Real-time noise cancellation for speech acquired in interactive functional magnetic resonance imaging studies

Abstract

Purpose: To present online scanner noise cancellation for speech acquired in functional magnetic resonance imaging (fMRI) studies.

Materials and methods: An online active noise cancellation method for speech acquired in fMRI studies was developed. The approach consists of two automated steps: 1) creation of an MR noise template in a short "test" fMRI scan; 2) application of the template for automatic recognition and subtraction of the MR noise from the acquired microphone signal during an fMRI study. The method was applied in an experimental paradigm where a subject and an investigator communicated in an interactive verbal generation task during fMRI.

Results: By applying online active noise cancellation, the quality of the subject's speech was substantially improved. The present approach was found to be flexible, reliable, and easy to implement, providing a method for fMRI studies that investigate the neural correlates of interactive speech communication.

Conclusion: Using online noise cancellation it is possible to improve the quality of acquired speech in fMRI. This approach may be recommended for interactive fMRI studies.