Fractional mesh-free linear diffusion method for image enhancement and segmentation for automatic tumor classification. (April 2020)

• Record Type: Journal Article
• Authors: Chandra, Saroj Kumar; Bajpai, Manish Kumar
• Is Part Of: Biomedical signal processing and control. Volume 58(2020)
• Publication Date: 2020-04
• Subjects: Linear diffusion -- Partial differential equation -- Radial basis function -- Brain tumor segmentation -- Image enhancement -- Computer aided diagnostic; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2019.101841
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A novel fractional mesh-free diffusion method has been proposed in the present manuscript for image enhancement and segmentation. An unconditional Crank–Nicolson finite difference scheme has been used to solve proposed methods. Alternate Direction Implicit (ADI) approach has been applied to simplify Crank–Nicolson finite difference method. The proposed methods (i.e. image enhancement and segmentation) have been used in CAD model for automatic tumor classification. Abstract: Computer aided diagnostic (CAD) models have shown outstanding performance in identifying many kind of diseases. Tumor identification is one of the most useful application of CAD model. Benign and malignant are two categories of tumor cells. Both categories share some textural features due to which tumor classification becomes complex and difficult task. In the present manuscript, a novel CAD model is being presented for classifying tumor cells automatically. The proposed model has been divided into four modules (image enhancement, segmentation, feature extraction and classification). All the modules are equally important in tumor classification. The manuscript focuses on effect of first two modules namely image enhancement and segmentation on tumor classification. In the present work, a novel linear fractional mesh-free partial differential equation (FPDE) based image enhancement method has been proposed to improve quality of the images. The proposed enhancement model is able to preserve fine details in smooth areas also non-linearly increases high frequency information. A novel fractional mesh-free segmentation method has been proposed to extract tumor region. It has been found that the method is able segment tumor region more accurately. Thirteen textural features have been used for training and testing. Support vector machine (SVM) classifier has been used to classify extracted tumor region. Quantitative analysis of proposed image enhancement, segmentation and CAD model have been done with other popular models. Higher performance has been observed using proposed models.