[Paper Review] Bayesian Inference Applied to the Electromagnetic Inverse Problem

David Schmidt, John George+1 more|ArXiv.org|Jun 16, 1998

Image and Signal Denoising Methods17 references4 citations

TL;DR

This paper introduces a Bayesian inference framework to address the ill-posed electromagnetic inverse problem in magnetoencephalography (MEG), generating a distribution of plausible neural activation solutions rather than a single estimate. By combining prior knowledge with measured data within a unified probabilistic model, it identifies highly probable features—such as location, extent, and number of active brain regions—demonstrated using both simulated and real visual evoked response MEG data.

ABSTRACT

We present a new approach to the electromagnetic inverse problem that explicitly addresses the ambiguity associated with its ill-posed character. Rather than calculating a single ``best'' solution according to some criterion, our approach produces a large number of likely solutions that both fit the data and any prior information that is used. While the range of the different likely results is representative of the ambiguity in the inverse problem even with prior information present, features that are common across a large number of the different solutions can be identified and are associated with a high degree of probability. This approach is implemented and quantified within the formalism of Bayesian inference which combines prior information with that from measurement in a common framework using a single measure. To demonstrate this approach, a general neural activation model is constructed that includes a variable number of extended regions of activation and can incorporate a great deal of prior information on neural current such as information on location, orientation, strength and spatial smoothness. Taken together, this activation model and the Bayesian inferential approach yield estimates of the probability distributions for the number, location, and extent of active regions. Both simulated MEG data and data from a visual evoked response experiment are used to demonstrate the capabilities of this approach.

Motivation & Objective

• To resolve the inherent ambiguity and ill-posed nature of the electromagnetic inverse problem in MEG and EEG by moving beyond single-solution estimates.
• To incorporate diverse prior information—such as spatial smoothness, location, orientation, and strength—of neural currents into the solution process.
• To quantify uncertainty in neural source localization by generating a distribution of likely solutions rather than a deterministic estimate.
• To identify features common across many sampled solutions as those with high posterior probability, enhancing reliability of inference.
• To demonstrate the method’s effectiveness using both simulated MEG data and real visual evoked response data from human subjects.

Proposed method

• Formalizes the inverse problem within a Bayesian framework, using Bayes' theorem to combine prior distributions with likelihood of measured data.
• Employs a flexible neural activation model that allows variable numbers of extended active regions with tunable spatial and directional constraints.
• Uses Markov Chain Monte Carlo (MCMC) sampling to explore the posterior distribution over possible source configurations.
• Incorporates prior information on current strength, orientation, and spatial smoothness through appropriate prior distributions.
• Applies the framework to both synthetic data and real MEG recordings from a visual stimulation experiment.
• Quantifies the probability distribution over the number, location, and extent of active brain regions based on the sampled posterior.

Experimental results

Research questions

• RQ1 How can prior knowledge about neural current distribution be systematically integrated into the solution of the electromagnetic inverse problem?
• RQ2 What is the distribution of plausible neural source configurations that are consistent with measured MEG data and prior constraints?
• RQ3 Which features of neural activation (e.g., location, extent) are robust across multiple sampled solutions and thus highly probable?
• RQ4 How does the Bayesian approach compare to traditional single-solution methods in terms of uncertainty quantification and reliability?
• RQ5 Can the method reliably recover known activation patterns in both simulated and real MEG data under realistic noise conditions?

Key findings

• The Bayesian approach successfully generated a distribution of plausible neural activation configurations that fit both simulated and real MEG data while incorporating prior knowledge.
• Features common across a large fraction of sampled solutions—such as the location and extent of activation—were identified with high posterior probability, indicating robust inference.
• The method provided quantitative estimates of the probability distribution over the number of active brain regions, enabling uncertainty quantification in source count.
• In the visual evoked response experiment, the approach correctly localized activation in the visual cortex with high confidence, consistent with known neuroanatomy.
• The integration of spatial smoothness and directional priors significantly improved the stability and interpretability of the inferred source distributions.
• The framework demonstrated that even with ill-posed data, meaningful probabilistic inference is possible when prior information is properly encoded.