Tristan van Doormaal

UMC Utrecht

In short, Tristan van Doormaals' research entails the following: 

AutoMatic Brain MRI SegmentaTion using artificial IntelligenCe to Augment assessment and surgery of maLignant tumors: The MISTICAL project

'In the Netherlands per year +/- 4500 patients undergo an operation for a malignant brain tumor. MISTICAL aims at creating fully automatic 3D brain MRI segmentation algorithms for these tumors using artificial intelligence. With the resulting segmentations, patients will understand their disease better. Thereby neurosurgeons can gain insight and prepare for surgery in 3D. Finally, during surgery, surgeons can be guided with holograms based on these segmentations. MISTICAL involves direct integration of the segmentation algorithms and holography in a cloud pipeline, linked to the hospital environment.'

 

Publications

• NnU-Net versus mesh growing algorithm as a tool for the robust and timely segmentation of neurosurgical 3D images in contrast-enhanced T1 MRI scans
  
  • PubMed, February 20, 2024
  • Click here to view the publication. 
• Accuracy of an nnUNet neural network for the automatic segmentation of intracranial aneurysms, their parent vessels and major cerebral arteries from magnetic resonance imaging-Time of flight (MRI-TOF)
  • AJNR (American journal of neuroradiology), November 22, 2024
  • Click here to view the publication.

 

Brief summary of progress / results January 2025

This year was focused on optimization of our previously developed deep learning-based segmentation models. This resulted in better 3D segmentations of brain tumors and surrounding anatomy for surgeons and patients. Integration wioth AR and cloud was finalysed. We started and finalized first results on automatic trajectory planning. We started 2 clinical studies:

1. a first in-human study assessing the registration accuracy of the holograms using augmented reality (AR) technology. The first results show that study probably will reach its primary endpoint meaning showing non-inferiority of AR hologram registration versus standard Brainlab registration.
2. a study assessing the value of three-dimensional holographic image visualization in patient education. First results show increase in information recall and high patient satisfaction.

Brief summary of progress / results January 2024 

We expanded our manual segmentations and we updated our deep learning (DL) trained algorithm with this set. Early integrations of the DL based segmentation models into the Cloud have been put in place, and have been producing successful segmentations in current clinical cases. The current focus in model integration lies in the standardization of the containers, so that future model updates and new models may be provided more easily. Furthermore, additional segmentation models have been trained on skull base, the optic chiasm, the circle of Willis, Edema and different functional brain regions using different MRI series.

To be able to create an augmented reality overlay and guide surgeries better, we use the models’ skin segmentations to locate critical anatomical landmarks. Early results are promising, with the model being able to locate the relevant anatomical landmarks accurately.

Lastly, we have started a patient study regarding the effect of 3D visualization of preoperative imaging on information recall and patient satisfaction with imaging. This study compares the use of conventional 2D imaging and 3D imaging on patient information recall and will continue in the next year.

Brief summary of progress / results January 2023

In the first year of the MISTICAL project, an initial model to automatically extract the skin, brain, ventricles and brain tumor from a brain scan was trained, tested and integrated into a cloud-based software package so that patients and physicians can directly view it holographically (in 3D). The quality and accuracy of the model exceed previously used models. Future project years will be focused on improving the model and testing its effect on patient care.