diff --git a/content/08_ICA_Based_Denoising.md b/content/08_ICA_Based_Denoising.md index 31ebb32..0464204 100644 --- a/content/08_ICA_Based_Denoising.md +++ b/content/08_ICA_Based_Denoising.md @@ -14,7 +14,7 @@ kernelspec: # Denoising Data with ICA ICA classification methods like `tedana` will produce two important outputs: component time series and component classifications. -The component classifications will indicate whether each componet is "good" (accepted) or "bad" (rejected). +The component classifications will indicate whether each component is "good" (accepted) or "bad" (rejected). To remove noise from your data, you can regress the "bad" components out of it, though there are multiple methods to accomplish this. Let's start by loading the necessary data. diff --git a/content/Acquiring_Multi_Echo_Data.md b/content/Acquiring_Multi_Echo_Data.md index 9d533db..282f4c2 100644 --- a/content/Acquiring_Multi_Echo_Data.md +++ b/content/Acquiring_Multi_Echo_Data.md @@ -87,8 +87,7 @@ used as an input to many functions in the tedana workflow. There are many ways to calculate T2* maps, with some using multi-echo acquisitions. We are not presenting an expansive review of this literature here, -but [Cohen-Adad et al. (2012)](https://doi.org/10.1016/j.neuroimage.2012.01.053) -and [Ruuth et al. (2019)](https://doi.org/10.1016/j.ejro.2018.12.006) are good places to start +but {cite:t}`cohen2012t2` and {cite:t}`ruuth2019comparison` are good places to start learning more about this topic. ## Acquisition parameter recommendations @@ -107,8 +106,8 @@ TE one would use for single-echo T2* weighted fMRI. ```{note} This is in contrast to the **dual echo** denoising method which uses a very early (~5ms) -first echo in order to clean data. For more information on this method, see -[Bright and Murphy (2013)](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3518782/). +first echo in order to clean data. +For more information on this method, see {cite:t}`bright2013removing`. ``` More than 3 echoes may be useful, because that would allow for more accurate @@ -124,7 +123,7 @@ We suggest new multi-echo fMRI users examine the :ref:`spreadsheet of publicatio multi-echo fMRI to identify studies with similar acquisition priorities, and use the parameters from those studies as a starting point. More complete recommendations and guidelines are discussed in the -[appendix of Dipasquale et al. (2017)](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0173289). +appendix of {cite:t}`dipasquale2017comparing`. ```{note} In order to increase the number of contrasts ("echoes") you may need to first increase the TR, shorten the @@ -132,6 +131,29 @@ first TE and/or enable in-plane acceleration. For typically used parameters see the **ME-fMRI parameters** section below. ``` + +## Additional considerations + + +### Complex reconstruction + +It is possible to retain phase data when reconstructing multi-echo fMRI data. +The phase data may be leveraged for a number of useful denoising and processing methods, +including NORDIC {cite:p}`vizioli2021lowering;dowdle2021nordic;dowdle2023evaluating` +and MEDIC dynamic distortion correction {cite:p}`van2023framewise`. + +It's important to remember that retaining phase data for each echo will effectively double the +amount of data you end up with. +This can also cause problems with online reconstruction, +for example with Siemens machines running XA30. + + +### No-excitation-pulse noise volumes + +In order to best use NORDIC, +researchers should acquire no-RF noise volumes at the end of their fMRI runs. + + ## ME-fMRI parameters The following section highlights a selection of parameters collected from published papers that have diff --git a/content/MR_Physics.md b/content/MR_Physics.md index 2886fb8..86d9a42 100644 --- a/content/MR_Physics.md +++ b/content/MR_Physics.md @@ -1,69 +1,66 @@ # MR Physics +Magnetic Resonance Imaging (MRI), +also known as Magnetic Resonance Tomography or Nuclear Magnetic Resonance Imaging, +is one of the non-invasive imaging techniques that have superior soft tissue contrasts and potential physiological and functional applications. +This type of radiation has not enough energy to remove an electron from an atom but just to excite it to a higher energy state. +Since the 1980s, MRI has been a mainstay of non-invasive diagnostic radiology because it does not expose the body to radiation. +It is frequently used in neuroimaging for the diagnosis and monitoring of diseases, +and it has not yet shown any adverse effects from exposure, +which is a major benefit over other imaging modalities. +MRI enables to perform dynamic studies due to its speed of acquisition.