rapidtide 3.1__py3-none-any.whl → 3.1.1__py3-none-any.whl

rapidtide/linfitfiltpass.py

@@ -16,7 +16,7 @@
 #   limitations under the License.
 #
 #
-from typing import Any, Callable
+from typing import Any
 
 import numpy as np
 from numpy.typing import NDArray
@@ -33,9 +33,8 @@ def _procOneRegressionFitItem(
     vox: int,
     theevs: NDArray,
     thedata: NDArray,
-    rt_floatset: type = np.float64,
-    rt_floattype: str = "float64",
-) -> tuple[int, Any, Any, Any, Any, Any, NDArray, NDArray]:
+    rt_floattype: np.dtype = np.float64,
+) -> tuple[int, float, float, float, float | NDArray, Any, NDArray, NDArray]:
     """
     Perform single regression fit on voxel data and return fit results.
 
@@ -47,28 +46,26 @@ def _procOneRegressionFitItem(
     ----------
     vox : int
         Voxel index.
-    theevs : numpy.ndarray
+    theevs : NDArray
         Experimental design matrix. If 2D, dimension 0 is number of points,
         dimension 1 is number of evs.
-    thedata : numpy.ndarray
+    thedata : NDArray
         Dependent variable data corresponding to the evs.
-    rt_floatset : type, optional
-        Data type for floating-point results, default is ``np.float64``.
     rt_floattype : str, optional
-        String representation of the floating-point type, default is ``"float64"``.
+        String representation of the floating-point type, default is ``np.float64``.
 
     Returns
     -------
-    tuple[int, Any, Any, Any, Any, Any, numpy.ndarray, numpy.ndarray]
+    tuple[int, float, float, float, float, Any, NDArray, NDArray]
         A tuple containing:
         - voxel index (`int`)
-        - intercept term (`Any`)
-        - signed square root of R-squared (`Any`)
-        - R-squared value (`Any`)
-        - fit coefficients (`Any` or `numpy.ndarray`)
+        - intercept term (`float`)
+        - signed square root of R-squared (`float`)
+        - R-squared value (`float`)
+        - fit coefficients (`float` or `NDArray`)
         - normalized fit coefficients (`Any`)
-        - data removed by fitting (`numpy.ndarray`)
-        - residuals (`numpy.ndarray`)
+        - data removed by fitting (`NDArray`)
+        - residuals (`NDArray`)
 
     Notes
     -----
@@ -80,7 +77,6 @@ def _procOneRegressionFitItem(
     Examples
     --------
     >>> import numpy as np
-    >>> from tide_fit import mlregress
     >>> theevs = np.array([[1, 2], [3, 4], [5, 6]], dtype=np.float64)
     >>> thedata = np.array([1, 2, 3], dtype=np.float64)
     >>> result = _procOneRegressionFitItem(0, theevs, thedata)
@@ -92,31 +88,31 @@ def _procOneRegressionFitItem(
     if theevs.ndim > 1:
         if thefit is None:
             thefit = np.matrix(np.zeros((1, theevs.shape[1] + 1), dtype=rt_floattype))
-        fitcoeffs = rt_floatset(thefit[0, 1:])
+        fitcoeffs = (thefit[0, 1:]).astype(rt_floattype)
         if fitcoeffs[0, 0] < 0.0:
             coeffsign = -1.0
         else:
             coeffsign = 1.0
         datatoremove = theevs[:, 0] * 0.0
         for j in range(theevs.shape[1]):
-            datatoremove += rt_floatset(rt_floatset(thefit[0, 1 + j]) * theevs[:, j])
+            datatoremove += (thefit[0, 1 + j] * theevs[:, j]).astype(rt_floattype)
         if np.any(fitcoeffs) != 0.0:
             pass
         else:
             R2 = 0.0
         return (
             vox,
-            rt_floatset(thefit[0, 0]),
-            rt_floatset(coeffsign * np.sqrt(R2)),
-            rt_floatset(R2),
+            thefit[0, 0],
+            coeffsign * np.sqrt(R2),
+            R2,
             fitcoeffs,
-            rt_floatset(thefit[0, 1:] / thefit[0, 0]),
+            (thefit[0, 1:] / thefit[0, 0]).astype(rt_floattype),
             datatoremove,
-            rt_floatset(thedata - datatoremove),
+            (thedata - datatoremove).astype(rt_floattype),
         )
     else:
-        fitcoeff = rt_floatset(thefit[0, 1])
-        datatoremove = rt_floatset(fitcoeff * theevs)
+        fitcoeff = (thefit[0, 1]).astype(rt_floattype)
+        datatoremove = (fitcoeff * theevs).astype(rt_floattype)
         if fitcoeff < 0.0:
             coeffsign = -1.0
         else:
@@ -125,13 +121,13 @@ def _procOneRegressionFitItem(
             R2 = 0.0
         return (
             vox,
-            rt_floatset(thefit[0, 0]),
-            rt_floatset(coeffsign * np.sqrt(R2)),
-            rt_floatset(R2),
+            thefit[0, 0],
+            coeffsign * np.sqrt(R2),
+            R2,
             fitcoeff,
-            rt_floatset(thefit[0, 1] / thefit[0, 0]),
+            (thefit[0, 1] / thefit[0, 0]).astype(rt_floattype),
             datatoremove,
-            rt_floatset(thedata - datatoremove),
+            (thedata - datatoremove).astype(rt_floattype),
         )
 
