.12% (left lung) and 94.38% (right lung).
• The ASM model also demonstrated significant accuracy in detecting tuberculosis: 88.33% (left lung) and 90.37% (right lung).

Conclusions:

• The ASM segmentation method, coupled with pre-segmentation grouping, serves as a valuable preliminary step for identifying tuberculosis and pulmonary nodules.
• The presented approach shows potential for future applications, including the measurement of cardiac size and dimensions.