rapidtide 3.0.11__py3-none-any.whl → 3.1.1__py3-none-any.whl

rapidtide/maskutil.py

@@ -18,9 +18,11 @@
 #
 import bisect
 import logging
+from typing import Any, Callable, Optional, Tuple, Union
 
 import numpy as np
 from nilearn import masking
+from numpy.typing import ArrayLike, NDArray
 from sklearn.decomposition import PCA
 
 import rapidtide.io as tide_io
@@ -30,16 +32,121 @@ import rapidtide.stats as tide_stats
 LGR = logging.getLogger("GENERAL")
 
 
-def resampmask(themask, thetargetres):
+def resampmask(themask: ArrayLike, thetargetres: float) -> NDArray:
+    """Resample a mask to a target resolution.
+
+    Parameters
+    ----------
+    themask : array_like
+        Input mask array to be resampled.
+    thetargetres : float
+        Target resolution for the resampled mask.
+
+    Returns
+    -------
+    NDArray
+        Resampled mask array with the specified target resolution.
+
+    Notes
+    -----
+    This function currently returns the input mask unchanged. A full implementation
+    would perform actual resampling operations to adjust the mask to the target
+    resolution.
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> mask = np.array([[0, 1, 0], [1, 1, 1], [0, 1, 0]])
+    >>> resampled = resampmask(mask, 0.5)
+    >>> print(resampled)
+    [[0 1 0]
+     [1 1 1]
+     [0 1 0]]
+    """
     resampmask = themask
     return themask
 
 
-def makeepimask(nim):
+def makeepimask(nim: Any) -> Any:
+    """Compute EPI mask from neuroimaging data.
+
+    This function computes an EPI (Echo Planar Imaging) mask from neuroimaging data
+    using the masking.compute_epi_mask function from nilearn.
+
+    Parameters
+    ----------
+    nim : Any
+        Neuroimaging data object (typically Nifti1Image or similar) from which
+        to compute the EPI mask. This can be a nibabel image object or array-like
+        data representing neuroimaging volumes.
+
+    Returns
+    -------
+    Any
+        EPI mask computed from the input neuroimaging data. The return type
+        depends on the underlying masking.compute_epi_mask implementation and
+        typically represents a binary mask image.
+
+    Notes
+    -----
+    This function is a wrapper around nilearn's masking.compute_epi_mask function
+    and is commonly used in neuroimaging preprocessing pipelines to automatically
+    generate brain masks from EPI functional MRI data.
+
+    Examples
+    --------
+    >>> import nibabel as nib
+    >>> from nilearn.masking import compute_epi_mask
+    >>> # Assuming 'img' is a nibabel image object
+    >>> mask = makeepimask(img)
+    >>> print(mask.shape)
+    """
     return masking.compute_epi_mask(nim)
 
 
-def maketmask(filename, timeaxis, maskvector, debug=False):
+def maketmask(
+    filename: str, timeaxis: ArrayLike, maskvector: NDArray, debug: bool = False
+) -> NDArray:
+    """Create a temporal mask from time interval data.
+
+    This function reads time interval data from a file and generates a binary mask
+    vector indicating which time points should be included in analysis. The mask
+    can be generated from either a simple vector of nonzero values or from time
+    intervals specified by start and duration values.
+
+    Parameters
+    ----------
+    filename : str
+        Path to the file containing time interval data. The file should contain
+        either a single vector of values or two rows of data (start times and durations).
+    timeaxis : ArrayLike
+        Array of time points corresponding to the fMRI time series. Used to map
+        time intervals to specific time indices.
+    maskvector : NDArray
+        Pre-allocated array to store the resulting temporal mask. Should be the
+        same length as the fMRI time series.
+    debug : bool, optional
+        If True, enables debug logging output. Default is False.
+
+    Returns
+    -------
+    NDArray
+        Binary mask vector where 1.0 indicates time points to include and 0.0
+        indicates time points to exclude.
+
+    Notes
+    -----
+    The function handles two input formats:
+    1. Single row: Each nonzero value indicates inclusion of the corresponding time point
+    2. Two rows: First row contains start times, second row contains durations
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> timeaxis = np.arange(0, 100, 2)  # 50 time points
+    >>> maskvector = np.zeros(50)
+    >>> mask = maketmask('time_intervals.txt', timeaxis, maskvector)
+    """
     inputdata = tide_io.readvecs(filename)
     theshape = np.shape(inputdata)
     if theshape[0] == 1:
@@ -61,16 +168,75 @@ def maketmask(filename, timeaxis, maskvector, debug=False):
 