{cite:p}`hovet2018mri;berger2002does` -Magnetic Resonance Imaging (MRI), also known as Magnetic Resonance Tomography or Nuclear Magnetic Resonance Imaging, -is one of the non-invasive imaging techniques that have superior soft tissue contrasts and potential physiological -and functional applications. This type of radiation has not enough energy to remove an electron from an atom but -just to excite it to a higher energy state. Since the 1980s, MRI has been a mainstay of non-invasive diagnostic radiology -because it does not expose the body to radiation. It is frequently used in neuroimaging for the diagnosis and monitoring -of diseases, and it has not yet shown any adverse effects from exposure, which is a major benefit over other imaging -modalities. MRI enables to perform dynamic studies due to it's speed of acquisition. [1][2] ## Basic Physics -Any atomic nucleous with an odd numer of nucleons has spin different from zero and so, a magnetic moment (magnetic dipole). +Any atomic nucleus with an odd number of neutrons has spin different from zero and so, +a magnetic moment (magnetic dipole). In the body, we can find several atoms with magnetic moment such as H, P, C, F, Na, which are sensitive to magnetic resonance. -Around 60% of the human body is made up of water that contains hydrogen, which is also present in proteins and lipids. +Around 60% of the human body is made up of water that contains hydrogen, +which is also present in proteins and lipids. For this reason, hydrogen is the most widely used in MRI. - -MRI bore contains a powerful magnet which generates an uniform magnetic field B0. Patiens are introduced in this magnetic field -and hydrogen atoms align to the magnetic field. According to Larmour's law, a magnetic dipole inside a magnetic field -precesses (spins) arround the magnetic field with a frequency proportional to the magnetic field strength. Hence, hydrogen atoms -precess arround the magnetic field generated by the MR with a frequency (Larmour frequency) that follows the equation: +MRI bore contains a powerful magnet which generates an uniform magnetic field B0. +Patiens are introduced in this magnetic field and hydrogen atoms align to the magnetic field. +According to Larmour's law, a magnetic dipole inside a magnetic field +precesses (spins) around the magnetic field with a frequency proportional to the magnetic field strength. +Hence, hydrogen atoms precess around the magnetic field generated by the MR with a frequency (Larmor frequency) that follows the equation: w = γ B0 ![](https://www.frontiersin.org/files/Articles/427144/frym-07-00023-HTML-r2/image_m/figure-2.jpg) -This precession can be parallel or antiparallel to B0. In the body the number of atoms that precess parallel is different to -the ones that precess antiparallel producing an small magnetic field which is proportional to B0 and also depends on the density -of hydrogen nuclei. So, the static magnetic field (B0) induces a slight magnetization of tissues. - - -Then, a radiofrequency pulse is emitted perpendicular to B0 with the same frequency that the spin precession frequency.Hydrogen atoms -abrosrb energy and spin out of equilibrium. Longitudinal magnetization (Mz) of protons in a parallel direction to B0 decreases, and a -transverse magnetization (Mx, My) appears. - -Then, when the RF dissapears, the magnetic momentum gradually goes back to te minimum -energy position (magnetic relaxation) while releasing energy. This emited signals are measured into the k-space which is an array -of numbers representing spatial frequencies in the MR image. (Each k-space point contains spatial frequency and phase information -about every pixel in the final image). Fourier transforme is performed to the k-space to obtain the final image. By varying the -sequence of RF pulses applied & collected, different types of images are created. - - -### MRI Sequences - -It's important to understand the meaning of **repetition time (TR)** and **echo time (TE)** in order to comprehend the main -MRI sequences. Time to Echo (TE) is the time between the delivery of the RF pulse and the receipt of the echo signal and +This precession can be parallel or antiparallel to B0. +In the body the number of atoms that precess parallel is different to +the ones that precess antiparallel producing an small magnetic field which is proportional to B0 and also depends on the density +of hydrogen nuclei. +So, the static magnetic field (B0) induces a slight magnetization of tissues. + +Then, a radiofrequency pulse is emitted perpendicular to B0 with the same frequency that the spin precession frequency. +Hydrogen atoms abrorb energy and spin out of equilibrium. +Longitudinal magnetization (Mz) of protons in a parallel direction to B0 decreases, +and a transverse magnetization (Mx, My) appears. + +Then, when the RF disappears, +the magnetic momentum gradually goes back to te minimum energy position (magnetic relaxation) while releasing energy. +These emitted signals are measured into the k-space, +which is an array of numbers representing spatial frequencies in the MR image. +(Each k-space point contains