Multi-objective noise estimator for the applications of de-noising and segmentation of MRI data. (September 2018)

• Record Type: Journal Article
• Authors: Singh, Munendra; Verma, Ashish; Sharma, Neeraj
• Is Part Of: Biomedical signal processing and control. Volume 46(2018)
• Page Start: 249
• Page End: 259
• Publication Date: 2018-09
• Subjects: Multi-objective particle swarm optimisation -- Noise estimation -- Kalman filter -- de-noising -- MRI; 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.2018.07.012
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The proposed noise estimation approach is adaptive and based on MOPSO. The maximisation of the image quality measures helps to achieve de-noising along with enhancement of the image features. To de-noise the MRI data, the present study proposes the MOPSO based non-local Kalman and MOPSO based LMMSE filters. The de-noising results have been compared with filters like Wavelet, Wiener, non-local mean, local Kalman and LMMSE filter. To show the impact of de-noising on segmentation of MRI data, enhanced FCM algorithm has been applied on filtered MRI data. Abstract: The present study proposes the noise estimation of Magnetic Resonance Imaging (MRI) data using multi-objective particle swarm optimisation (MOPSO). This adaptive noise estimation is based on the maximisation of the multiple quality measures, which enable the algorithm to achieve de-noising along with enhancement in the image features. The paper proposes two filtering approaches to de-noise MRI data. In first, MOPSO based noise estimation is followed by non-local statistics based Kalman filter, whereas, in the second approach, MOPSO based noise estimation is followed by Linear Minimum Mean Square Error (LMMSE) filter. The impact of de-noising on segmentation of MRI data has also been studied, for this purpose enhanced fuzzy c-means algorithm has been applied on filtered MRI data. The de-noising and segmentation performance of MOPSO-non local Kalman filter and MOPSO-LMMSE filters has been evaluated and compared with Wavelet filter, Wiener filter, non-local mean filter, standard Kalman and standard LMMSE filter. The proposed noise estimation approach followed by filtering is giving better de-noising and segmentation results as compared to standard filters considered.