Spinal Cord fMRI: How PCA Helps Clean Up the Noise

Researchers have tested a method that uses principal component analysis (PCA) to filter unwanted signals from spinal cord fMRI scans. The technique, called SpinalCompCor, picks out noise by looking at a region outside the spinal cord and cerebrospinal fluid. It then keeps only the most important components—usually about nine—to remove physiological interference like blood flow changes.

Study Overview

• Four studies examined the method:
• Two with motor tasks
• One with breathing exercises
• Two resting‑state scans

In each case, the PCA approach could explain a significant portion of background noise. However, when scientists compared brain activity maps from groups of participants, adding these PCA regressors did not noticeably improve the results.

Why It Might Not Work as Expected

• Some PCA components overlap with signals related to the actual tasks, which can make cleaning less effective.
• This overlap may be more pronounced when imaging settings differ or data quality varies.

Recommendations

Experts advise using SpinalCompCor only when actual physiological recordings (e.g., heart or breathing traces) are missing. The PCA method may not consistently match the precision of recordings across different experiments or equipment setups.

Bottom line: While PCA can help reduce noise in spinal cord imaging, its benefits are limited and must be weighed against potential complications.