 
 def readamask(
-    maskfilename,
-    nim_hdr,
-    xsize,
-    istext=False,
-    valslist=None,
-    thresh=None,
-    maskname="the",
-    tolerance=1.0e-3,
-    debug=False,
-):
+    maskfilename: str,
+    nim_hdr: Any,
+    xsize: int,
+    istext: bool = False,
+    valslist: Optional[list] = None,
+    thresh: Optional[float] = None,
+    maskname: str = "the",
+    tolerance: float = 1.0e-3,
+    debug: bool = False,
+) -> NDArray:
+    """
+    Read and process a mask file, returning a binary mask array.
+
+    This function reads a mask from either a text file or NIfTI format, applies
+    optional thresholding or value selection, and returns a binary mask array
+    compatible with the input data dimensions.
+
+    Parameters
+    ----------
+    maskfilename : str
+        Path to the mask file. Can be in text format (if `istext=True`) or NIfTI format.
+    nim_hdr : Any
+        Header information from the NIfTI file of the input data. Used for spatial
+        dimension matching.
+    xsize : int
+        Expected size of the first dimension of the mask array.
+    istext : bool, optional
+        If True, the mask is read as a text file. Default is False.
+    valslist : list of int, optional
+        List of values to include in the mask. If provided, only voxels matching
+        these values are set to 1. Default is None.
+    thresh : float, optional
+        Threshold value for binarizing the mask. If provided, voxels greater than
+        `thresh` are set to 1, others to 0. Default is None.
+    maskname : str, optional
+        Name of the mask for logging and error messages. Default is "the".
+    tolerance : float, optional
+        Tolerance for spatial dimension matching between the mask and input data.
+        Default is 1e-3.
+    debug : bool, optional
+        If True, print debug information. Default is False.
+
+    Returns
+    -------
+    NDArray
+        A binary mask array of type `uint16`, where 1 indicates included voxels
+        and 0 indicates excluded voxels.
+
+    Notes
+    -----
+    - If `istext=True`, the mask file is expected to contain numeric values
+      arranged in a single column or row.
+    - If `thresh` is provided, the mask is binarized based on the threshold.
+    - If `valslist` is provided, only voxels matching values in the list are set to 1.
+    - The function raises a `ValueError` if spatial dimensions of the mask and
+      input data do not match within the specified tolerance.
+
+    Examples
+    --------
+    >>> mask = readamask(
+    ...     maskfilename="mask.nii.gz",
+    ...     nim_hdr=hdr,
+    ...     xsize=64,
+    ...     thresh=0.5,
+    ...     maskname="brain"
+    ... )
+    >>> print(mask.shape)
+    (64, 64, 64)
+    """
     LGR.debug(f"readamask called with filename: {maskfilename} vals: {valslist}")
     if debug:
         print("getmaskset:")
@@ -108,19 +274,86 @@ def readamask(
 
