Article Preview
TopIntroduction
Lung cancer is disease with highest death rate as compared to breast, prostate, brain and cervical cancers. Even with number of available detection methodologies for lung cancer, life span of patient with stage III and IV lung cancer is still not improving as expected. Thus early stage lung nodules detection followed by the proper treatment is best choice to avoid conversion of early stage nodules into malignant tumors and to reduce mortality rate. Lung nodules possible size ranges from 3mm to 30mm. Small, non-cancerous and smooth boundary lung nodules are benign nodules. Large, cancerous and irregular boundary lung tumors are knows as malignant nodules. Non-detected benign nodules may get converts into cancer tumors. Based on nodule’s intensity variation, texture and additional connected components which are not a part of nodules, lung nodules are differentiated into well circumscribed, Juxta-Vascular Nodules (JVN) - attached to blood vessels, Juxta-Pleural Nodules (JPN) - attached to lung pleural and Ground Glass Opacity (GGO) nodules as in Figure 1. Further nodules at next level of classification on their solidity feature are solid, partly solid or non-solid nodules as in Figure 2.
Figure 1. a) Well-Circumscribed Nodule b) Juxta-Vascular Nodule c) Juxta-Pleural Nodule
Figure 2. a) Solid nodule b) Part-solid nodule c) Non-solid nodule
Medical imaging modalities like X-ray, Magnetic Resonance Imaging (MRI), Diffusion Weight MRI (DW-MRI), Computed Tomography (CT), Ultrasound (US) and other modalities have been part of medical routine for lung cancer detection. On medical images lung nodules appears as white patch. CT is GOLD STANDARD modality for detection of all types and stages lung nodules. Figure 3 shows the pixels value of all possible parts of lung CT image. As CT generates huge amount of images in single scan, interpretation of these images by radiologist to study the characteristic of nodules for further treatment is time consuming. Core of medical procedures to confirm the presence and severity of lung cancer are Computer Aided Detection (CAD) systems. These systems provides useful and accurate information about nodules for a radiologist to draw the useful conclusion about lung nodules (characteristic, size, benign or malignant), thus improves treatment option. With a use of available technologies and to make quick diagnosis of cancer, CAD systems can be implemented to provide remote E-health care services (Ashutosh Sharma & Rajiv Kumar 2019).
Figure 3. Histogram of original CT image that shows the pixel values of fat, bones, lung parenchyma and background.
TopLiterature Review
(K. Suzuki & K. Doi, 2004; K. Suzuki & K. Doi, 2005; Suzuki, Z. H. Shi, & J. Zhang, 2008) explained implementation of new CAD system based on supervised filter – Massive Training Artificial Neural Network (MTANN) to detect and classify lung nodules. With this new approach 97% sensitivity was achieved with reduced false positive rate.
(S. Diciotti, G. Picozzi, M. Falchini, M. Mascalchi, N. Villari, & G. Valli, 2008) proposed novel segmentation technique based on significant properties such as expert’s knowledge, geodic distance computation and user interaction to segment nodules with lesser diameter (phantom 5.3-11 and vivo 5-9.8). This method also considered JVN and JPN segmentation. CT images with 157 nodules were taken from Lung Image Database Consortium (LIDC) public database and achieved sensitivity of 88.5%.
(Q. Wang, E. Song, R. Jin, P. Han, X. Wang, Y. Zhou, & J. Zeng, 2009) explained lung nodules segmentation algorithm based on information present in adjacent CT slices and multidirectional relationship to segment different types of nodules. Data set 1 and data set 2 of CT images from LIDC were used for validation of presented methodology and 75% accuracy was achieved.