spatial frequency and phase information about every pixel in the final image). +Fourier transforme is performed to the k-space to obtain the final image. +By varying the sequence of RF pulses applied & collected, different types of images are created. + + +### MRI Sequences + +It's important to understand the meaning of **repetition time (TR)** and **echo time (TE)** in order to comprehend the main MRI sequences. +Time to Echo (TE) is the time between the delivery of the RF pulse and the receipt of the echo signal and the interval between subsequent pulse sequences delivered to the same slice is known as the repetition time (TR). -The most common sequences are T1-weighted and T2-weighted images. In neuroimaging, **T1-weighted** images are commonly used in anatomical -related studies, they are based on the study of the relaxation of the nuclei in the longitudinal component (Mz) of the magnetization -vector and are produced with short TR and TE.**T2-weighted** images are produced with longer TR and TE. They are based on study of the -variations of the component on the transverse plane of the magnetization during the relaxation, known as transverse relaxation (Mxy). - -There are many sequences that can be used depending on the objective. T - -## Multi-echo - - -## Bibliography - -[MRI-powered biomedical devices](https://doi.org/10.1080/13645706.2017.1402188) - -[Magnetic resonance imaging](https://doi.org/10.1136/bmj.324.7328.35 ) - -[nibib](https://www.nibib.nih.gov/science-education/science-topics/magnetic-resonance-imaging-mri) - - +The most common sequences are T1-weighted and T2-weighted images. +In neuroimaging, **T1-weighted** images are commonly used in anatomical related studies, +they are based on the study of the relaxation of the nuclei in the longitudinal component (Mz) of the magnetization +vector and are produced with short TR and TE. +**T2-weighted** images are produced with longer TR and TE. +They are based on study of the variations of the component on the transverse plane of the magnetization during the relaxation, +known as transverse relaxation (Mxy). +There are many sequences that can be used depending on the objective. diff --git a/content/Processing_Multi_Echo_Data.md b/content/Processing_Multi_Echo_Data.md index b6d8a78..c033e8b 100644 --- a/content/Processing_Multi_Echo_Data.md +++ b/content/Processing_Multi_Echo_Data.md @@ -56,18 +56,19 @@ Therefore, we suggest using the same slice timing for all echoes in an ME-EPI se ## 3. Perform distortion correction, spatial normalization, smoothing, and any rescaling or filtering **after** denoising Any step that will alter the relationship of signal magnitudes between echoes should occur after denoising and combining -of the echoes. For example, if echo is separately scaled by its mean signal over time, then resulting intensity gradients -and the subsequent calculation of voxelwise T2* values will be distorted or incorrect. An aggressive temporal filter -(e.g., a 0.1Hz low pass filter) or spatial smoothing could similarly distort the relationship between the echoes at each -time point. +of the echoes. +For example, if echo is separately scaled by its mean signal over time, +then resulting intensity gradients and the subsequent calculation of voxelwise T2* values will be distorted or incorrect. +An aggressive temporal filter (e.g., a 0.1Hz low pass filter) +or spatial smoothing could similarly distort the relationship between the echoes at each time point. ```{note} -We are assuming that spatial normalization and distortion correction, particularly non-linear normalization methods -with higher order interpolation functions, are likely to distort the relationship between echoes while rigid body -motion correction would linearly alter each echo in a similar manner. This assumption has not yet been empirically -tested and an affine normalzation with bilinear interpolation may not distort the relationship between echoes. +We are assuming that spatial normalization and distortion correction, +particularly non-linear normalization methods with higher order interpolation functions, +are likely to distort the relationship between echoes while rigid body motion correction would linearly alter each echo in a similar manner. +This assumption has not yet been empirically tested and an affine normalzation with bilinear interpolation may not distort the relationship between echoes. Additionally, there are benefits to applying only one spatial transform to data rather than applying one spatial -transform for motion correction and a later transform for normalization and distortion correction. Our advice -against doing normalization and distortion correction is a conservative choice and we encourage additional +transform for motion correction and a later transform for normalization and distortion correction. +Our advice against doing normalization and distortion