Local perturbation responses and checkerboard tests: Characterization tools for nonlinear MRI methods

Abstract

Purpose: Modern methods for MR image reconstruction, denoising, and parameter mapping are becoming increasingly nonlinear, black-box, and at risk of "hallucination." These trends mean that traditional tools for judging confidence in an image (visual quality assessment, point-spread functions (PSFs), g-factor maps, etc.) are less helpful than before. This paper describes and evaluates an approach that can help with assessing confidence in images produced by arbitrary nonlinear methods. Theory and Methods: We propose to characterize nonlinear methods by examining the images they produce before and after applying controlled perturbations to the measured data. This results in functions known as local perturbation responses (LPRs) that can provide useful insight into sensitivity, spatial resolution, and aliasing characteristics. LPRs can be viewed as generalizations of classical PSFs, and are are very flexible-they can be applied to arbitary nonlinear methods and arbitrary datasets across a range of different reconstruction, denoising, and parameter mapping applications. Importantly, LPRs do not require a ground truth image. Results: Impulse-based and checkerboard-pattern LPRs are demonstrated in image reconstruction and denoising scenarios. We observe that these LPRs provide insights into spatial resolution, signal leakage, and aliasing that are not available with other methods. We also observe that popular reference-based image quality metrics (eg, mean-squared error and structural similarity) do not always correlate with good LPR characteristics. Conclusions: LPRs are a useful tool that can be used to characterize and assess confidence in nonlinear MR methods, and provide insights that are distinct from and complementary to existing quality assessments.

NORDIC Increases the Sensitivity and Preserves the Spatiotemporal Precision of fMRI Responses

“…For the latter, prior work found that NORDIC had no impact on the functional point spread of the BOLD signal (Vizioli et al, 2021). To further validate these results we performed an initial evaluation using the local perturbation response (LPR) method (Chan and Haldar, 2021) and did not observe any image blurring on a checkerboard perturbation (Supplemental Figure S9, S11).…”

“…For the latter, prior work found that NORDIC had no impact on the functional point spread of the BOLD signal (Vizioli et al, 2021). To further validate these results we performed an initial evaluation using the local perturbation response (LPR) method (Chan and Haldar, 2021) and were able to recover the injected synthetic sparse signal, though sufficiently low intensity perturbations (i.e. below or near thermal noise level) were not perfectly recovered (See Supplement, Figures S8, S13).…”

“…Across all datasets, the NORDIC data were better able to predict the Standard data as measured by the coefficient of determination, including datasets using acquisition methods for which the thermal noise contribution is lower (i.e., datasets with larger voxels) relative to physiological fluctuations. In order to further examine the possibility of a loss of temporal precision we also probed the neighboring timepoints in the previously mentioned LPR analysis (Chan and Haldar, 2021) and did not observe any temporal spreading of the checkerboard perturbation (Supplemental Figure S9, S12).…”

Stabilizing deep tomographic reconstruction: Part A. Hybrid framework and experimental results

“…Our current evaluative metrics directly correspond to what were used in the Proceedings of the National Academy of Sciences ( PNAS ) study. 15 However, it will be valuable and interesting to assess the clinical influence of reconstructed results using other advanced assessment methods (local perturbation responses 54 and Frechet inception distance 55 ). In addition, the used deep tomographic networks are based on CNN architectures.…”

SNR Enhancement for Multi-TE MRSI Using Joint Low-Dimensional Model and Spatial Constraints

“…In addition, although the learned representation currently serves as a voxel-wise constraint, it can introduce timedependent bias into the final reconstruction thus frequencydependent spatial resolution. While the traditional metrics (like FWHM of point spread functions) are not sufficient to analyze such resolution effects with the nonlinearity of neural networks, a more careful spatial resolution analysis of the denoised reconstruction should be conducted in future research (e.g., using new analysis tools such as [59]).…”