 
@@ -146,7 +142,7 @@ def linfitfiltpass(
     fitcoeff: NDArray | None,
     fitNorm: NDArray | None,
     datatoremove: NDArray | None,
-    filtereddata: NDArray,
+    filtereddata: NDArray | None,
     nprocs: int = 1,
     alwaysmultiproc: bool = False,
     constantevs: bool = False,
@@ -155,8 +151,7 @@ def linfitfiltpass(
     procbyvoxel: bool = True,
     showprogressbar: bool = True,
     chunksize: int = 1000,
-    rt_floatset: type = np.float64,
-    rt_floattype: str = "float64",
+    rt_floattype: np.dtype = np.float64,
     verbose: bool = True,
     debug: bool = False,
 ) -> int:
@@ -207,10 +202,8 @@ def linfitfiltpass(
         If True, display a progress bar during processing.
     chunksize : int, default: 1000
         Size of chunks for multiprocessing.
-    rt_floatset : type, default: np.float64
+    rt_floattype : str, default: np.float64
         Data type for internal floating-point calculations.
-    rt_floattype : str, default: "float64"
-        String representation of the floating-point data type.
     verbose : bool, default: True
         If True, print verbose output.
     debug : bool, default: False
@@ -303,7 +296,6 @@ def linfitfiltpass(
                                     val,
                                     theevs,
                                     fmri_data[val, :],
-                                    rt_floatset=rt_floatset,
                                     rt_floattype=rt_floattype,
                                 )
                             )
@@ -313,7 +305,6 @@ def linfitfiltpass(
                                     val,
                                     theevs[val, :],
                                     fmri_data[val, :],
-                                    rt_floatset=rt_floatset,
                                     rt_floattype=rt_floattype,
                                 )
                             )
@@ -324,7 +315,6 @@ def linfitfiltpass(
                                     val,
                                     theevs,
                                     fmri_data[:, val],
-                                    rt_floatset=rt_floatset,
                                     rt_floattype=rt_floattype,
                                 )
                             )
@@ -334,7 +324,6 @@ def linfitfiltpass(
                                     val,
                                     theevs[:, val],
                                     fmri_data[:, val],
-                                    rt_floatset=rt_floatset,
                                     rt_floattype=rt_floattype,
                                 )
                             )
@@ -449,7 +438,6 @@ def linfitfiltpass(
                             vox,
                             theevs,
                             thedata,
-                            rt_floatset=rt_floatset,
                             rt_floattype=rt_floattype,
                         )
                     elif coefficientsonly:
@@ -467,7 +455,6 @@ def linfitfiltpass(
                                 vox,
                                 theevs[vox, :],
                                 thedata,
-                                rt_floatset=rt_floatset,
                                 rt_floattype=rt_floattype,
                             )
                         else:
@@ -484,7 +471,6 @@ def linfitfiltpass(
                                 vox,
                                 theevs,
                                 thedata,
-                                rt_floatset=rt_floatset,
                                 rt_floattype=rt_floattype,
                             )
                     else:
@@ -501,7 +487,6 @@ def linfitfiltpass(
                             vox,
                             theevs[vox, :],
                             thedata,
-                            rt_floatset=rt_floatset,
                             rt_floattype=rt_floattype,
                         )
                     itemstotal += 1
@@ -528,7 +513,6 @@ def linfitfiltpass(
                             timepoint,
                             theevs,
                             thedata,
-                            rt_floatset=rt_floatset,
                             rt_floattype=rt_floattype,
                         )
                     elif coefficientsonly:
@@ -546,7 +530,6 @@ def linfitfiltpass(
                                 timepoint,
                                 theevs[:, timepoint],
                                 thedata,
-                                rt_floatset=rt_floatset,
                                 rt_floattype=rt_floattype,
                             )
                         else:
@@ -563,7 +546,6 @@ def linfitfiltpass(
                                 timepoint,
                                 theevs,
                                 thedata,
-                                rt_floatset=rt_floatset,
                                 rt_floattype=rt_floattype,
                             )
                     else:
@@ -580,7 +562,6 @@ def linfitfiltpass(
                             timepoint,
                             theevs[:, timepoint],
                             thedata,
-                            rt_floatset=rt_floatset,
                             rt_floattype=rt_floattype,
                         )
 

