Functional electrophysiological imaging comes in many flavours – including electroencephalography, magnetoencephalography and magnetospinography – but all are prone to significant difficulties in the face of interference, which often occurs at magnitudes that are considerable in comparison to the signal-of-interest.

While many approaches exist to compensate for these disruptions, they typically rely on secondary measurements designed to record the properties of the interference – such as so-called "empty room" data – resulting in more complicated and often more time-consuming imaging procedures.

Kensuke Sekihara, from Tokyo Medical and Dental University, and colleagues, propose a new solution – the dual signal subspace projection (DSSP) – which can remove interference without the need for separate noise measurements. DSSP – which estimates the interference subspace as an intersection between two time-domain subspaces – is based on an existing, but different algorithm called temporal signal space separation (tSSS).

On a simplistic level, Sekihara explains, the algorithm works in two stages. First, it creates a pseudo sensor signal in which components coming from the source space (the heart, for example, in the case of cardiac imaging) are supressed. In this way, the pseudo sensor signal contains only the interference and sensor noise, and not the signal-of-interest.

"The second step looks for components that are highly correlated between the original sensor data and the pseudo sensor data," Sekihara explains, adding: "such components are identified as the interference". Once the interference signals have been identified, they can be statistically subtracted from the original image data.

Validating the algorithm

To demonstrate DSSP's effectiveness, the researchers used the algorithm to successfully remove interference from both simulated and real-life biomagnetic measurements. First, the researchers pitted DSSP against the stimulus-induced artefacts generated in spinal cord evoked field (SCEF) measurements of a healthy patient. Functional imaging can be a useful diagnostic tool for spinal cord disorders, where anatomical imaging is not always effective.

To demonstrate the versatility of DSSP, the researchers also applied their algorithm to a comparable problem found with magnetoencephalographic (MEG) recordings. When imaging patients with vagal nerve stimulators – which use electrical pulses to prevent epileptic seizures – interference from the stimulator implant and its wires can often completely obscure the signal of interest.

Overall, Sekihara says, "the algorithm is effective for removing overlapped interference in a wide variety of biomagnetic measurements".

Javier Escudero – an engineer from the University of Edinburgh, who was not involved in this study – calls the approach taken with DSSP "interesting", and notes the potential for the new algorithm to streamline clinical practice. "The most common biomagnetic measurement is MEG," he notes, adding: "The results of the new algorithm are promising … but further work is needed to demonstrate how well it can eliminate physiological artefacts in MEG. Until then, tSSS may still have the upper hand."

"[The researchers] have developed an interesting extension of the spatiotemporal signal space separation method for a variety of biomagnetic applications," comments Samu Taulu – developer of the tSSS method and a physicist at the University of Washington. "This work further strengthens the signal processing methodology aiming at enabling high-quality data in noisy conditions."

With their initial study complete, the researchers are now looking to apply the DSSP algorithm to the removal of eye-blink and cardiac artefacts also encountered in MEG imaging.

Related stories

• An easy-to-use foetal heartbeat monitor
• MR driven EIT eyes oncology, neurology
• CUBRIC: a focus on brain research
• MEG spots post-traumatic stress disorder