Machine learning—aided personalized DTI tractographic planning for deep brain stimulation of the superolateral medial forebrain bundle using HAMLET

Volker A. Coenen, Thomas E. Schlaepfer, Peter C. Reinacher, Hansjoerg Mast, Horst Urbach, Marco Reisert

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Background: Growing interest exists for superolateral medial forebrain bundle (slMFB) deep brain stimulation (DBS) in psychiatric disorders. The surgical approach warrants tractographic rendition. Commercial stereotactic planning systems use deterministic tractography which suffers from inherent limitations, is dependent on manual interaction (ROI definition), and has to be regarded as subjective. We aimed to develop an objective but patient-specific tracking of the slMFB which at the same time allows the use of a commercial surgical planning system in the context of deep brain stimulation. Methods: The HAMLET (Hierarchical Harmonic Filters for Learning Tracts from Diffusion MRI) machine learning approach was introduced into the standardized workflow of slMFB DBS tractographic planning on the basis of patient-specific dMRI. Rendition of the slMFB with HAMLET serves as an objective comparison for the refinement of the deterministic tracking procedure. Our application focuses on the tractographic planning of DBS (N = 8) for major depression and OCD. Results: Previous results have shown that only fibers belonging to the ventral tegmental area to prefrontal/orbitofrontal axis should be targeted. With the proposed technique, the deterministic tracking approach, that serves as the surgical planning data, can be refined, over-sprouting fibers are eliminated, bundle thickness is reduced in the target region, and thereby probably a more accurate targeting is facilitated. The HAMLET-driven method is meant to achieve a more objective surgical fiber display of the slMFB with deterministic tractography. Conclusions: The approach allows overlying the results of patient-specific planning from two different approaches (manual deterministic and machine learning HAMLET). HAMLET shows the slMFB as a volume and thus serves as an objective tracking corridor. It helps to refine results from deterministic tracking in the surgical workspace without interfering with any part of the standard software solution. We have now included this workflow in our daily clinical experimental work on slMFB DBS for psychiatric indications.

Original languageEnglish (US)
Pages (from-to)1559-1569
Number of pages11
JournalActa Neurochirurgica
Issue number8
StatePublished - Aug 1 2019


  • Brain
  • Deep brain stimulation
  • Depression
  • Machine learning
  • Medial forebrain bundle
  • Obsessive-compulsive disorder
  • Stereotaxy

ASJC Scopus subject areas

  • Surgery
  • Clinical Neurology


Dive into the research topics of 'Machine learning—aided personalized DTI tractographic planning for deep brain stimulation of the superolateral medial forebrain bundle using HAMLET'. Together they form a unique fingerprint.

Cite this