mincreshape - cuts a hyperslab out of a minc file (with dimension re-ordering)

mincreshape [<options>] <infile>.mnc <outfile>.mnc

Mincreshape's main job is to chop a hyperslab out of a minc file and put it into a new minc file. "What is a hyperslab?", you ask. It is simply a multi-dimensional box specified with a starting index (a vector giving a voxel coordinate) and a count vector (a number of voxels along each axis). A single slice out of a volume is a hyperslab (with a count of 1 in the slice direction), a small block pulled out of a large volume is a hyperslab, a single echo volume out of a multi-echo MRI dataset is a hyperslab, one time point out of a dynamic acquisition is a hyperslab - you get the idea. Check out the -start, -count and -dimrange options for more details on how to do this (and look at the examples!). If you are pulling out only one point along a dimension, you have the option of making the dimension disappear, so mincreshape gives you the ability to reduce the dimensionality of a minc file. As well, you aren't constrained to specify a hyperslab that is only within the input file, you can extend beyond the bounds of the dimensions in the input file, and furthermore you can give a count that will flip the data along a dimension.

As if all that is not enough mincreshape has the ability to re-order dimensions. The most obvious case is converting a transverse image into a coronal image. But you can type a list of dimension names to get an arbitrary order of dimensions.

You want more!?! Okay, okay. Mincreshape makes all of the minc library icv conversions available on the command line. For those who like things defined, an icv is an image conversion variable (don't ask me why I called it that) which basically lets you tell the data what it's going to look like. In other words, it does a bunch of conversions for you. These conversions include changing type, range and normalization of the pixel values, expanding or contracting images (by pixel duplication or averaging) to give a specified image size, and converting vector images to scalar.

Just so you don't get confused let me tell you clearly here: Mincreshape does all of the icv conversions first and then the hyperslab and dimension re-ordering stuff is applied to the result of that. So if you want to mix them together (like -imgsize, -start, -count), get it clear in your head first.

Okay, hold on to your seat: here's a list of options.

Note that options can be specified in abbreviated form (as long as they are unique) and can be given anywhere on the command line.

-clobber: Overwrite an existing file.

-noclobber: Don't overwrite an existing file (default).

-verbose: Print out progress information for each chunk of data copied (default). A chunk varies in size depending mostly on whether you're re-ordering dimensions or not and how big the internal buffer is allowed to be.

-quiet: Do not print out progress information.

-max_chunk_size_in_kb: Specify the maximum size of the copy buffer (in kbytes). Default is 4096 kbytes.

The default for type, sign and valid range is to use those of the input file. If type is specified, then both sign and valid range are set to the default for that type. If sign is specified, then valid range is set to the default for the type and sign.

-filetype: Don't do any type conversion (default).

-byte: Write out bytes values.

-short: Write out short integer values.

-int: Write out 32-bit integer values.

-long: Superseded by -int.

-float: Write out single-precision floating point values.

-double: Write out double-precision floating point values.

-signed: Write out values as signed integers (default for short and long). Ignored for floating point types.

-unsigned: Write out values as unsigned integers (default for byte). Ignored for floating point types.

-valid_range <min> <max>: specifies the valid range of output voxel values in their integer representation. Default is the full range for the type and sign. This option is ignored for floating point values.

-image_range: <min> <max> Normalize images to a given minimum and maximum real value (not voxel value).

-normalize: Normalize images to real minimum and maximum for the entire input file.

-nonormalize: Do not normalize images (default).

-nopixfill: Do not convert out-of-range values in input file, just copy them through.

-pixfill: Replace out-of-range values in input file by the smallest possible value (default).

-pixfillvalue <value>: Specify a new pixel value to replace out-of-range values in the input file.

-scalar: Convert vector images to scalar images (a vector image is one with vector_dimension as the fastest varying dimension). The vector dimension is removed and values are averaged.

-noscalar: Do not convert vector images to scalar images (default).

+direction: Flip images to give positive step value for spatial axes. Note that the flipping of spatial axes only applies to "image dimensions". These are the two fastest varying (non-vector) dimensions in the file. If you want to flip a non-image dimension, you can convert it to an image dimension with -dimsize <dimname>=-1 (the -1 means don't really change the size). Check out the examples.

-direction: Flip images to give negative step value for spatial axes.

-anydirection: Don't flip images along spatial axes (default).

+xdirection: Flip images to give positive xspace:step value (left-to-right).

-xdirection: Flip images to give negative xspace:step value (right-to-left).

-xanydirection: Don't flip images along x-axis.

+ydirection: Flip images to give positive yspace:step value (posterior-to-anterior).

-ydirection: Flip images to give negative yspace:step value (anterior-to-posterior).

-yanydirection: Don't flip images along y-axis.

+zdirection: Flip images to give positive zspace:step value (inferior-to-superior).