correction is a conservative choice and we encourage additional research to better understand how these steps can be applied before denoising. ``` diff --git a/content/Recommended_Reading.md b/content/Recommended_Reading.md index c627d9c..cf2a178 100644 --- a/content/Recommended_Reading.md +++ b/content/Recommended_Reading.md @@ -18,4 +18,4 @@ kernelspec: {cite:t}`POSSE2012665` includes an historical overview of multi-echo acquisition and research. {cite:t}`KUNDU201759` is a review of multi-echo denoising with a focus on the MEICA algorithm. The appendix of {cite:t}`OLAFSSON201543` includes a good explanation of the math underlying MEICA denoising. -The appendix of {cite:t}`10.1371/journal.pone.0173289` includes some recommendations for multi-echo acquisition. +The appendix of {cite:t}`dipasquale2017comparing` includes some recommendations for multi-echo acquisition. diff --git a/content/references.bib b/content/references.bib index c55013b..8e84651 100644 --- a/content/references.bib +++ b/content/references.bib @@ -1,6 +1,3 @@ ---- ---- - @article{power2018ridding, title={Ridding fMRI data of motion-related influences: Removal of signals with distinct spatial and physical bases in multiecho data}, author={Power, Jonathan D and Plitt, Mark and Gotts, Stephen J and Kundu, Prantik and Voon, Valerie and Bandettini, Peter A and Martin, Alex}, @@ -107,21 +104,6 @@ @article{OLAFSSON201543 abstract = {The recent introduction of simultaneous multi-slice (SMS) acquisitions has enabled the acquisition of blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI) data with significantly higher temporal sampling rates. In a parallel development, the use of multi-echo fMRI acquisitions in conjunction with a multi-echo independent component analysis (ME-ICA) approach has been introduced as a means to automatically distinguish functionally-related BOLD signal components from signal artifacts, with significant gains in sensitivity, statistical power, and specificity. In this work, we examine the gains that can be achieved with a combined approach in which data obtained with a multi-echo simultaneous multi-slice (MESMS) acquisition are analyzed with ME-ICA. We find that ME-ICA identifies significantly more BOLD-like components in the MESMS data as compared to data acquired with a conventional multi-echo single-slice acquisition. We demonstrate that the improved performance of MESMS derives from both an increase in the number of temporal samples and the enhanced ability to filter out high-frequency artifacts.} } -@article{10.1371/journal.pone.0173289, - doi = {10.1371/journal.pone.0173289}, - author = {Dipasquale, Ottavia AND Sethi, Arjun AND Laganà, Maria Marcella AND Baglio, Francesca AND Baselli, Giuseppe AND Kundu, Prantik AND Harrison, Neil A. AND Cercignani, Mara}, - journal = {PLOS ONE}, - publisher = {Public Library of Science}, - title = {Comparing resting state fMRI de-noising approaches using multi- and single-echo acquisitions}, - year = {2017}, - month = {03}, - volume = {12}, - url = {https://doi.org/10.1371/journal.pone.0173289}, - pages = {1-25}, - abstract = {Artifact removal in resting state fMRI (rfMRI) data remains a serious challenge, with even subtle head motion undermining reliability and reproducibility. Here we compared some of the most popular single-echo de-noising methods—regression of Motion parameters, White matter and Cerebrospinal fluid signals (MWC method), FMRIB’s ICA-based X-noiseifier (FIX) and ICA-based Automatic Removal Of Motion Artifacts (ICA-AROMA)—with a multi-echo approach (ME-ICA) that exploits the linear dependency of BOLD on the echo time. Data were acquired using a clinical scanner and included 30 young, healthy participants (minimal head motion) and 30 Attention Deficit Hyperactivity Disorder patients (greater head motion). De-noising effectiveness was assessed in terms of data quality after each cleanup procedure, ability to uncouple BOLD signal and motion and preservation of default mode network (DMN) functional connectivity. Most cleaning methods showed a positive impact on data quality. However, based on the investigated metrics, ME-ICA was the most robust. It minimized the impact of motion on FC even for high motion participants and preserved DMN functional connectivity structure. The high-quality results obtained using ME-ICA suggest that using a multi-echo EPI sequence, reliable rfMRI data can be obtained in a clinical setting.