Three-dimensional fluid-attenuated inversion recovery sequence for visualisation of subthalamic nucleus for deep brain stimulation in Parkinson's disease

Visualization of small brain nuclei with a high-spatial resolution, clinically available whole-body PET scanner

OBJECTIVE

To verify the visibility of physiological 18F-fluorodeoxyglucose (18F-FDG) uptake in nuclei in and around the brainstem by a whole-body (WB) silicon photomultiplier positron emission tomography (SiPM-PET) scanner with point-spread function (PSF) reconstruction using various iteration numbers.

METHODS

Ten healthy subjects (5 men, 5 women; mean age, 56.0 ± 5.0 years) who underwent 18F-FDG PET/CT using a WB SiPM-PET scanner and magnetic resonance imaging (MRI) of the brain including a spin-echo three-dimensional sampling perfection with application-optimized contrasts using different flip angle evolutions fluid-attenuated inversion recovery (3D-FLAIR) and a 3D-T1 magnetization-prepared rapid gradient-echo (T1-MPRAGE) images were enrolled. Each acquired PET image was reconstructed using ordered-subset expectation maximization (OSEM) with iteration numbers of 4, 16, 64, and 256 (subset 5 fixed) + time-of-flight (TOF) + PSF. The reconstructed PET images and 3D-FLAIR images for each subject were registered to individual T1-MPRAGE volumes using normalized mutual information criteria. For each MR-coregistered individual PET image, the pattern of FDG uptake in the inferior olivary nuclei (ION), dentate nuclei (DN), midbrain raphe nuclei (MRN), inferior colliculi (IC), mammillary bodies (MB), red nuclei (RN), subthalamic nuclei (STN), lateral geniculate nuclei (LGN), medial geniculate nuclei (MGN), and superior colliculi (SC) was visually classified into the following three categories: good, clearly distinguishable FDG accumulation; fair, obscure contour of FDG accumulation; poor, FDG accumulation indistinguishable from surrounding uptake.

RESULTS

Among individual 18F-FDG PET images with OSEM iterations of 4, 16, 64, and 256 + TOF + PSF, the iteration numbers that showed the best visibility in each structure were as follows: ION, MRN, LGN, MGN, and SC, iteration 64; DN, iteration 16; IC, iterations 16, 64, and 256; MB, iterations 64 and 256; and RN and STN, iterations 16 and 64, respectively. Of the four iterations, the 18F-FDG PET image of iteration 64 visualized FDG accumulation in small structures in and around the brainstem most clearly (good, 98 structures; fair, 2 structures).

CONCLUSIONS

A clinically available WB SiPM-PET scanner is useful for visualizing physiological FDG uptake in small brain nuclei, using a sufficiently high number of iterations for OSEM with TOF and PSF reconstructions.

Computational Neurosurgery in Deep Brain Stimulation

Computational methods and technologies are critical for neurosurgery in general and in deep brain stimulation (DBS) in particular. They increasingly inform every aspect of clinical DBS therapy, from presurgical planning and hardware implantation to postoperative adjustment of stimulation parameters. Computational methods also occupy a prominent position within the DBS research sphere, where they facilitate efforts to better understand DBS' underlying mechanisms and optimize and individualize its delivery. This chapter provides a high-level overview of the various computational tools and methods that have been applied to DBS. First, we discuss the invaluable contribution of computational neuroimaging (primarily magnetic resonance imaging) to DBS, targeting and the role of postoperative methods of image analysis-specifically, electrode localization, volume of activated tissue modeling, and sweet-spot mapping-in precisely localizing DBS' targets in the brain and discerning optimal treatment loci. We then address the growing field of connectomics, which leverages specific magnetic resonance imaging (MRI) sequences and post-acquisition processing algorithms to explore how DBS operates at the level of brain-wide networks. Next, the search for electrophysiological and imaging-based biomarkers of optimal DBS therapy is explored. We lastly touch on the incipient field of spatial characterization analysis and discuss the ongoing development of adaptive, closed-loop DBS systems.

Comparison of Image Quality and Radiation Dose Between Single-Energy and Dual-Energy Images for the Brain With Stereotactic Frames on Dual-Energy Cerebral CT

Background

Preoperative stereotactic planning of deep brain stimulation (DBS) using computed tomography (CT) imaging in patients with Parkinson's disease (PD) is of clinical interest. However, frame-induced metal artifacts are common in clinical practice, which can be challenging for neurosurgeons to visualize brain structures.

