A cautionary note on the use of SIFT in pathological connectomes

We recently presented a hypothetical example1 that demonstrates paradoxical behavior of the spherical‐deconvolution informed filtering of tractograms (SIFT) method2 in pathology. The Letter to the Editor by Smith and colleagues3 contends that the paradoxical behavior of the SIFT method exemplified by our example can be remedied by: (i) connection density normalization and (ii) use of a more realistic number of streamlines. Here, we revisit our example and show that these two modifications do not alleviate the paradoxical conclusions that result when SIFT is used. Specifically, we demonstrate that the SIFT method augmented with the two modifications proposed by Smith and colleagues can cause paradoxical changes in connectivity, which lead to wrong conclusions about the location of pathological fibers. SIFT can cause normal fiber bundles without any pathology to show the same reduction in connectivity as pathological fibers. While we did not evaluate SIFT in our original inves-