More comments in tsnr.py and README

cni · Sep 9, 2022 · f513f30 · f513f30
1 parent a38d02b
commit f513f30
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 4 deletions.
diff --git a/README.md b/README.md
@@ -9,7 +9,8 @@ Calculates the temporal SNR (tSNR) and signal fluctuation to noise ratio (SFNR)
 - Gear outputs:
  - tSNR map ``*_tsnr.nii.gz``, SFNR map ``*_sfnr.nii.gz``
  - Plot of the temporal drift of mean signal within ROI ``*_mean_signal.png``, plot of the temporal drift of center of mass ``*_cm_drift.png``, plot of Weisskoff analysis (radius of decorrelation) ``*_rdc.png``
- - JSON file ``*_results.json`` containing statistic results e.g. mean tSNR, SFNR in ROI
+ - JSON file ``*_results.json`` containing statistic results, e.g. mean tSNR, SFNR in ROI
+ - Other intermediate results
 
 The script ``tsnr.py`` can also be executed seperately to generate tSNR maps on 4D NIFTIs containing time series images:
 

diff --git a/tsnr.py b/tsnr.py
@@ -68,14 +68,14 @@
  num_tpoints = noise.shape[3]
  center_of_mass_t = np.zeros((3, num_tpoints))
  center_of_mass_drift = np.zeros((3, num_tpoints))
- # calculate the center of mass drift in mm
+ # calculate the center of mass drift in mm over the time series
  for t in range(num_tpoints):
  center_of_mass_t[:, t] = ndimage.measurements.center_of_mass(tseries[...,t])
  if t == 0:
  center_of_mass = list(int(s) for s in ndimage.measurements.center_of_mass(tseries[...,t]))
  else:
  center_of_mass_drift[:, t] = np.multiply(center_of_mass_t[:, t] - center_of_mass_t[:, 0], pixdim)
- # calculate the radius of decorrelation according to the Weisskoff analysis
+ # calculate the radius of decorrelation according to the Weisskoff analysis. Radius of correlation measures how much correlation there is among the adjacent voxels due to noise contamination. Smaller radius of correlation indicates less contamination
  roi_length = range(1, args.roi_size+1)
  roi_std_detrend = []
  for r in roi_length:
@@ -92,7 +92,7 @@
  res = roi_mean - poly(range(num_tpoints))
  roi_std_detrend.append(np.std(res))
  rdc = roi_std_detrend[0] / roi_std_detrend[args.roi_size-1]
- # signal intensity within ROI over the time series
+ # mean signal intensity within ROI over the time series. Temporal variation within 0.05% is preferred
  roi_signal_mean = roi_mean 
  roi_signal_mean_fitted = poly(range(num_tpoints))
  sfnr_center = np.mean(sfnr[np.where(roi_mask[:,:,:,0])])