-zdirection: Flip images to give negative zspace:step value (superior-to-inferior).

-zanydirection: Don't flip images along z-axis.

-keepaspect: Preserve aspect ratio when resizing images. This means that voxels are replicated (or averaged) the same number of times along each image dimension.

-nokeepaspect: Do not force preservation of aspect ratio when resizing images (default).

-imgsize <size>: Specify the desired image size (used if -rowsize or -colsize not given).

-rowsize <size>: Specify the desired number of rows in the image.

-colsize <size>: Specify the desired number of columns in the image.

-dimsize <dimension>=<size> Specify the size of a named dimension (<dimension>=<size>). Note that the resizing only applies to "image dimensions" - usually the two fastest-varying (non-vector) dimensions. To do dimension resizing, all fastest-varying dimensions up to the named dimension are turned into image dimensions, and these are then affected by the direction options. The dimension name and size must be in one command-line argument, so if you use spaces (which is okay), remember to use quotes to hide them from the shell.

-transverse: Write out transverse slices.

-sagittal: Write out sagittal slices.

-coronal: Write out coronal slices.

-dimorder <dim1>,<dim2>,<dim3>,...: Specify dimension order, where <dim?> are the names of the dimensions. You can give fewer dimensions than exist in the file: they are assumed to be the fastest varying dimensions in the output file (so -transverse is exactly equivalent to -dimorder zspace,yspace,xspace). Again, spaces are allowed between names, but remember to hide them from the shell with quotes.

-dimrange <dim>=<start>[,<count>]: Specify the range of dimension subscripts for dimension <dim>. If <count> is missing or 0, then it is taken to mean 1, but remove the dimension from the output file (a count of 1 will keep a dimension of size 1). A negative <count> means flip the data along that dimension - in this case <start> specifies the highest voxel coordinate for the dimension (-dimrange xspace=3,-3 gives a flipped version of -dimrange xspace=1,3). The options -start and -count provide an alternative way to specify the same information.

-start <coord0>,<coord1>,<coord2>,...: Specifies the starting corner of the hyperslab (coordinates go from slowest varying dimension to fastest). If fewer coordinates are given than dimensions exist in the file, then they are assumed to apply to the slowest varying dimensions and the remaining coordinates are set to 0. See -dimrange for more details. Both -start and -count give vectors that correspond to input file dimensions after the image conversion (icv) options have been applied.

-count <size0>,<size1>,<size2>,...: Specifies edge lengths of hyperslab to read (coordinates go from slowest varying dimension to fastest). If fewer sizes are given than dimensions exist in the file, then they are assumed to apply to the slowest varying dimensions and the remaining sizes are set to the full size of the dimension. See -dimrange for more details. Both -start and -count give vectors that correspond to input file dimensions after the image conversion (icv) options have been applied.

-nofill: Use value zero for points outside of the input volume (default).

-fill: Use a fill value for points outside of input volume (minimum possible value).

-fillvalue <fillvalue>: Specify a fill value for points outside of the input volume (this is a real value, not a pixel value).

-help: Print summary of command-line options and abort.

Assume that we have a volume with dimensions zspace, yspace, xspace (that's transverse) and sizes 128, 256, 256. If we want to get slice 40 out of it (keeping the coordinate information for the zspace dimension), then we can use

mincreshape original.mnc new.mnc \ -dimrange zspace=40,1

Alternatively, we could use

mincreshape original.mnc new.mnc \ -start 40,0,0 -count 1,256,256

Or simply

mincreshape original.mnc new.mnc \ -start 40 -count 1

If we wanted to get rid of the zspace dimension, we could use

mincreshape original.mnc new.mnc \ -dimrange zspace=40,0

Let's get a block out of the middle and flip it along xspace:

mincreshape original.mnc new.mnc \ -start 40,10,240 -count 1,200,-200

But why restrain outselves? Let's go out of bounds:

mincreshape original.mnc new.mnc \ -start 40,-100,340 -count 1,200,-200

If you want a sagittal volume, use this:

mincreshape original.mnc new.mnc -sagittal

How about some sideways heads - flip x and y. And convert to byte to save space while we're at it:

mincreshape original.mnc new.mnc \ -dimorder xspace,yspace -byte

You like to store volumes in x,y,z order (that's z varying fastest! I know some people who do it!)? Okay.

mincreshape original.mnc new.mnc \ -dimorder xspace,yspace,zspace

But you're a minimalist (and don't mind taking a chance). Here's the same thing (but it might break for another file):

mincreshape original.mnc new.mnc \ -dimorder zspace

Let's make sure that all dimensions have a negative step attribute (see option +direction for some details):

mincreshape original.mnc new.mnc \ -direction -dimsize zspace=-1

Peter Neelin

Copyright 1994 by Peter Neelin