rapidtide/makelaggedtcs.py

@@ -49,7 +49,7 @@ def _procOneVoxelMakelagtc(
         3. `timeaxis` - Time axis array for evaluation
     **kwargs : Any
         Additional keyword arguments that can override default options:
-        - `rt_floatset` : float type for computations (default: np.float64)
+        - `rt_floattype` : float type for computations (default: np.float64)
         - `debug` : boolean for debug printing (default: False)
 
     Returns
@@ -82,11 +82,11 @@ def _procOneVoxelMakelagtc(
     """
     # unpack arguments
     options = {
-        "rt_floatset": np.float64,
+        "rt_floattype": np.float64,
         "debug": False,
     }
     options.update(kwargs)
-    rt_floatset = options["rt_floatset"]
+    rt_floattype = options["rt_floattype"]
     debug = options["debug"]
     (lagtcgenerator, thelag, timeaxis) = voxelargs
     if debug:
@@ -94,7 +94,7 @@ def _procOneVoxelMakelagtc(
 
     # question - should maxlag be added or subtracted?  As of 10/18, it is subtracted
     #  potential answer - tried adding, results are terrible.
-    thelagtc = rt_floatset(lagtcgenerator.yfromx(timeaxis - thelag))
+    thelagtc = (lagtcgenerator.yfromx(timeaxis - thelag)).astype(rt_floattype)
 
