Lung nodule detection and diagnosis using circle detection through plain radiographs

• Main Authors: De La Cruz, Ria Rodette, Roque, Trizia Roby Ann, Rosas, John Daryl, Vera Cruz, Charles Vincent M., Cordel, Macario O., II, Ilao, Joel P.
• Format: text
• Published: Animo Repository 2013
• Subjects: Pattern recognition systems; Image processing—Digital techniques; Radiography, Medical—Digital techniques; Lungs—Radiography; Lungs—Diseases—Diagnosis
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/5441
• Institution: De La Salle University

In this paper, we present a system that locates pulmonary nodules in digital chest radiographs through pattern recognition. Digital radiographs that are already diagnosed with lung nodules underwent histogram equalization in order to address varying illumination levels across different regions in the radiographs, and make the radiograph samples more comparable. Laplacian of Gaussian filtering is next applied in order to highlight the edges of pathological features like nodule-shaped blobs in each radiograph. Circular Hough Transform (CHT) was utilized in tandem with pixel-based image processing techniques in locating possible nodules. These system reports the count and sizes of the candidate nodules. We report an overall system accuracy of 73.33% when classifying digitized radiographs as either with nodules or without nodules.