Performance analysis of bias correction techniques in brain MR images
- PDF / 1,247,940 Bytes
- 7 Pages / 595.276 x 790.866 pts Page_size
- 122 Downloads / 223 Views
ORIGINAL RESEARCH
Performance analysis of bias correction techniques in brain MR images Farzana1,2
•
Mohamed Sathik1,2 • Shajun Nisha1,2
Received: 27 September 2019 / Accepted: 26 June 2020 Ó Bharati Vidyapeeth’s Institute of Computer Applications and Management 2020
Abstract Bias field is a smooth intensity variation, which emanate during the process of image procurement. Bias removal is an essential prerequisite while incorporating computer assisted diagnosis. Several bias correction algorithms are proposed till date. This paper scrutinizes the prominent bias correction algorithms, LEMS, MICO, BCFCM and N3. These investigations carried over the substantial amount of T1 and T2 weighted brain MR images with different bias spectrum from Brain Web website. Algorithms efficiency are analyzed in spectral wise, slice wise, and type wise. Based upon the performance indicators coefficient of variation and coefficient of joint variation, the algorithms are assessed and ranked. The result concludes that which algorithm exterminates the bias field, presents in brain MR images in an efficient manner. Keywords Bias field BCFCM IIH INU LEMS MICO N3
1 Introduction There are several non invasive medical imaging modalities are available such as X-ray, computed tomography, magnetic resonance imaging (MRI), radiography, medical ultrasonography, single photon emission computed tomography (SPECT), and positron emission tomography & Farzana [email protected] 1
Department of Computer Science, Sadakathullah Appa College, Tirunelveli, India
2
Affiliation of Manonmaniam Sundaranar University, Abhishekapatti, Tirunelveli, Tamilnadu 627012, India
(PET), which helps to study and diagnose the anatomy related specifics. Among these, MRI is most preferred imaging modality because of its non ionizing nature and factual portrayal character for soft tissues. Acquired medical MR images are often corrupted by bias field. It arises due to imperfection in magnetic field of the scanner or patient movement during the scan process [1–6]. This can be defined as an intensity overlap. Bias is also named as intensity in homogeneity (IIH), intensity non uniformity (INU). Bias field hardly noticeable to the beholder. However it becomes prominent issue while working over image processing techniques such as registration, segmentation, etc., [7]. Eliminating the bias field is an essential pre processing step. Correction techniques are branched under two categories namely prospective and retrospective [8]. First one is used to amend scanner related bias filed and later one is for anatomy related bias field correction. Many prospective and retrospective approaches have been recommended till date. Prospective based bias field can be exempted, by constructing special hardware sequences such as multicoil imaging and phantom based calibration. These techniques fail to erode patient dependent bias. There are several retrospective techniques available to correct IIH such as filtering, segmentation, histogram and surface based bias correction
Data Loading...