     return (
         vox,
@@ -189,8 +189,7 @@ def makelaggedtcs(
     alwaysmultiproc: bool = False,
     showprogressbar: bool = True,
     chunksize: int = 1000,
-    rt_floatset: type = np.float64,
-    rt_floattype: str = "float64",
+    rt_floattype: np.dtype = np.float64,
     debug: bool = False,
 ) -> int:
     """
@@ -204,14 +203,14 @@ def makelaggedtcs(
     ----------
     lagtcgenerator : Any
         A callable or object that generates lagged timecourses for a single voxel.
-    timeaxis : numpy.ndarray
+    timeaxis : NDArray
         1D array representing the time axis (e.g., TRs or time points).
-    lagmask : numpy.ndarray
+    lagmask : NDArray
         3D or 4D boolean or integer array defining the voxels to process.
         Non-zero entries indicate voxels to be processed.
-    lagtimes : numpy.ndarray
+    lagtimes : NDArray
         1D array of lag times (in seconds or time units) to be applied.
-    lagtc : numpy.ndarray
+    lagtc : NDArray
         4D array of shape (ntimepoints, nvoxels, nlags) to store the output lagged
         timecourses. This is updated in-place.
     LGR : logging.Logger, optional
@@ -226,12 +225,9 @@ def makelaggedtcs(
     chunksize : int, optional
         Size of chunks to process in each step when using multiprocessing.
         Default is 1000.
-    rt_floatset : type, optional
-        The floating-point type to use for internal computations.
-        Default is `np.float64`.
     rt_floattype : str, optional
         String representation of the floating-point type.
-        Default is "float64".
+        Default is `np.float64`.
     debug : bool, optional
         If True, print debug information. Default is False.
 
@@ -249,7 +245,6 @@ def makelaggedtcs(
     Examples
     --------
     >>> import numpy as np
-    >>> from some_module import makelaggedtcs
     >>> timeaxis = np.arange(100)
     >>> lagtimes = np.array([0, 1, 2])
     >>> lagmask = np.ones((10, 10, 10), dtype=bool)
@@ -294,7 +289,7 @@ def makelaggedtcs(
         alwaysmultiproc,
         showprogressbar,
         chunksize,
-        rt_floatset=rt_floatset,
+        rt_floattype=rt_floattype,
     )
     if LGR is not None:
         LGR.info(f"\nLagged timecourses created for {volumetotal} voxels")

rapidtide/maskutil.py

@@ -44,7 +44,7 @@ def resampmask(themask: ArrayLike, thetargetres: float) -> NDArray:
 
     Returns
     -------
-    numpy.ndarray
+    NDArray
         Resampled mask array with the specified target resolution.
 
     Notes
@@ -445,7 +445,7 @@ def getregionsignal(
     signalgenmethod: str = "sum",
     pcacomponents: Union[float, str] = 0.8,
     signame: str = "global mean",
-    rt_floatset: type = np.float64,
+    rt_floattype: type = np.float64,
     debug: bool = False,
 ) -> Tuple[NDArray, NDArray]:
     """
@@ -482,7 +482,7 @@ def getregionsignal(
         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_floatset : type, optional
+    rt_floattype : type, optional
         Data type for internal computations. Default is np.float64.
     debug : bool, optional
         If True, print debugging information. Default is False.
@@ -522,7 +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]
@@ -532,36 +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 "
@@ -569,6 +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")
@@ -593,7 +609,7 @@ def saveregionaltimeseries(
     suffix: str = "",
     signalgenmethod: str = "sum",
     pcacomponents: Union[float, str] = 0.8,
-    rt_floatset: type = np.float64,
+    rt_floattype: type = np.float64,
     debug: bool = False,
 ) -> Tuple[NDArray, NDArray]:
     """
@@ -632,7 +648,7 @@ def saveregionaltimeseries(
         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_floatset : type, default=np.float64
+    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
@@ -678,7 +694,7 @@ def saveregionaltimeseries(
         signalgenmethod=signalgenmethod,
         pcacomponents=pcacomponents,
         signame=tcdesc,
-        rt_floatset=rt_floatset,
+        rt_floattype=rt_floattype,
         debug=debug,
     )
     tide_io.writebidstsv(

rapidtide/miscmath.py

@@ -565,12 +565,12 @@ def stdnormalize(vector: NDArray) -> NDArray:
 
     Parameters
     ----------
-    vector : numpy.ndarray
+    vector : NDArray
         Input vector to be standardized.
 
