From trisanna.sprung-much at mail.mcgill.ca Mon Mar 20 11:53:21 2017 From: trisanna.sprung-much at mail.mcgill.ca (Trisanna Sprung-Much) Date: Mon, 20 Mar 2017 11:53:21 -0400 Subject: [MINC-users] computing similarity across images Message-ID: Hi there Is there a tool that can be used to confirm that 2 images are in fact images of the same patient's brain? For instance computing some sort of similarity index based on 2 T1 images or outputing a probability for them being images of the same brain? Thanks! Trisanna -- Ph.D. Candidate McGill University Integrated Program in Neuroscience Psychology From vladimir.fonov at gmail.com Mon Mar 20 13:12:42 2017 From: vladimir.fonov at gmail.com (Vladimir S. Fonov) Date: Mon, 20 Mar 2017 13:12:42 -0400 Subject: [MINC-users] computing similarity across images In-Reply-To: References: Message-ID: <37b6eab8-6cdc-5025-cc82-ec700240ed41@gmail.com> Hello, in general this is not a trivial task. But from the practical point of view - did you try to use the contents of "patient:full_name" field of the image header? On 2017-03-20 11:53 AM, Trisanna Sprung-Much wrote: > Hi there > > Is there a tool that can be used to confirm that 2 images are in fact > images of the same patient's brain? For instance computing some sort of > similarity index based on 2 T1 images or outputing a probability for them > being images of the same brain? > > Thanks! > > Trisanna > > -- > Ph.D. Candidate > McGill University > Integrated Program in Neuroscience > Psychology > _______________________________________________ > MINC-users at bic.mni.mcgill.ca > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > From zijdenbos at gmail.com Mon Mar 20 13:22:31 2017 From: zijdenbos at gmail.com (Alex Zijdenbos) Date: Mon, 20 Mar 2017 13:22:31 -0400 Subject: [MINC-users] computing similarity across images In-Reply-To: <37b6eab8-6cdc-5025-cc82-ec700240ed41@gmail.com> References: <37b6eab8-6cdc-5025-cc82-ec700240ed41@gmail.com> Message-ID: Hi Trisanna, I assume you are talking about cases when the patient name/id may have been removed or is unreliable? I have done this in the past by performing a direct linear registration between two scans, and then calculating a similarity metric (e.g., xcorr) between them (using minccmp for example). Assuming there has not been too much change in the individual's brain between the scans, this can get you at least an indication that might be useful; but as Vladimir said, it's in general not trivial - and a measure like this will likely give you false positives. -- A On Mon, Mar 20, 2017 at 1:12 PM, Vladimir S. Fonov wrote: > Hello, > > in general this is not a trivial task. > > But from the practical point of view - did you try to use the contents of > "patient:full_name" field of the image header? > > > > On 2017-03-20 11:53 AM, Trisanna Sprung-Much wrote: > >> Hi there >> >> Is there a tool that can be used to confirm that 2 images are in fact >> images of the same patient's brain? For instance computing some sort of >> similarity index based on 2 T1 images or outputing a probability for them >> being images of the same brain? >> >> Thanks! >> >> Trisanna >> >> -- >> Ph.D. Candidate >> McGill University >> Integrated Program in Neuroscience >> Psychology >> _______________________________________________ >> MINC-users at bic.mni.mcgill.ca >> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users >> >> > _______________________________________________ > MINC-users at bic.mni.mcgill.ca > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > > From claude at bic.mni.mcgill.ca Mon Mar 20 13:28:45 2017 From: claude at bic.mni.mcgill.ca (Claude LEPAGE) Date: Mon, 20 Mar 2017 13:28:45 -0400 Subject: [MINC-users] computing similarity across images In-Reply-To: Message-ID: <201703201728.v2KHSjiI009695@login1.bic.mni.mcgill.ca> Hi, Do you have only a few scans or hundreds of them? If only a few scans, linear registration with visual inspection will do the trick. Claude > > Hi Trisanna, > > I assume you are talking about cases when the patient name/id may have been > removed or is unreliable? > > I have done this in the past by performing a direct linear registration > between two scans, and then calculating