 
 def getmaskset(
-    maskname,
-    includename,
-    includevals,
-    excludename,
-    excludevals,
-    datahdr,
-    numspatiallocs,
-    extramask=None,
-    extramaskthresh=0.1,
-    istext=False,
-    tolerance=1.0e-3,
-    debug=False,
-):
+    maskname: str,
+    includename: Optional[str],
+    includevals: Optional[list],
+    excludename: Optional[str],
+    excludevals: Optional[list],
+    datahdr: Any,
+    numspatiallocs: int,
+    extramask: Optional[str] = None,
+    extramaskthresh: float = 0.1,
+    istext: bool = False,
+    tolerance: float = 1.0e-3,
+    debug: bool = False,
+) -> Tuple[Optional[NDArray], Optional[NDArray], Optional[NDArray]]:
+    """
+    Construct and return masks for inclusion, exclusion, and an additional mask.
+
+    This function builds masks based on provided parameters, including optional
+    inclusion and exclusion criteria, as well as an extra mask. It performs
+    validation to ensure that the resulting masks are not empty or overly restrictive.
+
+    Parameters
+    ----------
+    maskname : str
+        Name of the mask being constructed, used for logging and labeling.
+    includename : str, optional
+        File name or identifier for the mask to be used for inclusion.
+    includevals : list of float, optional
+        List of values to include in the inclusion mask. If ``None``, all values
+        are included.
+    excludename : str, optional
+        File name or identifier for the mask to be used for exclusion.
+    excludevals : list of float, optional
+        List of values to exclude from the exclusion mask. If ``None``, no values
+        are excluded.
+    datahdr : Any
+        Header information for the data, passed to mask reading functions.
+    numspatiallocs : int
+        Number of spatial locations in the data.
+    extramask : str, optional
+        File name or identifier for an additional mask to be applied.
+    extramaskthresh : float, default=0.1
+        Threshold value for the extra mask, used when reading the mask.
+    istext : bool, default=False
+        If ``True``, treat input files as text-based.
+    tolerance : float, default=1e-03
+        Tolerance for floating-point comparisons when reading masks.
+    debug : bool, default=False
+        If ``True``, print debug information during execution.
+
+    Returns
+    -------
+    tuple of (Optional[NDArray], Optional[NDArray], Optional[NDArray])
+        A tuple containing:
+        - ``internalincludemask``: The inclusion mask, reshaped to ``numspatiallocs``.
+        - ``internalexcludemask``: The exclusion mask, reshaped to ``numspatiallocs``.
+        - ``internalextramask``: The extra mask, reshaped to ``numspatiallocs``.
+
+    Notes
+    -----
+    - If both inclusion and exclusion masks are specified, the function ensures
+      that at least one voxel remains after applying both masks.
+    - If an extra mask is specified, it is applied in combination with the inclusion
+      and exclusion masks.
+    - The function raises a ``ValueError`` if any of the resulting masks are invalid:
+      e.g., empty inclusion mask, or masks that leave no voxels.
+
+    Examples
+    --------
+    >>> maskname = "brain"
+    >>> includename = "brain_include.nii"
+    >>> includevals = [1]
+    >>> excludename = "ventricles.nii"
+    >>> excludevals = [1]
+    >>> datahdr = header
+    >>> numspatiallocs = 10000
+    >>> includemask, excludemask, extramask = getmaskset(
+    ...     maskname, includename, includevals, excludename, excludevals,
+    ...     datahdr, numspatiallocs
+    ... )
+    """
     internalincludemask = None
     internalexcludemask = None
     internalextramask = None
@@ -204,17 +437,81 @@ def getmaskset(
 
 
 def getregionsignal(
-    indata,
-    filter=None,
-    Fs=1.0,
-    includemask=None,
-    excludemask=None,
-    signalgenmethod="sum",
-    pcacomponents=0.8,
-    signame="global mean",
-    rt_floatset=np.float64,
-    debug=False,
-):
+    indata: NDArray,
+    filter: Optional[Any] = None,
+    Fs: float = 1.0,
+    includemask: Optional[NDArray] = None,
+    excludemask: Optional[NDArray] = None,
+    signalgenmethod: str = "sum",
+    pcacomponents: Union[float, str] = 0.8,
+    signame: str = "global mean",
+    rt_floattype: type = np.float64,
+    debug: bool = False,
+) -> Tuple[NDArray, NDArray]:
+    """
+    Compute a global signal from a 2D array of voxel data using specified methods.
+
+    This function computes a global signal from input data by applying optional masking,
+    and then combining voxel signals using one of several methods: sum, meanscale, PCA,
+    or random. The resulting signal can be filtered and normalized.
+
+    Parameters
+    ----------
+    indata : ndarray
+        Input 2D array of shape (n_voxels, n_timepoints) containing voxel time series.
+    filter : optional
+        A filter object with an `apply` method to apply to the computed signal.
+        Default is None.
+    Fs : float, optional
+        Sampling frequency (Hz) used for filtering. Default is 1.0.
+    includemask : ndarray, optional
+        Binary mask to include specific voxels. Voxels not included will be ignored.
+        Default is None.
+    excludemask : ndarray, optional
+        Binary mask to exclude specific voxels. Voxels marked as 1 will be excluded.
+        Default is None.
+    signalgenmethod : str, optional
+        Method used to generate the global signal. Options are:
+        - "sum": Mean of selected voxels (default).
+        - "meanscale": Scale each voxel by its mean before averaging.
+        - "pca": Use PCA to reduce dimensionality and compute signal.
+        - "random": Generate a random signal.
+        Default is "sum".
+    pcacomponents : float or str, optional
+        Number of PCA components to use. If float, specifies number of components;
+        if "mle", uses maximum likelihood estimation. Default is 0.8.
+    signame : str, optional
+        Name of the signal for logging purposes. Default is "global mean".
+    rt_floattype : type, optional
+        Data type for internal computations. Default is np.float64.
+    debug : bool, optional
+        If True, print debugging information. Default is False.
+
+    Returns
+    -------
+    tuple of ndarray
+        A tuple containing:
+        - normalized_global_signal : ndarray
+          The normalized global signal of shape (n_timepoints,).
+        - final_mask : ndarray
+          The final voxel mask used in computation, shape (n_voxels,).
+
+    Notes
+    -----
+    - The function applies `includemask` and `excludemask` sequentially to define
+      which voxels are used in signal computation.
+    - For "pca" method, PCA is applied to the transposed scaled voxel data.
+    - If filtering is applied, the signal is filtered in-place using the provided filter.
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> from sklearn.decomposition import PCA
+    >>> indata = np.random.rand(100, 50)
+    >>> signal, mask = getregionsignal(indata, signalgenmethod="sum")
+    >>> print(signal.shape)
+    (50,)
+    """
     # Start with all voxels
     themask = indata[:, 0] * 0 + 1
 
