A novel Computer Aided Detection for identifying lung nodules on chest radiographs

Lung cancer is the leading cause of cancer death in the United States. It usually exhibits its presence with the formation of pulmonary nodules. Nodules are round or oval-shaped growth present in the lung. Chest radiographs are used by radiologists to detect and treat such nodules but they are quite difficult to detect with human eye and are sometimes misinterpreted with lesions present. Thus, automated analysis of such data is very essential and would be of valuable help in lung cancer screening. A new computer aided detection (CAD) system in chest radiography is proposed in this paper. The algorithmic steps include (i) local contrast enhancement; (ii) automated anatomical segmentation; (iii) detection of nodule candidates; (iv) feature extraction; (v) candidate classification. In this research, we present facets of the proposed algorithm using a publicly available dataset and we explore into new set of features and classifiers. The publicly available database was created by the Standard Digital Image Database Project Team of the Scientific Committee of the Japanese Society of Radiological Technology (JRST). The JRST dataset comprises of 154 chest radiographs containing one radiologist confirmed nodule each. In this research, we compute a rich set of 117 features for each potential candidate. Local contrast enhancement is achieved using a Gaussian low pass filter. Anatomical segmentation is performed using an active shape model. Potential candidate nodules can then be determined by using an adaptive distance- based threshold algorithm limited to delineated lung fields. Later, a set of features are computed for each potential candidate. Based on those tailored features, a classifier/neural network system can be used to identify the candidates as either true positives or false positives. This CAD system would aid in providing a second opinion to radiologists. Algorithm will be trained using Riverain Database and would be tested later in JRST database.https://ecommons.udayton.edu/stander_posters/1794/thumbnail.jp