Title:Lung Cancer Detection from CT Images: Modified Adaptive Threshold
Segmentation with Support Vector Machines and Artificial Neural Network
Classifier
Volume: 20
Author(s): Sneha S. Nair*, V. N. Meena Devi and Saju Bhasi
Affiliation:
- Department of Physics, Noorul Islam Centre for Higher Education, Kumarakovil, Kanyakumari District, Tamilnadu, India
Keywords:
Accuracy, Classifier, Computed tomography, Diagnosis, Lidc dataset, Lung cancer.
Abstract:
Background:
The most difficult aspect of diagnosing lung cancer is early diagnosis. According to the American Cancer Society, each year, there are around 11
million newly diagnosed instances of cancer worldwide. Radiologists often turn to Computed Tomography (CT) scans to diagnose respiratory
conditions, which can reveal if lung tissue remains normal or abnormal. However, there is an increased chance of inaccuracy and delay; therefore,
radiologists are concerned with the physical segmentation of nodules.
Objective:
The objective of the research is to implement an advanced modified threshold segmentation and classification model for early and accurate
detection of lung cancer from CT images.
Methods:
Using the Support Vector Machines (SVM) classifier as well as the Artificial Neural Network (ANN) classifier, the authors propose using
Modified adaptive threshold segmentation as a segmentation approach for cancer detection. Here, Lung Image Database Consortium (LIDC)
datasets, a collection of CT scans, are used as the video frames in an investigation to authorize the recitation of the suggested technique.
Results:
Both quantitative as well as qualitative analyses are used to analyze the segmentation function of the anticipated algorithm. Both the ANN and
SVM classifiers used in the suggested technique for lung cancer diagnosis achieve world-record levels of accuracy, with the former achieving a
96.3% detection rate and the latter a 97% rate of accuracy.
Conclusion:
This innovation may have a major impact on the worldwide rate of lung cancer rate due to its ability to detect lung tumors in their earliest stages
when they are most amenable to being avoided and treated. This method is useful because it provides more information and facilitates quick,
precise decision-making for doctors diagnosing lung cancer in their patients.
