 |
 |
Dear Richard I actually but the subject of my previous response was a reply to spm digest, sorry for that. Here comes my response once more: If you scan a gel phantom you will be convinced that there is low frequency drift coming from the scanner it self. This is due to a lot of things but probably mainly due to heating induced by mechanical vibrations. I would pick the cut-off frequency based on your particular scanner, (most recent scanners that I have seen have little non-white noise faster than the default 128s period used in SPM). Regarding aliasing of cardiac and respiratory noise it is in no way guaranteed to end up at low frequencies. In fact for typical heart and respiration rates and TR values around 2s the variation i e.g. heart rate is so large that there will be cardiac noise at all frequencies, not just the low ones. you can read more here: Foerster BU, Tomasi D, Caparelli EC. Magnetic field shift due to mechanical vibration in functional magnetic resonance imaging. Magn Reson Med. 2005 Nov;54(5):1261-7. PubMed PMID: 16215962; PubMed Central PMCID: PMC2408718. Lund TE, Madsen KH, Sidaros K, Luo WL, Nichols TE. Non-white noise in fMRI: does modelling have an impact? Neuroimage. 2006 Jan 1;29(1):54-66. Epub 2005 Aug 11. PubMed PMID: 16099175. Smith AM, Lewis BK, Ruttimann UE, Ye FQ, Sinnwell TM, Yang Y, Duyn JH, Frank JA. Investigation of low frequency drift in fMRI signal. Neuroimage. 1999 May;9(5):526-33. PubMed PMID: 10329292. https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=SPM;651c37f7.00 Best Torben Torben Ellegaard Lund Associate Professor, PhD The Danish National Research Foundation's Center of Functionally Integrative Neuroscience (CFIN) Aarhus University Aarhus University Hospital Building 10G, 5th floor, room 31 Noerrebrogade 44 8000 Aarhus C Denmark Phone: +4589494380 Fax: +4589494400 http://www.cfin.au.dk [log in to unmask] Den Uge:9 04/03/2011 kl. 17.56 skrev Richard Binney: > Dear all, > > Something has recently been said to me that has made me think, and I want to get my head around this a bit more..... > > At what point does high pass filtering becomes ineffective? It has not been a concern for me in the past, but recently it has been suggested to me that a rule of thumb is to set the high pass filter at a minimum of 1.5 x your longest SOA. That seems reasonable if this is equal to or less than 128secs. I say that because I beleive that this is the optimum filter setting to remove low-freq noise associated with cardiac/respiratory noise, etc. But what if you use a filter at a lower frequency (e.g., 260secs)? It has been said to me that this would be OK. But aren't you running a risk of alliasing signal of interest at frequencies lower than 0.01Hz with cardiac/respiratory noise etc? IF this is an acceptable risk with a filter of say 260secs, at what point does it becomes unacceptable and the filter becomes ineffective? Eg., Say you had an enomrous longest SOA of 800 secs, you might want to use a filter set at 1000secs (0.001Hz) just to be sure. would this filter be effectively redundant? > > What other factors might speak to this? Jittered SOAs for example? Using a range of SOAs would of course spread the signal of interest across frequencies increasing sensitivity. Does this have an effect of reducing the risk of alliasing with low-freq noise or are you still losing a significant proprotion of your signal in the <0.1Hz frequencies (either due to noise or a 128 sec filter)? In what manner should this inform your high-pass filter? Should you be concerned with the longest SOA or the mean SOA (fundamental frequency) of your signal? > > Thanks in advance for your help, > > Richard