Home Blog Focused Ultrasound for Psychiatric Disorders: AI Improves Prediction of Capsulotomy Success

Focused Ultrasound for Psychiatric Disorders: AI Improves Prediction of Capsulotomy Success

Published:

Key Points

  • Researchers found that machine learning models can predict the success of some focused ultrasound treatments. 
  • Clinicians can use the new predictive tool for better patient selection. 

Focused Ultrasound Capsulotomy: Predicting the Probability of Successful Lesioning Based on Skull Morphology 

A recent study published in the Journal of Neurosurgery sheds light on how machine learning (ML) models can improve the technical success of using MR-guided focused ultrasound anterior capsulotomy (MRgFUS-AC) to treat patients with obsessive-compulsive disorder (OCD) and major depressive disorder. 

While the procedure offers significant benefits, clinical data have shown that up to 25% of patients who pass the initial screening for an approved skull density ratio (SDR) experience treatment failure related to patient-specific skull characteristics. 

The research team at Sunnybrook Health Sciences Centre and the University of Toronto led by Benjamin Davidson, MD, PhD, and Maged Goubran, PhD, analyzed 60 MRgFUS-AC treatments performed with the Insightec Exablate Neuro device between 2017 and 2024. From a total of 157 lesions, 31 were deemed treatment failures. The findings revealed that higher SDR, thinner skulls, and lower incident angles were strongly associated with better outcomes. Interestingly, other variables, such as the brain, cerebrospinal fluid, and head volumes, showed weak correlations with lesion success. 

Researchers then trained and evaluated 12 ML models on these factors to assess their ability to predict technical success. A logistic regression model achieved the highest overall mean F1 score (0.89 ± 0.04) and highest accuracy (0.81 ± 0.07), with perfect recall in the test cohort, indicating it correctly identified all successful lesions. The authors concluded that this prediction tool can be used to assist clinicians in evaluating patient suitability for MRgFUS-AC, and that it could be particularly valuable for patients with borderline SDR values, where treatment decisions are often more complex. 

By refining candidate selection with this ML model, the study offers a practical path to reducing failure rates and expanding access to an innovative psychiatric treatment. 

See the Journal of Neurosurgery 

Related Stories 

Canadian Hospital Now Offering Focused Ultrasound Treatment for OCD to International Patients February 2025 

Monica’s Story: Patient with OCD Shares her Journey with Focused Ultrasound 2025 

Focused Ultrasound Durable for Obsessive-Compulsive Disorder November 2024 

Long-Term Clinical Results: Focused Ultrasound Safe and Effective for Obsessive-Compulsive Disorder (OCD), Depression October 2024 

Meet Anya: OCD Patient Finds Peace with Focused Ultrasound August 2024 

Optimizing Focused Ultrasound Capsulotomy for Obsessive-Compulsive Disorder (OCD) July 2021 

Focused Ultrasound for Psychiatric Disorders: Clinical Trial Results Published May 2020 

Meet Jeffrey: The First Patient in a Clinical Trial for OCD November 2019 

Focused Ultrasound Clinical Trial Begins for Patients with Obsessive-Compulsive Disorder September 2017