     Returns
     -------
-    numpy.ndarray
+    NDArray
         Standardized vector with zero mean and unit variance. If the input vector
         has zero standard deviation, the demeaned vector is returned unchanged.
 

rapidtide/patchmatch.py

@@ -21,7 +21,6 @@ import math
 import os
 import sys
 import warnings
-from typing import Any
 
 import numpy as np
 from numpy.typing import NDArray
@@ -41,15 +40,15 @@ def interpolate_masked_voxels(
     from the unmasked region. Supports boundary extrapolation and multiple interpolation methods.
 
     Parameters:
-        data (np.ndarray): A 3D numpy array containing the data.
-        mask (np.ndarray): A 3D binary numpy array of the same shape as `data`,
+        data (NDArray): A 3D numpy array containing the data.
+        mask (NDArray): A 3D binary numpy array of the same shape as `data`,
                            where 1 indicates masked voxels and 0 indicates unmasked voxels.
         method (str): Interpolation method ('linear', 'nearest', or 'cubic').
         extrapolate (bool): Whether to extrapolate values for masked voxels outside the convex hull
                             of the unmasked points.
 
     Returns:
-        np.ndarray: A new 3D array with interpolated (and optionally extrapolated)
+        NDArray: A new 3D array with interpolated (and optionally extrapolated)
                     values replacing masked regions.
     """
     if data.shape != mask.shape:
@@ -97,7 +96,7 @@ def get_bounding_box(mask: NDArray, value: int, buffer: int = 0) -> tuple[tuple,
 
     Parameters
     ----------
-    mask : np.ndarray
+    mask : NDArray
         A 3D binary mask where non-zero values indicate the masked region.
     value : int
         The masked region value to compute the bounding box for.
@@ -411,19 +410,19 @@ def _smooth_array(
 
     Parameters
     ----------
-    arr : :class:`numpy.ndarray`
+    arr : :class:`NDArray`
         4D array, with image number as last dimension. 3D arrays are also
         accepted.
 
-    affine : :class:`numpy.ndarray`
+    affine : :class:`NDArray`
         (4, 4) matrix, giving affine transformation for image. (3, 3) matrices
         are also accepted (only these coefficients are used).
         If `fwhm='fast'`, the affine is not used and can be None.
 
-    fwhm : scalar, :class:`numpy.ndarray`/:obj:`tuple`/:obj:`list`, 'fast' or None, optional
+    fwhm : scalar, :class:`NDArray`/:obj:`tuple`/:obj:`list`, 'fast' or None, optional
         Smoothing strength, as a full-width at half maximum, in millimeters.
         If a nonzero scalar is given, width is identical in all 3 directions.
-        A :class:`numpy.ndarray`, :obj:`tuple`, or :obj:`list` must have 3 elements,
+        A :class:`NDArray`, :obj:`tuple`, or :obj:`list` must have 3 elements,
         giving the FWHM along each axis.
         If any of the elements is zero or None, smoothing is not performed
         along that axis.
@@ -443,7 +442,7 @@ def _smooth_array(
 
     Returns
     -------
-    :class:`numpy.ndarray`
+    :class:`NDArray`
         Filtered `arr`.
 
     Notes
@@ -517,7 +516,7 @@ def difference_of_gaussian(
     Parameters
     ----------
     fdata : numpy.memmap from Niimg-like object
-    affine : :class:`numpy.ndarray`
+    affine : :class:`NDArray`
         (4, 4) matrix, giving affine transformation for image. (3, 3) matrices
         are also accepted (only these coefficients are used).
     fwhmNarrow : int

rapidtide/peakeval.py

@@ -79,8 +79,7 @@ def peakevalpass(
     interptype: str = "univariate",
     showprogressbar: bool = True,
     chunksize: int = 1000,
-    rt_floatset: type = np.float64,
-    rt_floattype: str = "float64",
+    rt_floattype: np.dtype = np.float64,
 ) -> tuple[int, dict]:
     """
 
@@ -100,7 +99,6 @@ def peakevalpass(
     interptype
     showprogressbar
     chunksize
-    rt_floatset
     rt_floattype
 
