DeepNavNet: Automated Landmark Localization for Neuronavigation.

Christine A Edwards,Aaron E Rusheen,Abbas Z Kouzani,Abhinav Goyal,Kendall H Lee

doi:10.3389/fnins.2021.670287

Christine A Edwards, Aaron E Rusheen + Show 3 more

Open Access

https://doi.org/10.3389/fnins.2021.670287

Copy DOI

Abstract

Functional neurosurgery requires neuroimaging technologies that enable precise navigation to targeted structures. Insufficient image resolution of deep brain structures necessitates alignment to a brain atlas to indirectly locate targets within preoperative magnetic resonance imaging (MRI) scans. Indirect targeting through atlas-image registration is innately imprecise, increases preoperative planning time, and requires manual identification of anterior and posterior commissure (AC and PC) reference landmarks which is subject to human error. As such, we created a deep learning-based pipeline that consistently and automatically locates, with submillimeter accuracy, the AC and PC anatomical landmarks within MRI volumes without the need for an atlas. Our novel deep learning pipeline (DeepNavNet) regresses from MRI scans to heatmap volumes centered on AC and PC anatomical landmarks to extract their three-dimensional coordinates with submillimeter accuracy. We collated and manually labeled the location of AC and PC points in 1128 publicly available MRI volumes used for training, validation, and inference experiments. Instantiations of our DeepNavNet architecture, as well as a baseline model for reference, were evaluated based on the average 3D localization errors for the AC and PC points across 311 MRI volumes. Our DeepNavNet model significantly outperformed a baseline and achieved a mean 3D localization error of 0.79 ± 0.33 mm and 0.78 ± 0.33 mm between the ground truth and the detected AC and PC points, respectively. In conclusion, the DeepNavNet model pipeline provides submillimeter accuracy for localizing AC and PC anatomical landmarks in MRI volumes, enabling improved surgical efficiency and accuracy.

Highlights

Imaging technologies such as magnetic resonance imaging (MRI) provide visualization of brain structures that enable neurosurgeons to plan accurate and safe surgical trajectories (Edwards et al, 2017)
Automated Landmark Localization for Neuronavigation software (Edwards et al, 2018). This software allows surgeons to view the neuroimaging data to derive 3D coordinates of the brain target(s) to plan the surgical trajectory path(s). This approach has been predominantly used to implant electrodes for deep brain stimulation and to deliver focused ultrasonic ablation, both of which provide therapeutic relief of debilitating movement and psychiatric disorders (Elias et al, 2016; Edwards et al, 2017, 2018)
To create DeepNavNet, this application was adapted to train models to learn a mapping from MRI volumes to volumetric heatmaps concentrated around corresponding AC and PC anatomical landmark locations

Summary

INTRODUCTION

Imaging technologies such as magnetic resonance imaging (MRI) provide visualization of brain structures that enable neurosurgeons to plan accurate and safe surgical trajectories (Edwards et al, 2017). A random forest regression approach demonstrated increased accuracy from millimeters to submillimeter localization errors compared to atlas-based methods for detecting AC and PC points within a limited set of 100 T1-weighted MR volumes acquired uniformly from an institution (Liu and Dawant, 2015) These regression forests were built upon texture-based features (i.e., local binary patterns), which have largely been replaced by more robust deeply-learned features in the computer vision community (LeCun et al, 2015). The aforementioned deep learning based methods have demonstrated state-of-the-art and promising results for a variety of medical image modalities and applications Even so, these methods neither report submillimeter accuracy in localization of landmarks in neuroimaging data, nor report the localization of AC and PC points as required for targeted functional neurosurgery planning. We present the development and validation of a novel deep learning-based heatmap regression pipeline that is built upon an implementation of a 3D residual network (Liu et al, 2018) that demonstrates submillimeter accuracy in localization of the AC and PC landmarks within MRI volumes

MATERIALS AND METHODS

RESULTS

DISCUSSION

CONCLUSION

ETHICS STATEMENT