No‐reference image quality assessment of magnetic resonance images with high‐boost filtering and local features

Abstract

Purpose: Subjective quality assessment of displayed magnetic resonance (MR) images plays a key role in diagnosis and the resultant treatment. Therefore, this study aims to introduce a new no-reference (NR) image quality assessment (IQA) method for the objective, automatic evaluation of MR images and compare its judgments with those of similar techniques. Methods: A novel NR-IQA method was developed. The method uses a sequence of scaled images filtered to enhance high-frequency components and preserve lowfrequency parts. Since the human visual system (HVS) is sensitive to local image variations and local features often mimic the attraction of the HVS to high-frequency image regions, they were detected in the filtered images and described. Then, the statistics of obtained descriptors were used to build a quality model via the Support Vector Regression method. Results: The method was compared with 21 state-of-the-art techniques for NR-IQA on a new dataset of 70 distorted MR images assessed by 31 experienced radiologists, using typical evaluation criteria for the comparison of NR measures. The introduced method significantly outperforms the compared approaches, in terms of the correlation with human judgments. Conclusions: It is demonstrated that the presented NR-IQA method for the assessment of MR images is superior to the state-of-the-art NR techniques. The method would be beneficial for a wide range of image processing applications, assessing their outputs and affecting the directions of their development. K E Y W O R D Shigh-boost filtering, image quality assessment, local features, magnetic resonance images, no-reference, subjective tests

Fusion of Deep Convolutional Neural Networks for No-Reference Magnetic Resonance Image Quality Assessment

“…Volumetric and artifact-specific features were used by Pizarro et al [10] to train the SVM classifier. In previous authors' works on the MRIQA, the entropy of local intensity extrema was used for direct quality prediction [7] or high-boost filtering followed by the detection and description of local features [8] was used with an SVR-based quality model.…”

“…The proposed approach is evaluated on two MRIQA benchmark datasets. The first dataset, denoted for convenience as DB1, contains 70 MR images [8], while the second one (DB2) has been created for the needs of this study and contains 240 MR images.…”

“…The collection consists of images captured under different conditions affecting the image quality (IPAT software to made GeneRalized Autocalibrating Partially Parallel Acquisitions (GRAPPA); GRAPPA3 [41]). Apart from the images, the benchmark contains the MOS, ranging from 1 to 5, which was obtained in tests with a group of radiologists [8]. The resolution of images in the dataset is between 192 × 320 and 512 × 512.…”

“…The introduced approach is represented by three fusion models: Resnet-50_GoogLeNet _ResNet-18 (R50GR18), ResNet-50_GoogLeNet_MobileNet-V2 (R18GR50M), and MobileNet-V2_ResNet-50 (MR50). They are experimentally compared with 17 state-of-the-art techniques: NFERM [47], SEER [20], DEEPIQ [27], MEON [46], SNRTOI [48], NOREQI [22], BPRI [24], HOSA [17], NOMRIQA [8], IL-NIQE [23], GM-LOG [18], GWHGLBP [21], BRISQUE [15], SISBLIM [49], metricQ [50], SINDEX [51], and ENMIQA [7]. The NOM-RIQA, ENMIQA, and SNRTOI are designed for MR images, whereas DEEPIQ and MEON are deep learning approaches devoted to natural images.…”

“…Furthermore, some early methods adapt solutions from the IQA of natural images [5], train them on images assessed by the SNR [2], or add additional features to characterize MR images [6]. Some approaches employ the entropy of local intensity extrema [7] or specific image filtering to facilitate the usage of local features [8]. Other works are devoted to a binary classification of noisy MR images [9,10] or for classification of images with prior detection of selected distortion types [11,12].…”