     Returns

rapidtide/ppgproc.py

@@ -1623,7 +1623,7 @@ class HeartRateExtractor:
 
         Parameters
         ----------
-        ppg_signal : numpy.ndarray
+        ppg_signal : NDArray
             Input photoplethysmography (PPG) signal.
         hr_range : tuple of float, optional
             Expected heart rate range in beats per minute (BPM). Default is (40.0, 180.0).
@@ -1636,9 +1636,9 @@ class HeartRateExtractor:
         frequency : float or None
             Dominant frequency in Hz. Returns None if no valid peak is found in the
             specified range.
-        psd : numpy.ndarray
+        psd : NDArray
             Power spectral density values corresponding to the frequency array.
-        freqs : numpy.ndarray
+        freqs : NDArray
             Frequency values corresponding to the power spectral density.
 
         Notes

rapidtide/qualitycheck.py

@@ -226,9 +226,9 @@ def gethistmetrics(
 
     Parameters
     ----------
-    themap : numpy.ndarray
+    themap : NDArray
         The input data array from which metrics are computed.
-    themask : numpy.ndarray
+    themask : NDArray
         A boolean or numeric mask array used to select valid data points.
     thedict : dict[str, Any]
         Dictionary to store computed metrics. Keys will be updated with statistical

rapidtide/refinedelay.py

@@ -20,7 +20,7 @@ from typing import Any, Callable, Optional, Tuple, Union
 
 import numpy as np
 import numpy.polynomial.polynomial as poly
-from numpy.typing import ArrayLike, NDArray
+from numpy.typing import NDArray
 from scipy.interpolate import CubicSpline, UnivariateSpline
 from scipy.ndimage import median_filter
 from statsmodels.robust import mad
@@ -101,7 +101,7 @@ def trainratiotooffset(
     ----------
     lagtcgenerator : Any
         An object that provides the `yfromx` method for generating lagged time courses.
-    timeaxis : numpy.ndarray
+    timeaxis : NDArray
         The time axis (in seconds) for the fMRI data.
     outputname : str
         Base name for output files (e.g., BIDS entity).
@@ -555,6 +555,7 @@ def getderivratios(
     TimingLGR: Optional[Any],
     optiondict: dict,
     regressderivs: int = 1,
+    timemask: Optional[NDArray] = None,
     starttr: Optional[int] = None,
     endtr: Optional[int] = None,
     debug: bool = False,
@@ -611,6 +612,8 @@ def getderivratios(
         Dictionary of options for regression and processing.
     regressderivs : int, optional
         Number of derivative regressors to include (default is 1).
+    timemask : NDArray, optional
+       Mask of timepoints to include in regression filtering.
     starttr : Optional[int], optional
         Start timepoint for processing (default is 0).
     endtr : Optional[int], optional
@@ -672,6 +675,11 @@ def getderivratios(
         print(f"\t{starttr=}")
         print(f"\t{endtr=}")
 
+    if timemask is not None:
+        trimmedtimemask = timemask[starttr:endtr]
+    else:
+        trimmedtimemask = None
+
     voxelsprocessed_regressionfilt, regressorset, evset = tide_regressfrommaps.regressfrommaps(
         fmri_data_valid[:, starttr:endtr],
         validvoxels,
@@ -701,6 +709,7 @@ def getderivratios(
         showprogressbar=optiondict["showprogressbar"],
         alwaysmultiproc=optiondict["alwaysmultiproc"],
         coefficientsonly=True,
+        timemask=trimmedtimemask,
         debug=debug,
     )
 
@@ -726,7 +735,7 @@ def filterderivratios(
     patchthresh: float = 3.0,
     gausssigma: float = 0,
     filetype: str = "nifti",
-    rt_floattype: str = "float64",
+    rt_floattype: np.dtype = np.float64,
     verbose: bool = True,
     debug: bool = False,
 ) -> Tuple[NDArray, NDArray, float]: