anatomical.select_rois applies the spatial high-pass twice (duplicated line since v1.0.0.1) #1270

detection/anatomical.py::select_rois contains the same statement twice:

if settings.get("highpass_spatial", 0):
    img = np.clip(normalize99(img), 0, 1)
    img -= gaussian_filter(img, diameter[1] * settings["highpass_spatial"])
    img -= gaussian_filter(img, diameter[1] * settings["highpass_spatial"])

v0.14.4 has the subtraction once. The second line high-passes the already high-passed image, i.e. a stronger filter than highpass_spatial implies, which changes the image Cellpose segments (and therefore the ROI set) for anatomical-mode users who set highpass_spatial. If two passes are intended it would help to say so in the parameter description; otherwise the duplicate line can be dropped.

Generated by Claude Code