Computer-Aided Detection of Pulmonary Embolism: An Ant Colony Optimized Wavelet Neural Network Model

Abstract

Pulmonary Embolism (PE) is the disease outcome as of a blood clogging within the legs that travels to
the respiratory organ (lungs). The chance of blood clots is accrued by cancer, prolonged bed rest,
smoking, stroke, bound genetic conditions. Some other positive cases which includes embolization of air,
fat, or amniotic fluid in the body. Diagnosing relies on signs and symptoms together with check results.
This paper inspects the replacing the conventional classification approach intended for diagnosis of PEs
in contrast-enhanced CT pictures with the help of Ant Colony Optimized Wavelet Neural Network
(ACOWNN) approach. The nominated abnormality segmentation rule relies on applying normalized grey
level co-occurrence matrix to segregate the texture features from the Scanned images and categorizing
the doubtful sections by exploiting an optimized classifier. Consequently the segregation of false
detections, like tissue and parenchyma diseases was made successfully. The system was trained with 20
CT medical data sets and tested with 11 different CT images illustrated that this model generalized well
for this application. The accuracy of this Computer Aided Diagnosis (CAD) scheme is evaluated by
means of Receiver operative Characteristic (ROC) curve. With this curvature we can calculate the
sensitivity and specificity rates which can be used to identifying the true and false positives from the given
data sets. The result from the ROC curve of this proposed method yields a sensitivity of 94.16% and
specificity of 92.10%.