a similarity metric (e.g., xcorr) > between them (using minccmp for example). Assuming there has not been too > much change in the individual's brain between the scans, this can get you > at least an indication that might be useful; but as Vladimir said, it's in > general not trivial - and a measure like this will likely give you false > positives. > > -- A > > On Mon, Mar 20, 2017 at 1:12 PM, Vladimir S. Fonov > wrote: > > > Hello, > > > > in general this is not a trivial task. > > > > But from the practical point of view - did you try to use the contents of > > "patient:full_name" field of the image header? > > > > > > > > On 2017-03-20 11:53 AM, Trisanna Sprung-Much wrote: > > > >> Hi there > >> > >> Is there a tool that can be used to confirm that 2 images are in fact > >> images of the same patient's brain? For instance computing some sort of > >> similarity index based on 2 T1 images or outputing a probability for them > >> being images of the same brain? > >> > >> Thanks! > >> > >> Trisanna > >> > >> -- > >> Ph.D. Candidate > >> McGill University > >> Integrated Program in Neuroscience > >> Psychology > >> _______________________________________________ > >> MINC-users at bic.mni.mcgill.ca > >> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > >> > >> > > _______________________________________________ > > MINC-users at bic.mni.mcgill.ca > > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > > > > > _______________________________________________ > MINC-users at bic.mni.mcgill.ca > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > From trisanna.sprung-much at mail.mcgill.ca Mon Mar 20 13:46:45 2017 From: trisanna.sprung-much at mail.mcgill.ca (Trisanna Sprung-Much) Date: Mon, 20 Mar 2017 13:46:45 -0400 Subject: [MINC-users] computing similarity across images In-Reply-To: <37b6eab8-6cdc-5025-cc82-ec700240ed41@gmail.com> References: <37b6eab8-6cdc-5025-cc82-ec700240ed41@gmail.com> Message-ID: thanks Vlad - I was referring to a situation where all patient info has been removed! Sorry for the confusion. -- Ph.D. Candidate McGill University Integrated Program in Neuroscience Psychology On Mon, Mar 20, 2017 at 1:12 PM, Vladimir S. Fonov wrote: > Hello, > > in general this is not a trivial task. > > But from the practical point of view - did you try to use the contents of > "patient:full_name" field of the image header? > > > > On 2017-03-20 11:53 AM, Trisanna Sprung-Much wrote: > >> Hi there >> >> Is there a tool that can be used to confirm that 2 images are in fact >> images of the same patient's brain? For instance computing some sort of >> similarity index based on 2 T1 images or outputing a probability for them >> being images of the same brain? >> >> Thanks! >> >> Trisanna >> >> -- >> Ph.D. Candidate >> McGill University >> Integrated Program in Neuroscience >> Psychology >> _______________________________________________ >> MINC-users at bic.mni.mcgill.ca >> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users >> >> > _______________________________________________ > MINC-users at bic.mni.mcgill.ca > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users > From jp at biospective.com Wed Mar 29 11:36:00 2017 From: jp at biospective.com (Jean-Philippe Coutu) Date: Wed, 29 Mar 2017 11:36:00 -0400 Subject: [MINC-users] Dimension data type Message-ID: Dear MINC users, In the minc header, how does the data type of the dimension matter? For instance, one could start with a 2D slice with this for its yspace: int yspace ; yspace:length = 1 ; yspace:varid = "MINC standard variable" ; yspace:vartype = "dimension____" ; yspace:version = "MINC Version 1.0" ; yspace:comments = "Y increases from patient posterior to anterior" ; yspace:spacing = "regular__" ; yspace:alignment = "centre" ; yspace:start = -4.1 ; yspace:direction_cosines = 0., 1., 0. ; yspace:units = "mm" ; yspace:spacetype = "native____" ; yspace:step = 0.1 ; (both xspace and zspace are also int in this case, but with length >1) Now if one attempts to replicate this slice in the yspace dimension with this command: mincconcat -verbose -concat_dimension yspace -filestarts -0.1,0,0.1 slice.mnc slice.mnc slice.mnc out.mnc The mincheader of out.mnc will now display