}, - number = {3}, -} - @ARTICLE{Kundu2013-po, title = "Integrated strategy for improving functional connectivity mapping using multiecho {fMRI}", @@ -196,3 +178,105 @@ @article{spreng2022neurocognitive doi={10.1038/s41597-022-01231-7}, abstract={Central to understanding human behavior is a comprehensive mapping of brain-behavior relations within the context of lifespan development. Reproducible discoveries depend upon well-powered samples of reliable data. We provide to the scientific community two, 10-minute, multi-echo functional MRI (ME-fMRI) runs, and structural MRI (T1-MPRAGE), from 181 healthy younger (ages 18–34 y) and 120 older adults (ages 60–89 y). T2-FLAIR MRIs and behavioral assessments are available in a majority subset of over 250 participants. Behavioral assessments include fluid and crystallized cognition, self-reported measures of personality, and socioemotional functioning. Initial quality control and validation of these data is provided. This dataset will be of value to scientists interested in BOLD signal specifically isolated from ME-fMRI, individual differences in brain-behavioral associations, and cross-sectional aging effects in healthy adults. Demographic and behavioral data are available within the Open Science Framework project “Goal-Directed Cognition in Older and Younger Adults” (http://osf.io/yhzxe/), which will be augmented over time; neuroimaging data are available on OpenNeuro (https://openneuro.org/datasets/ds003592)} } + +@article{cohen2012t2, + title={T2* mapping and B0 orientation-dependence at 7 T reveal cyto-and myeloarchitecture organization of the human cortex}, + author={Cohen-Adad, Julien and Polimeni, Jonathan R and Helmer, Karl G and Benner, Thomas and McNab, Jennifer A and Wald, Lawrence L and Rosen, Bruce R and Mainero, Caterina}, + journal={Neuroimage}, + volume={60}, + number={2}, + pages={1006--1014}, + year={2012}, + publisher={Elsevier} +} + +@article{ruuth2019comparison, + title={Comparison of reconstruction and acquisition choices for quantitative T2* maps and synthetic contrasts}, + author={Ruuth, Riikka and Kuusela, Linda and M{\"a}kel{\"a}, Teemu and Melkas, Susanna and Korvenoja, Antti}, + journal={European journal of radiology open}, + volume={6}, + pages={42--48}, + year={2019}, + publisher={Elsevier} +} + +@article{bright2013removing, + title={Removing motion and physiological artifacts from intrinsic BOLD fluctuations using short echo data}, + author={Bright, Molly G and Murphy, Kevin}, + journal={Neuroimage}, + volume={64}, + pages={526--537}, + year={2013}, + publisher={Elsevier} +} + +@article{dipasquale2017comparing, + title={Comparing resting state fMRI de-noising approaches using multi-and single-echo acquisitions}, + author={Dipasquale, Ottavia and Sethi, Arjun and Lagan{\`a}, Maria Marcella and Baglio, Francesca and Baselli, Giuseppe and Kundu, Prantik and Harrison, Neil A and Cercignani, Mara}, + journal={PloS one}, + volume={12}, + number={3}, + pages={e0173289}, + year={2017}, + publisher={Public Library of Science San Francisco, CA USA} +} + +@article{hovet2018mri, + title={MRI-powered biomedical devices}, + author={Hovet, Sierra and Ren, Hongliang and Xu, Sheng and Wood, Bradford and Tokuda, Junichi and Tse, Zion Tsz Ho}, + journal={Minimally Invasive Therapy \& Allied Technologies}, + volume={27}, + number={4}, + pages={191--202}, + year={2018}, + publisher={Taylor \& Francis} +} + +@article{berger2002does, + title={How does it work?: Magnetic resonance imaging}, + author={Berger, Abi}, + journal={BMJ: British Medical Journal}, + volume={324}, + number={7328}, + pages={35}, + year={2002}, + publisher={BMJ Publishing Group} +} + +@article{vizioli2021lowering, + title={Lowering the thermal noise barrier in functional brain mapping with magnetic resonance imaging}, + author={Vizioli, Luca and Moeller, Steen and Dowdle, Logan and Ak{\c{c}}akaya, Mehmet and De Martino, Federico and Yacoub, Essa and U{\u{g}}urbil, Kamil}, + journal={Nature communications}, + volume={12}, + number={1}, + pages={5181}, + year={2021}, + publisher={Nature Publishing Group UK London} +} + +@article{dowdle2021nordic, + title={NORDIC Increases the Sensitivity and Preserves the Spatiotemporal Precision of fMRI Responses}, + author={Dowdle, Logan T and Vizioli, Luca and Moeller, Steen and Ak{\c{c}}akaya, Mehmet and Olman, Cheryl and Ghose, Geoffrey and Yacoub, Essa and U{\u{g}}urbil, K{\^a}mil}, + journal={bioRxiv}, + pages={2021--08}, + year={2021}, + publisher={Cold Spring Harbor Laboratory} +} + +@article{dowdle2023evaluating, + title={Evaluating increases in sensitivity from NORDIC for diverse fMRI acquisition strategies}, + author={Dowdle, Logan T and Vizioli, Luca and Moeller, Steen and Ak{\c{c}}akaya, Mehmet and Olman, Cheryl and Ghose, Geoffrey and Yacoub, Essa and U{\u{g}}urbil, K{\^a}mil}, + journal={NeuroImage}, + volume={270}, + pages={119949}, + year={2023}, + publisher={Elsevier} +} + +@article{van2023framewise, + title={Framewise multi-echo distortion correction for superior functional MRI}, + author={Van, Andrew N and Montez, David F and Laumann, Timothy O and Suljic, Vahdeta and Madison, Thomas and Baden, Noah J and Ramirez-Perez, Nadeshka and Scheidter, Kristen M and Monk, Julia S and Whiting, Forrest I and others}, + journal={Biorxiv}, + year={2023}, + publisher={Cold Spring Harbor Laboratory Preprints} +}