@@ -225,7 +522,7 @@ def getregionsignal(
         themask = themask * (1 - excludemask)
 
     # combine all the voxels using one of the three methods
-    globalmean = rt_floatset(indata[0, :])
+    globalmean = (indata[0, :]).astype(rt_floattype)
     thesize = np.shape(themask)
     numvoxelsused = int(np.sum(np.where(themask > 0.0, 1, 0)))
     selectedvoxels = indata[np.where(themask > 0.0), :][0]
@@ -235,36 +532,49 @@ def getregionsignal(
     if signalgenmethod == "sum":
         globalmean = np.mean(selectedvoxels, axis=0)
         globalmean -= np.mean(globalmean)
+        if debug:
+            print("Sum method")
+            print(f"getregionsignal: {globalmean.shape=}")
     elif signalgenmethod == "meanscale":
         themean = np.mean(indata, axis=1)
         for vox in range(0, thesize[0]):
             if themask[vox] > 0.0:
                 if themean[vox] != 0.0:
                     globalmean += indata[vox, :] / themean[vox] - 1.0
+        if debug:
+            print("Meanscale method")
+            print(f"getregionsignal: {globalmean.shape=}")
     elif signalgenmethod == "pca":
         themean = np.mean(indata, axis=1)
         thevar = np.var(indata, axis=1)
-        scaledvoxels = selectedvoxels * 0.0
+        scaledvoxels = np.zeros_like(selectedvoxels)
         for vox in range(0, selectedvoxels.shape[0]):
             scaledvoxels[vox, :] = selectedvoxels[vox, :] - themean[vox]
             if thevar[vox] > 0.0:
                 scaledvoxels[vox, :] = selectedvoxels[vox, :] / thevar[vox]
         try:
-            thefit = PCA(n_components=pcacomponents).fit(np.transpose(scaledvoxels))
+            thefit = PCA(n_components=pcacomponents).fit(scaledvoxels)
         except ValueError:
             if pcacomponents == "mle":
                 LGR.warning("mle estimation failed - falling back to pcacomponents=0.8")
-                thefit = PCA(n_components=0.8).fit(np.transpose(scaledvoxels))
+                thefit = PCA(n_components=0.8).fit(scaledvoxels)
             else:
                 raise ValueError("unhandled math exception in PCA refinement - exiting")
 
         varex = 100.0 * np.cumsum(thefit.explained_variance_ratio_)[len(thefit.components_) - 1]
-        thetransform = thefit.transform(np.transpose(scaledvoxels))
-        if debug:
-            print(f"getregionsignal: {thetransform.shape=}")
-        globalmean = np.mean(thetransform, axis=0)
+        # thetransform = thefit.transform(np.transpose(scaledvoxels))
+        thetransform = thefit.transform(scaledvoxels)
+        cleanedvoxels = thefit.inverse_transform(thetransform) * thevar[:, None]
+        globalmean = np.mean(cleanedvoxels, axis=0)
         globalmean -= np.mean(globalmean)
         if debug:
+            print("PCA method")
+            print(
+                f"getregionsignal: {cleanedvoxels.shape=}, {thetransform.shape=}, {scaledvoxels.shape=}, {globalmean.shape=}"
+            )
+            print(
+                f"getregionsignal: {(thefit.components_).shape=}, {thefit.n_samples_=}, {thefit.n_features_in_=}"
+            )
             print(f"getregionsignal: {varex=}")
         LGR.info(
             f"Using {len(thefit.components_)} component(s), accounting for "
@@ -272,6 +582,9 @@ def getregionsignal(
         )
     elif signalgenmethod == "random":
         globalmean = np.random.standard_normal(size=len(globalmean))
+        if debug:
+            print("Random method")
+            print(f"getregionsignal: {globalmean.shape=}")
     else:
         raise ValueError(f"illegal signal generation method: {signalgenmethod}")
     LGR.info(f"used {numvoxelsused} voxels to calculate {signame} signal")
@@ -283,22 +596,95 @@ def getregionsignal(
 
