Kidney segmentation in MR images using active contour model driven by fractional-based energy minimization

  • PDF / 1,300,444 Bytes
  • 8 Pages / 595.276 x 790.866 pts Page_size
  • 86 Downloads / 193 Views

DOWNLOAD

REPORT


ORIGINAL PAPER

Kidney segmentation in MR images using active contour model driven by fractional-based energy minimization Ala’a R. Al-Shamasneh1 · Hamid A. Jalab1 · Palaiahnakote Shivakumara1 · Rabha W. Ibrahim2,3 · Unaizah H. Obaidellah1 Received: 22 September 2019 / Revised: 29 February 2020 / Accepted: 10 March 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020

Abstract In the field of diagnosis and treatment planning of kidney-related diseases, accurate kidney segmentation is challenging due to intensity inhomogeneity caused by imperfections during image acquisition process. This study presents a model, which consists of a novel fractional energy minimization for segmenting human kidney organ from MR images. Unlike existing active contour models (Chan–Vese), which uses gradient-based energy minimization that is sensitive to inhomogeneous intensity values, we propose a novel fractional Mittag–Leffler’s function for energy minimization, a technique more suitable to cope with the mentioned challenges. The proposed model exploits the special property of fractional calculus in maintaining high-frequency contour features while enhancing the low-frequency detail of texture in smooth area. Experimental results performed on complex kidney images using the proposed method show that the proposed model outperforms the existing models in terms of sensitivity, accuracy, Jaccard index and Dice coefficient. Keywords Fractional calculus · Mittag–Leffler’s function · Medical imaging · Active contour · Kidney segmentation

1 Introduction Nowadays, medicine and health care are among the most important research areas that serve as a foundation for efficient expert systems. For example, the invention of advanced technologies and automatic computer-aided tools

B

Hamid A. Jalab [email protected] Ala’a R. Al-Shamasneh [email protected] Palaiahnakote Shivakumara [email protected] Rabha W. Ibrahim [email protected] Unaizah H. Obaidellah [email protected]

1

Faculty of Computer Science and Information Technology, University Malaya, Kuala Lumpur, Malaysia

2

Informetrics Research Group, Ton Duc Thang University, Ho Chi Minh City, Vietnam

3

Faculty of Mathematics and Statistics, Ton Duc Thang University, Ho Chi Minh City, Vietnam

have become an integral part of biomedical image analysis to reduce and facilitate the labor work of medical practitioners. One such use of biomedical imaging is on kidney image analysis to study various factors related to kidney diseases, such as chronic kidney disease and acute kidney injury. Often, kidney-related diseases cause loss of renal function leading to kidney failure, mortality and other severe complications (i.e., cardiovascular diseases). As a result, expert systems for segmenting accurate kidney from the MR image is essential, which can assist practitioners in predicting kidney problems. In processing the MR images, one of crucial tasks is segmentation, an area which is still considered an open issue due to the characteristically high inhomogeneity and complex f