this: double yspace(yspace) ; yspace:length = 3 ; yspace:dimorder = "yspace" ; yspace:varid = "MINC standard variable" ; yspace:vartype = "dimension____" ; yspace:version = "MINC Version 1.0" ; yspace:comments = "Y increases from patient posterior to anterior" ; yspace:alignment = "centre" ; yspace:direction_cosines = 0., 1., 0. ; yspace:units = "mm" ; yspace:spacetype = "native____" ; yspace:spacing = "regular__" ; yspace:start = -4.2 ; yspace:step = 0.0999999999999999 ; So it went from int to double (both xspace and zspace remain int), and in addition, something weird happened where yspace became yspace(yspace). Now if one was to run something like mincreshape -normalize on either the slice.mnc or out.mnc, all xspace, yspace and zspace will now become double rather than int. Is this important for any minc operation? Can it be 'normalized' back to int somehow? Thanks, JP From bert at phalarope.com Wed Mar 29 23:19:33 2017 From: bert at phalarope.com (Robert D. Vincent) Date: Wed, 29 Mar 2017 23:19:33 -0400 Subject: [MINC-users] Dimension data type In-Reply-To: References: Message-ID: Hi JP, The long answer is that the type of the dimension variables does not usually matter, because for a regularly sampled dimension, the dimension variable's dataset is not used - only its attributes are meaningful. Many of the MINC tools just create a scalar integer dataset for regular dimensions. The notation yspace(yspace) indicates that the variable "yspace" is a vector indexed by the dimension "yspace", which is what is expected for an irregularly sampled dimension. In these dimensions the dimension dataset stores a vector of sampling points. It has a floating-point type to allow for real-valued sample points. However, it sometimes happens that a regularly sampled dimension still has this vector dataset stored - if you look at its values, it will probably just be a vector of regular, monotonic values, something like : -4.2, -4.1, -4.0, etc. in your case. It looks as though mincconcat is storing this vector even though it is not needed. So the short answer is: everything is probably working fine. -bert On Wed, Mar 29, 2017 at 11:36 AM, Jean-Philippe Coutu wrote: > Dear MINC users, > > In the minc header, how does the data type of the dimension matter? > > For instance, one could start with a 2D slice with this for its yspace: > > int yspace ; > yspace:length = 1 ; > yspace:varid = "MINC standard variable" ; > yspace:vartype = "dimension____" ; > yspace:version = "MINC Version 1.0" ; > yspace:comments = "Y increases from patient posterior to anterior" ; > yspace:spacing = "regular__" ; > yspace:alignment = "centre" ; > yspace:start = -4.1 ; > yspace:direction_cosines = 0., 1., 0. ; > yspace:units = "mm" ; > yspace:spacetype = "native____" ; > yspace:step = 0.1 ; > > (both xspace and zspace are also int in this case, but with length >1) > > Now if one attempts to replicate this slice in the yspace dimension with > this command: > > mincconcat -verbose -concat_dimension yspace -filestarts -0.1,0,0.1 > slice.mnc slice.mnc slice.mnc out.mnc > > The mincheader of out.mnc will now display this: > > double yspace(yspace) ; > yspace:length = 3 ; > yspace:dimorder = "yspace" ; > yspace:varid = "MINC standard variable" ; > yspace:vartype = "dimension____" ; > yspace:version = "MINC Version 1.0" ; > yspace:comments = "Y increases from patient posterior to anterior" ; > yspace:alignment = "centre" ; > yspace:direction_cosines = 0., 1., 0. ; > yspace:units = "mm" ; > yspace:spacetype = "native____" ; > yspace:spacing = "regular__" ; > yspace:start = -4.2 ; > yspace:step = 0.0999999999999999 ; > > So it went from int to double (both xspace and zspace remain int), and in > addition, something weird happened where yspace became yspace(yspace). > > Now if one was to run something like mincreshape -normalize on either the > slice.mnc or out.mnc, all xspace, yspace and zspace will now become double > rather than int. > > Is this important for any minc operation? Can it be 'normalized' back to > int somehow? > > Thanks, > JP > _______________________________________________ > MINC-users at bic.mni.mcgill.ca > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users >