NEUROIMAGING AI FOR TMS TARGET LOCALISATION: EVIDENCE FOR INDIVIDUALISED COIL PLACEMENT IN DEPRESSION AND NEUROREHABILITATION

Abstract

Transcranial magnetic stimulation has moved steadily from a research instrument to a clinically established non-invasive treatment for conditions including pharmacoresistant major depression, obsessive-compulsive disorder and post-stroke motor impairment. Its effectiveness, however, is not uniform across patients, and a significant proportion of this variability traces back to imprecision in coil positioning. The standard anatomical approach — placing the coil over a scalp landmark estimated to overlie the left dorsolateral prefrontal cortex — frequently misses the functionally defined target by margins large enough to reduce therapeutic yield. Artificial intelligence applied to structural and functional MRI provides an alternative: automated, subject-specific identification of cortical targets based on connectivity fingerprints rather than population-average anatomy. A review of 31 studies published between 2017 and 2024 indicates that AI-guided targeting reduces positioning error by 35–60% relative to landmark methods, and that functional connectivity-guided DLPFC stimulation achieves antidepressant response rates of 58.4% versus 41.2% for anatomical targeting in controlled comparisons. Reinforcement learning algorithms for adaptive parameter optimisation have shown further gains in cortical excitability induction. The implications of these findings for TMS service development within Uzbekistan's neurological infrastructure are considered.