Computer-Aided Diagnosis (CAD) to Detect Abnormalities in Lung Pediatric Radiography using Particle Swarm Optimization Method

M. L.E. Yuliansyah, P. Prajitno, D. S. Soejoko

Research output: Contribution to journalConference articlepeer-review

1 Citation (Scopus)


The diagnosis of lung organ requires accurate analysis and interpretation. Abnormal parts such as nodules are sometimes covered by other complex lung tissue that is normal tissue. Therefore innovation is needed in analyzing and classifying normal tissue and the nodule. This study developed a Computer-Aided Diagnosis (CAD) for radiographic of pediatric lung using segmentation Particle Swarm Optimization (PSO) method to detect the abnormality in lung image. Particle Swarm Optimization (PSO)-based segmentation method is combined with Fuzzy C-Means (FCM) clustering method and Wiener filter to refine the lung region and search for abnormalities, especially for pneumonia and tuberculosis, based on the value of the image pixel. The performance evaluation of this CAD was done by calculating the Receiver Operating Characteristics (ROC) using 136 images and compared with the reference from doctor evaluation. The overall error of this method is 11.43% or has an accuracy value of 88.57%, while its sensitivity is 90.00%, specificity is 85.00%, and precision is 93.75%. This method has a good success rate in detecting abnormal lung image. However, this segmentation method cannot detect abnormalities located on the edge of the lung, caused by the superposition of the rib image.

Original languageEnglish
Article number012003
JournalJournal of Physics: Conference Series
Issue number1
Publication statusPublished - 15 Jun 2020
Event3rd Annual Scientific Meeting on Medical Physics and Biophysics, PIT-FMB in conjunction with the 17th South-East Asia Congress of Medical Physics, SEACOMP 2019 - Bali, Indonesia
Duration: 8 Aug 201910 Aug 2019


Dive into the research topics of 'Computer-Aided Diagnosis (CAD) to Detect Abnormalities in Lung Pediatric Radiography using Particle Swarm Optimization Method'. Together they form a unique fingerprint.

Cite this