Quality assessment of anatomical MRI images from Generative Adversarial Networks: human assessment and image quality metrics

Background: Generative Adversarial Networks (GANs) can synthesize brain images from image or noise input. So far, the gold standard for assessing the quality of the generated images has been human expert ratings. However, due to limitations of human assessment in terms of cost, scalability, and the limited sensitivity of the human eye to more subtle statistical relationships, a more automated approach towards evaluating GANs is required. New method: We investigated to what extent visual quality can be assessed using image quality metrics and we used group analysis and spatial independent components analysis to verify that the GAN reproduces multivariate statistical relationships found in real data. Reference human data was obtained by recruiting neuroimaging experts to assess real Magnetic Resonance (MR) images and images generated by a Wasserstein GAN. Image quality was manipulated by exporting images at different stages of GAN training. Results: Experts were sensitive to changes in image quality as evidenced by ratings and reaction times, and the generated images reproduced group effects (age, gender) and spatial correlations moderately well. We also surveyed a number of image quality metrics which consistently failed to fully reproduce human data. While the metrics Structural Similarity Index Measure (SSIM) and Naturalness Image Quality Evaluator (NIQE) showed good overall agreement with human assessment for lower-quality images (i.e. images from early stages of GAN training), only a Deep Quality Assessment (QA) model trained on human ratings was sensitive to the subtle differences between higher-quality images. Conclusions: We recommend a combination of group analyses, spatial correlation analyses, and both distortion metrics (SSIM, NIQE) and perceptual models (Deep QA) for a comprehensive evaluation and comparison of brain images produced by GANs.

Colourlab Image Database: Geometric Distortions

Over the years, a high number of different objective image quality metrics have been proposed. While some image quality metrics show a high correlation with subjective scores provided in different datasets, there still exists room for improvement. Different studies have pointed to evaluating the quality of images affected by geometrical distortions as a challenge for current image quality metrics. In this work, we introduce the Colourlab Image Database: Geometric Distortions (CID:GD) with 49 different reference images made specifically to evaluate image quality metrics. CID:GD is one of the first datasets which include three different types of geometrical distortions; seam carving, lens distortion, and image rotation. 35 state-ofthe-art image quality metrics are tested on this dataset, showing that apart from a handful of these objective metrics, most are not able to show a high performance. The dataset is available at <ext-link ext-link-type="url" xlink:href="http://www.colourlab.no/cid">www.colourlab.no/cid</ext-link>.

Establishing scanning protocols for a CT lung cancer screening trial in the UK

Objectives: To develop a CT scanning protocol for lung cancer screening which achieved low radiation dose and a high level of objectively assessed image quality. Methods: An anthropomorphic chest phantom and a commercially available lung screening image quality phantom were scanned on a series of scan protocols from a previous UK lung screening pilot and on an alternative protocol. The chest phantom scans were used to assess the CT dose metrics on community-based mobile CT scanners and comparisons were made with published recommended doses. Scans of the image quality phantom were objectively assessed against the RSNA Quantitative Imaging Biomarkers Alliance (QIBA) recommendations. Protocol adjustments were made to ensure that the recommended dose and image quality levels were both achieved. Results: The alternative scan protocol yielded doses up to 72% lower than on the previously used protocols with a CTDIvol of 0.6mGy for the 55 kg equivalent phantom and 1.3mGy with an additional 6 cm of tissue equivalent material in place. Scans on the existing protocols failed on two of the QIBA image quality metrics (edge enhancement and 3D resolution aspect ratio). Following adjustments to the reconstruction parameters of the resulting image quality met all six QIBA recommendations. Radiologist review of phantom images with this scan protocol deemed them suitable for a lung screening trial. Conclusions: Scan protocols yielding low radiation doses and high levels of objectively assessed image quality which meet published criteria can be established through the use of specific anthropomorphic and image quality phantoms, and are deliverable in community-based lung cancer screening. Advances in knowledge: Development of a standard methodology for establishing CT lung screening scanning protocols Use of QIBA recommendations as objective image quality metrics Standardised lung phantoms are essential tools for setting up lung screening protocols

Quantitative Image Quality Metrics of the Low-Dose 2D/3D Slot Scanner Compared to Two Conventional Digital Radiography X-ray Imaging Systems

The aim of this study was to determine the quantitative image quality metrics of the low-dose 2D/3D EOS slot scanner X-ray imaging system (LDSS) compared with conventional digital radiography (DR) X-ray imaging systems. The effective detective quantum efficiency (eDQE) and effective noise quantum equivalent (eNEQ) were measured using chest and knee protocols. Methods: A Nationwide Evaluation of X-ray Trends (NEXT) of a chest adult phantom and a PolyMethylmethacrylate (PMMA) phantom were used for the chest and knee protocols, respectively. Quantitative image quality metrics, including effective normalised noise power spectrum (eNNPS), effective modulation transfer function (eMTF), eDQE and eNEQ of the LDSS and DR imaging systems were assessed and compared. Results: In the chest acquisition, the LDSS imaging system achieved significantly higher eNEQ and eDQE than the DR imaging systems at lower and higher spatial frequencies (0.001 > p ≤ 0.044). For the knee acquisition, the LDSS imaging system also achieved significantly higher eNEQ and eDQE than the DR imaging systems at lower and higher spatial frequencies (0.001 > p ≤ 0.002). However, there was no significant difference in eNEQ and eDQE between DR systems 1 and 2 at lower and higher spatial frequencies (0.10 < p < 1.00) for either chest or knee protocols. Conclusion: The LDSS imaging system performed well compared to the DR systems. Thus, we have demonstrated that the LDSS imaging system has the potential to be used for clinical diagnostic purposes.

Quality Assessment of 2.5D Prints Using 2D Image Quality Metrics

Quality assessment is an important aspect in a variety of application areas. In this work, the objective quality assessment of 2.5D prints was performed. The work is done on camera captures under both diffuse (single-shot) and directional (multiple-shot) illumination. Current state-of-the-art 2D full-reference image quality metrics were used to predict the quality of 2.5D prints. The results showed that the selected metrics can detect differences between the prints as well as between a print and its 2D reference image. Moreover, the metrics better detected differences in the multiple-shot set-up captures than in the single-shot set-up ones. Although the results are based on a limited number of images, they show existing metrics’ ability to work with 2.5D prints under limited conditions.