Background: A Computer Assisted Diagnosis (CAD) is the recently used framework
that supports for the early detection and diagnosis of the tumor affected parts of the human body.
The diversities in the shape, image intensities and the abnormalities make the manual approaches
as the challenging one during the reading process of chest Computed Tomography (CT) scans. The
lack of image clarity and the loss of edge information during the filtering process reduce the accuracy
of segmentation and detection of tumor regions in the lung images.
Discussion: To solve these issues, this paper proposes the integrated framework of Enhanced Hopfield
Neural Network with the Point Kernel-based Classifier (PKC) for lung tumor detection. Initially,
the pixel-intensity-based filtering removes the noise present in the image with edge information
preserving capability. After that, the lung region is extracted from the preprocessed image
with the use of proposed EHNN. Sequentially, the Local Tetra Patterns (LTrP) is employed to extract
the patterns for describing the image information clearly. Finally, the Point Kernel-based
Classifier (PKC) is used with the LTrP features for accurately classifying the normal and tumor affected
Conclusion: Moreover, the proposed EHNN-PKC is compared with the existing techniques by using
various measures for analyzing the effectiveness of EHNN-PKC.
Keywords: Background Normalization, Filtering, Hopfield Neural Network (HNN), Kernel-based Classification, Local Tetra Pattern (LTrP)
Rights & PermissionsPrintExport