Thanks. How to extract the beta value of a ROI?
Jiansong
> From: Robert Welsh <[log in to unmask]>
> Reply-To: Robert Welsh <[log in to unmask]>
> Date: Tue, 19 Oct 2004 18:07:45 -0400
> To: [log in to unmask]
> Subject: Re: [SPM] % signal change
>
> They are the beta images you'll find in the directory that contains your
> results.
>
> -+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
> Robert C. Welsh, PhD
> Research Investigator
> Department of Radiology
> University of Michigan
> (734) - 764 - 2412 (fax)
> [log in to unmask]
>
>
>>>> Jiansong Xu <[log in to unmask]> 10/19/04 9:41 AM >>>
> How to get these beta(task) and beta(constant term)?
>
> Thanks
>
> Jiansong
>
>
>
>> From: Will Penny <[log in to unmask]>
>> Reply-To: Will Penny <[log in to unmask]>
>> Date: Mon, 18 Oct 2004 17:17:54 +0100
>> To: [log in to unmask]
>> Subject: Re: [SPM] % signal change
>>
>> Simo Vanni wrote:
>>
>>> Dear Will,
>>>
>>> I wonder if you got my last e-mail to SPM list about % signal change
> - I
>>> did not see it elsewhere than in the JISCMAIL web site. Sorry to
> bother
>>> you personally, but we are in a process of replying to referee's
>>> comments and have to proceed with this issue. My question was that
> when
>>> we do the :
>>>
>>> percentsignalchange = (beta(task)-beta(constant term))/beta(constant
> term)
>>>
>>
>>
>> I think I might've got it wrong there, sorry, ... see below
>>
>>
>>> we get negative numbers due to high beta(constant term). Would it be
>>> correct to do simply:
>>>
>>> beta(task)/beta(constant term)
>>>
>>> to get the % signal change.
>>>
>>
>>> In addition, would it be better to call this % model signal change.
>>>
>>
>>
>>
>> Commonly people report just beta(task) as the percent signal change
>> associated with that task. This is because the data has usually been
>> normalised by multiplying every voxel in every scan by '100/g' where g
> is the
>> average value over all time points and scans in that session. This
> happens
>> automatically in an SPM analysis.
>> This means that time series should, on average, have a value of 100.
>> So, beta(task) is then naturally in units of percent of 'global'
> signal, g.
>>
>> If you wish to report in units of percent of 'local' signal, then
> because the
>> average local signal is simply beta(constant term) then
>>
>> % local signal = beta(task) x 100 / beta(constant term)
>>
>> So if the % global signal = 0.9 (ie. beta(task)=0.9) and
> beta(constant term)
>> equals 150 at that voxel then the % local signal = 0.6.
>>
>> Hope this make sense,
>>
>> Will.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>>
>>>
>>> Best wishes,
>>>
>>> Simo Vanni
>>>
>>
>>
>> --
>> William D. Penny
>> Wellcome Department of Imaging Neuroscience
>> University College London
>> 12 Queen Square
>> London WC1N 3BG
>>
>> Tel: 020 7833 7475
>> FAX: 020 7813 1420
>> Email: [log in to unmask]
>> URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|