Supervised denoising of diffusion-weighted magnetic resonance images using a convolutional neural network and transfer learning

Jurek, Jakub; Materka, Andrzej; Ludwisiak, Kamil; Majos, Agata; Gorczewski, Kamil; Cepuch, Kamil; Zawadzka, Agata

doi:10.1016/j.bbe.2022.12.006

Cited by 10 publications

(3 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A slightly alternative view is that model fitting can be considered a denoising strategy on its own and one can denoise based on a forward (non-)parametric prediction, e.g. (Xiang T. et al, 2023;Jurek et al, 2023a;Tian et al, 2021;Fadnavis et al, 2020). Following this approach, the measured signal informs a prediction maker, which generates a deterministic sample from what the data for a diffusion-weighted volume should look like given all the other volumes.…”

Section: Denoising As Filtering or Deterministic Prediction?mentioning

confidence: 99%

“…A number of interesting questions regarding denoising applications remain open and we list here some of them as potential future work using the framework proposed. For instance, the interaction of denoising with other pre-processing steps can be evaluated (especially those that increase the rank of the non-thermal noise matrix and affect PCA-based methods), as explored with phase-correction (Jurek et al, 2023b;Liu et al, 2022;Cole et al 2021;Pizzolato et al, 2020;Cordero-Grande et al, 2019) or motion correction previously (Cieslak et al, 2022;Moeller et al, 2021a;Schilling et al, 2021). In addition, it would be interesting to explore whether signal transformations aiming to reduce the noise-floor bias "post-denoising" (e.g., Koay's method of moments (Koay and Basser, 2006;Koay et al, 2009), the Variance Stabilization Transform (Ma et al, 2020b)), including the noise-floor as part of the model fitting (Jbabdi et al, 2012)) are as advantageous as performing denoising in the complex domain, as this has been generally explored only for magnitude-based denoising (Hutchinson et al, 2017).…”

Section: Limitations and Future Workmentioning

confidence: 99%

See 1 more Smart Citation

Denoising Diffusion MRI: Considerations and implications for analysis

Manzano-Patron

Moeller

Andersson

et al. 2023

Preprint

View full text Add to dashboard Cite

Development of diffusion MRI (dMRI) denoising approaches has experienced considerable growth over the last years. As noise can inherently reduce accuracy and precision in measurements, its effects have been well characterised both in terms of uncertainty increase in dMRI-derived features and in terms of biases caused by the noise floor, the smallest measurable signal given the noise level. However, gaps in our knowledge still exist in objectively characterising dMRI denoising approaches in terms of both of these effects and assessing their efficacy. In this work, we reconsider what a denoising method should and should not do and we accordingly define criteria to characterise the performance. We propose a comprehensive set of evaluations, including i) benefits in improving signal quality and reducing uncertainty, ii) gains in reducing biases and the noise floor, iii) preservation of spatial resolution, iv) agreement of denoised data against a gold standard, v) efficacy in enabling noise-prone applications, such as ultra-high resolution imaging. We further provide newly acquired complex datasets (magnitude and phase) with multiple repeats that sample different SNR regimes to highlight performance differences under different scenarios. Without loss of generality, we subsequently apply a number of exemplar patch-based denoising algorithms to these datasets, including Non-Local Means, Marchenko-Pastur PCA (MPPCA) in the magnitude and complex domain and NORDIC, and compare them with respect to the above criteria and against a gold standard complex average of multiple repeats. We demonstrate that all tested denoising approaches reduce noise-related variance, but not always biases from the elevated noise floor. They all induce a spatial resolution penalty, but its extent can vary depending on the method and the implementation. Some denoising approaches agree with the gold standard more than others and we demonstrate challenges in even defining such a standard. Overall, we show that dMRI denoising performed in the complex domain is advantageous to magnitude domain denoising with respect to all the above criteria.

show abstract

Section: Denoising As Filtering or Deterministic Prediction?mentioning

confidence: 99%

Section: Limitations and Future Workmentioning

confidence: 99%

Denoising Diffusion MRI: Considerations and implications for analysis

Manzano-Patron

Moeller

Andersson

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…An additional potential application of AI is denoising of diffusion MRI or resting-state functional MRI (fMRI) data [ 6 ]. These modalities are widely used for studying brain connectivity and the networks that are targeted in DBS and other neuromodulation techniques.…”

Section: Ai and Dbs Target Identificationmentioning

confidence: 99%

AI and deep brain stimulation: what have we learned?

Limousin

Akram

2023

Nat Rev Neurol

View full text Add to dashboard Cite

Oncecorrections are submitted, we cannot routinely make further changes to the article.2. Note that the eproof should be amended in only one browser window at any one time; otherwise changes will be overwritten.3. Author surnames have been highlighted. Please check these carefully and adjust if the first name or surname is marked up incorrectly. Note that changes here will affect indexing of your article in public repositories such as PubMed. Also, carefully check the spelling and numbering of all author names and affiliations, and the corresponding email address(es). 4. Au: In the sentence beginning ‘DBS conventionally involves…’, please check that the edits have not changed your meaning. 5. If ref. 3 (preprint) has now been published in final peer-reviewed form, please update the reference details if appropriate.

show abstract