corrected inverse Fourier transform; mistake in using ifft caused abe…

…rrant convolution
ccnmaastricht · Jan 17, 2022 · 4d0b9cc · 4d0b9cc
1 parent b9c2ed0
commit 4d0b9cc
Show file tree

Hide file tree

Showing 4 changed files with 16 additions and 15 deletions.
diff --git a/code/python/cni_tlbx/HGR.py b/code/python/cni_tlbx/HGR.py
@@ -435,7 +435,7 @@ def __convolution__(self, x):
         kernel = np.append(self.hrf, np.zeros(n_samples))
         x = np.vstack((x, np.zeros((np.ceil(self.l_hrf / self.p_sampling).astype(int),
                                     self.n_features))))
-        x_conv = np.abs(
+        x_conv = np.real(
                     ifft(fft(x, axis=0) *
                          np.expand_dims(fft(kernel),
                                         axis=1), axis=0))

diff --git a/code/python/cni_tlbx/IRM.py b/code/python/cni_tlbx/IRM.py
@@ -110,7 +110,7 @@ def get_hrf(self):
         else:
             hrf = ifft(self.hrf_fft, axis = 0)[0:self.l_hrf]
 
-        return np.abs(hrf)
+        return np.real(hrf)
 
     def get_stimulus(self):
         '''
@@ -130,7 +130,7 @@ def get_timecourses(self):
             predicted timecourses
         '''
 
-        return np.abs(ifft(self.tc_fft, axis = 0)[0:self.n_samples, :])
+        return np.real(ifft(self.tc_fft, axis = 0)[0:self.n_samples, :])
 
     def set_hrf(self, hrf):
         '''
@@ -256,7 +256,7 @@ def mapping(self, data, threshold = 100, mask = []):
         if self.hrf_fft.ndim==1:
             tc = np.transpose(
                 zscore(
-                    np.abs(
+                    np.real(
                         ifft(self.tc_fft *
                              np.expand_dims(self.hrf_fft,
                                             axis = 1), axis = 0)), axis = 0))
@@ -287,7 +287,7 @@ def mapping(self, data, threshold = 100, mask = []):
 
                 tc = np.transpose(
                     zscore(
-                        np.abs(
+                        np.real(
                             ifft(self.tc_fft *
                                  np.expand_dims(self.hrf_fft[:, v],
                                                 axis = 1), axis = 0)), axis = 0))

diff --git a/code/python/cni_tlbx/gadgets.py b/code/python/cni_tlbx/gadgets.py
@@ -66,11 +66,12 @@ def two_gamma(timepoints):
         hemodynamic response function
     '''
 
-    hrf = (6 * timepoints**5 * np.exp(-timepoints)) / gamma(6) \
-                - 1 / 6 * (16 * timepoints ** 15 * np.exp(-timepoints)) \
+    hrf = (6 * (timepoints**5) * np.exp(-timepoints)) / gamma(6) \
+                - 1 / 6 * (16 * (timepoints ** 15) * np.exp(-timepoints)) \
                     / gamma(16)
     return hrf
 
+
 def size(x,length):
     '''
     Parameters
@@ -246,7 +247,7 @@ def convolve(self, x):
 
         x_fft = fft(np.vstack((x.reshape(1,-1), np.zeros((self.l_subsampled - 1, self.n_channels)))), axis=0)
         self.x_conv = np.vstack((self.x_conv, np.zeros((1, self.n_channels))))
-        self.x_conv[self.step:self.step + self.l_subsampled, :] += np.abs(ifft(
+        self.x_conv[self.step:self.step + self.l_subsampled, :] += np.real(ifft(
             x_fft * np.expand_dims(self.hrf_fft , axis=1), axis=0))
 
         return self.x_conv[self.step, :]

diff --git a/code/python/cni_tlbx/pRF.py b/code/python/cni_tlbx/pRF.py
@@ -116,7 +116,7 @@ def get_hrf(self):
         else:
             hrf = ifft(self.hrf_fft, axis=0)[0:self.l_hrf]
 
-        return np.abs(hrf)
+        return np.real(hrf)
 
     def get_stimulus(self):
         '''
@@ -141,7 +141,7 @@ def get_timecourses(self):
             predicted timecourses
         '''
 
-        return np.abs(ifft(self.tc_fft, axis=0)[0:self.n_samples, :])
+        return np.real(ifft(self.tc_fft, axis=0)[0:self.n_samples, :])
 
     def set_hrf(self, hrf):
         '''
@@ -287,7 +287,7 @@ def get_timecourses(self):
             predicted timecourses
         '''
 
-        return np.abs(ifft(self.tc_fft, axis=0)[0:self.n_samples, :])
+        return np.real(ifft(self.tc_fft, axis=0)[0:self.n_samples, :])
 
     def set_hrf(self, hrf):
         '''
@@ -511,7 +511,7 @@ def _mapping_slope(self, data, threshold, mask):
      if self.hrf_fft.ndim == 1:
          tc = np.transpose(
              zscore(
-                 np.abs(
+                 np.real(
                      ifft(self.tc_fft *
                           np.expand_dims(self.hrf_fft,
                                          axis=1), axis=0)), axis=0))
@@ -545,7 +545,7 @@ def _mapping_slope(self, data, threshold, mask):
 
              tc = np.transpose(
                  zscore(
-                     np.abs(
+                     np.real(
                          ifft(self.tc_fft *
                               np.expand_dims(self.hrf_fft[:, v],
                                              axis=1), axis=0)), axis=0))
@@ -620,7 +620,7 @@ def _mapping_sigma(self, data, threshold, mask):
      if self.hrf_fft.ndim == 1:
          tc = np.transpose(
              zscore(
-                 np.abs(
+                 np.real(
                      ifft(self.tc_fft *
                           np.expand_dims(self.hrf_fft,
                                          axis=1), axis=0)), axis=0))
@@ -653,7 +653,7 @@ def _mapping_sigma(self, data, threshold, mask):
 
              tc = np.transpose(
                  zscore(
-                     np.abs(
+                     np.real(
                          ifft(self.tc_fft *
                               np.expand_dims(self.hrf_fft[:, v],
                                              axis=1), axis=0)), axis=0))