Predicting Lithium Response with Functional Brain Imaging

Predicting Lithium Response with Brain Imaging

Peter ForsterBipolar Treatment, Testing Leave a Comment

Predicting lithium response with brain imaging may be an option in the future if the preliminary results of a study published in 2017 are confirmed.

Researchers from the University of Cincinnati used three tools to create a program that was remarkably successful at predicting which bipolar patients would respond to lithium.

20 first onset bipolar patients who received adequate trials of lithium and baseline scans using functional magnetic resonance imaging and proton magnetic resonance spectroscopy were used to create an algorithm for distinguishing between lithium responders and non-responders.

The researchers used artificial intelligence technology (specifically, genetic fuzzy tree design) which allows a computer to identify a pattern in complex data.

The resulting “Lithium Intelligent Agent” correctly classified those with responding to lithium with an accuracy of between 80 and 88% in training and validation trials.

Implications

The study suggests that it may be possible to predict who will respond to lithium using a pair of brain scans. Unfortunately, it may be quite a while before this results in a test that can be implemented outside a research setting since both of the functional brain imaging techniques used are quite expensive and proton magnetic resonance spectroscopy is not widely available.

What is perhaps more interesting is that it points to the fact that there are differences in brain functioning that can be seen using existing brain imaging technologies that are linked to lithium response.

References

Fleck DE, Ernest N, Adler CM, Cohen K, Eliassen JC, Norris M, Komoroski RA, Chu WJ, Welge JA, Blom TJ, DelBello MP, Strakowski SM. Prediction of lithium response in first-episode mania using the LITHium Intelligent Agent (LITHIA): Pilot data and proof-of-concept. Bipolar Disord. 2017 Jun;19(4):259-272. doi: 10.1111/bdi.12507. Epub 2017 Jun 2. PubMed PMID: 28574156; PubMed Central PMCID:  PMC5517343.