 
 def saveregionaltimeseries(
-    tcdesc,
-    tcname,
-    fmridata,
-    includemask,
-    fmrifreq,
-    outputname,
-    filter=None,
-    initfile=False,
-    excludemask=None,
-    filedesc="regional",
-    suffix="",
-    signalgenmethod="sum",
-    pcacomponents=0.8,
-    rt_floatset=np.float64,
-    debug=False,
-):
+    tcdesc: str,
+    tcname: str,
+    fmridata: NDArray,
+    includemask: NDArray,
+    fmrifreq: float,
+    outputname: str,
+    filter: Optional[Any] = None,
+    initfile: bool = False,
+    excludemask: Optional[NDArray] = None,
+    filedesc: str = "regional",
+    suffix: str = "",
+    signalgenmethod: str = "sum",
+    pcacomponents: Union[float, str] = 0.8,
+    rt_floattype: type = np.float64,
+    debug: bool = False,
+) -> Tuple[NDArray, NDArray]:
+    """
+    Save regional time series data from fMRI data to a BIDS-compatible TSV file.
+
+    This function extracts regional signal time courses from fMRI data using the
+    specified masking and filtering parameters, then writes the results to a
+    BIDS-style TSV file. The function supports various signal generation methods
+    and can handle both inclusive and exclusive masking.
+
+    Parameters
+    ----------
+    tcdesc : str
+        Description of the time course for the output file header
+    tcname : str
+        Name of the time course to be used in the output file column header
+    fmridata : NDArray
+        4D fMRI data array (time x x x y z)
+    includemask : NDArray
+        Binary mask defining regions to include in the analysis
+    fmrifreq : float
+        Sampling frequency of the fMRI data (Hz)
+    outputname : str
+        Base name for the output file (without extension)
+    filter : Optional[Any], default=None
+        Filter to apply to the time series data
+    initfile : bool, default=False
+        If True, initializes a new file; if False, appends to existing file
+    excludemask : Optional[NDArray], default=None
+        Binary mask defining regions to exclude from the analysis
+    filedesc : str, default="regional"
+        Description string for the output file name
+    suffix : str, default=""
+        Suffix to append to the column name in the output file
+    signalgenmethod : str, default="sum"
+        Method for generating the signal ('sum', 'mean', 'pca', etc.)
+    pcacomponents : Union[float, str], default=0.8
+        Number of PCA components to use (or fraction of variance explained)
+    rt_floattype : np.dtype, default=np.float64
+        Data type for floating point operations
+    debug : bool, default=False
+        If True, enables debug mode for additional logging
+
+    Returns
+    -------
+    Tuple[NDArray, NDArray]
+        Tuple containing:
+        - thetimecourse : NDArray
+          The extracted time course data
+        - themask : NDArray
+          The mask used for extraction
+
+    Notes
+    -----
+    The function uses `getregionsignal` to compute the regional signal and
+    `tide_io.writebidstsv` to write the output file in BIDS TSV format.
+    The output file name follows the pattern:
+    {outputname}_desc-{filedesc}_timeseries.tsv
+
+    Examples
+    --------
+    >>> import numpy as np
+    >>> fmri_data = np.random.rand(100, 10, 10, 10)
+    >>> mask = np.ones((10, 10, 10))
+    >>> timecourse, mask_used = saveregionaltimeseries(
+    ...     tcdesc="mean_signal",
+    ...     tcname="signal",
+    ...     fmridata=fmri_data,
+    ...     includemask=mask,
+    ...     fmrifreq=2.0,
+    ...     outputname="sub-01_task-rest",
+    ...     filter=None,
+    ...     initfile=True
+    ... )
+    """
     thetimecourse, themask = getregionsignal(
         fmridata,
         filter=filter,
@@ -308,7 +694,7 @@ def saveregionaltimeseries(
         signalgenmethod=signalgenmethod,
         pcacomponents=pcacomponents,
         signame=tcdesc,
-        rt_floatset=rt_floatset,
+        rt_floattype=rt_floattype,
         debug=debug,
     )
     tide_io.writebidstsv(