Page 27 - TD-2-1

Tumor Discovery An approach for classification of lung nodules

CT imaging is less expensive and produces a variety of using the point spread function. Deep et al. presented
[15]
cross-sectional images of complete chest within a single an algorithm for segmenting different nodule types such
breath hold. For the analysis and early detection of lung as juxtravascular, pleura tail, juxtapleural, solid, and non-
nodules, it is now regarded as the best imaging technique. solid nodules.
Majority of the pulmonary nodules are benign; however, Segmentation technique such as region growing and
a small populace of them grow to be malignant. The some hybrid fuzzy connectivity models was implemented.
radiologists examine the CT scan to conclude whether a Another segmentation scheme proposed by Dehmeshki
nodule presents a chance for malignancy. The radiologist et al. used two different datasets taken from the LIDC
[16]
finds it challenging and time-consuming to detect some database. Two different techniques such as dynamic
[6]
nodules in the CT because of non-pathological features . programming model and multidirection fusion techniques
To get around this, radiologists opt for a computer- are used to know the information, relationship between
assisted approach as a backup method to validate their adjacent slices and to reduce segmentation error.
interpretation. Computer-based processes like computer- Dehmeshki et al. utilized a three-step procedure to find
[17]
aided detection (CAD) act as the radiologist’s “second pair lung nodules. Initially, lung regions are segmented using
of eyes” to analyze medical pictures for any problematic an adaptive threshold algorithm. Second, lung vessel was
regions. CAD is a topical technique designed to improve removed using active contour model (ACM) and finally,
the radiologists’ ability to find even the smallest lung the suspicious nodules were located using a Hessian
nodules at their earliest stages. matrix (selected shape filter). Delogu et al. developed
[18]
The main objective of CAD is to improve disease a fully automated segmentation method for detecting
identification by lowering the false negative (FN) rate the pulmonary nodule. The authors used region growing
brought on by observational omission. CAD was created approach for a set of 130 CT images. Only 84 images
[19]
with the purpose of improving systematic clinical decision- produced satisfactory result. Dheepak et al. proposed
making and the performance of detection in medical a methodology for segmentation of juxtapleural lung
imaging modalities. The aim of this paper is to focus on the nodules. The authors used two techniques for detecting
architecture of different stages of CAD design with fruitful the nodules from the lung. They are region growing and
[20]
results to assist the radiologists in detecting lung nodule at shape curvature-based techniques. Doi et al. developed
early stage. a method for segmenting the lung with the help of CT data.
The method was fully automatic and was composed of two
2. Literature survey steps. They are robust active shape model (RASM) and
optimal surface finding method. Dolejsi et al. proposed
[21]
In the field of lung cancer and in relation to the work a CAD to reduce the lung volume and juxtapleural nodule
of this study, there are a number of existing models and from thoracic CT images. For segmenting the lung
algorithms. The stages of CAD systems can be used to volume and nodule in juxtapleural, region growing and a
categorize the existing work. 3D-mass-spring mdel (MSM) was used. Elizabeth et al.
[22]
Various techniques, such as region growing [7,8] , watershed proposed a region-based ACM based on local divergence
segmentation , fuzzy logic, active contours, intensity-based energies. This model was designed for blurred boundary
[9]
thresholding, graph search algorithm , etc., were used for and noisy images. The author used regularization function
[10]
region of interest (ROI) image segmentation Segmentation to smoothen the boundary from different noise level.
is an essential step in nodule detection. The survey in The system performance was evaluated with Chan-Vese’s
various existing segmentation techniques for lung nodule (CV) model, region scalable fitting and local Gaussian
[23]
is discussed in detail. Dai et al. developed a segmentation fitting. Enquobahrie et al. developed an edge detection
[11]
algorithm known as segmentation by registration. The model, which precisely diffuses the edge space. Farag
[24]
authors compared their work with algorithms such as et al. proposed a newly developed nodule segmentation
automatic region growing, interactive region growing and algorithm which was stable, accurate, and automated. Farag
[25]
voxel classification. Daneshmand et al. demonstrated a et al. developed a segmentation model to segment the
[12]
precise technique designed for toning lung nodules in chest juxtavascular and ground glass (GG) nodules. The authors
CT scans. The region growing, optimal thresholding and proposed parametric mixture model for juxtavascular
optimal cube registration were used in this system Dawoud nodules and ACM for detecting leakage boundary.
et al. proposed an adaptive border marching which is a A nonlinear level set method proposed by Farag
[13]
geometric-based algorithm used for segmenting the lung. et al. used adaptive velocity function and edge stopping
[26]
De Nunzio et al. performed phantom experiment for CT function to employ a noise-free segmentation model.
[14]
image. The density of the nodule and its size was measured Gambhir et al. focused on segmenting the lung nodule
[27]
Volume 2 Issue 1 (2023) 2 https://doi.org/10.36922/td.317