Objectives

To evaluate the image quality and radiation exposure of patients with stereotactic frame brain CT acquired using a dual-source CT (DSCT) system in single- and dual-energy modes.

Materials and Methods

We included 60 consecutive patients with Parkinson's disease (PD) and randomized them into two groups. CT images of the brain were performed using DSCT (Group A, an 80/Sn150 kVp dual-energy mode; Group B, a 120 kVp single-energy mode). One set of single-energy images (120 kVp) and 10 sets of virtual monochromatic images (50-140 keV) were obtained. Subjective image analysis of overall image quality was performed using a five-point Likert scale. For objective image quality evaluation, CT values, image noise, signal-to-noise ratio (SNR), and contrast-to-noise (CNR) were calculated. The radiation dose was recorded for each patient.

Results

The mean effective radiation dose was reduced in the dual-energy mode (1.73 mSv ± 0.45 mSv) compared to the single-energy mode (3.16 mSv ± 0.64 mSv) (p < 0.001). Image noise was reduced by 46-52% for 120-140 keV VMI compared to 120 kVp images (both p < 0.01). CT values were higher at 100-140 keV than at 120 kVp images. At 120-140 keV, CT values of brain tissue showed significant differences at the level of the most severe metal artifacts (all p < 0.05). SNR was also higher in the dual-energy mode 90-140 keV compared to 120 kVp images, showing a significant difference between the two groups at 120-140 keV (all p < 0.01). The CNR was significantly better in Group A for 60-140 keV VMI compared to Group B (both p < 0.001). The highest subjective image scores were found in the 120 keV images, while 110-140 keV images had significantly higher scores than 120 kVp images (all p < 0.05).

Conclusion

DSCT images using dual-energy modes provide better objective and subjective image quality for patients with PD at lower radiation doses compared to single-energy modes and facilitate brain tissue visualization with stereotactic frame DBS procedures.

Quantitative Analysis for the Delineation of the Subthalamic Nuclei on Three-Dimensional Stereotactic MRI Before Deep Brain Stimulation Surgery for Medication-Refractory Parkinson’s Disease

Delineation of the subthalamic nuclei (STN) on MRI is critical for deep brain stimulation (DBS) surgery in patients with Parkinson’s disease (PD). We propose this retrospective cohort study for quantitative analysis of MR signal-to-noise ratio (SNR), contrast, and signal difference-to-noise ratio (SDNR) of the STN on pre-operative three-dimensional (3D) stereotactic MRI in patients with medication-refractory PD. Forty-five consecutive patients with medication-refractory PD who underwent STN-DBS surgery in our hospital from January 2018 to June 2021 were included in this study. All patients had whole-brain 3D MRI, including T2-weighted imaging (T2WI), T2-weighted fluid-attenuated inversion recovery (FLAIR), and susceptibility-weighted imaging (SWI), at 3.0 T scanner for stereotactic navigation. The signal intensities of the STN, corona radiata, and background noise were obtained after placing regions of interest (ROIs) on corresponding structures. Quantitative comparisons of SNR, contrast, and SDNR of the STN between MR pulse sequences, including the T2WI, FLAIR, and SWI. Subgroup analysis regarding patients’ sex, age, and duration of treatment. We used one-way repeated measures analysis of variance for quantitative comparisons of SNR, contrast, and SDNR of the STN between different MR pulse sequences, and we also used the dependent t-test for the post hoc tests. In addition, we used Mann–Whitney U test for subgroup analyses. Both the contrast (0.33 ± 0.07) and SDNR (98.65 ± 51.37) were highest on FLAIR (all p < 0.001). The SNR was highest on SWI (276.16 ± 115.5), and both the SNR (94.23 ± 31.63) and SDNR (32.14 ± 17.23) were lowest on T2WI. Subgroup analyses demonstrated significantly lower SDNR on SWI for patients receiving medication treatment for ≥13 years (p = 0.003). In conclusion, on 3D stereotactic MRI of medication-refractory PD patients, the contrast and SDNR for the STN are highest on FLAIR, suggesting the optimal delineation of STN on FLAIR.

A literature review of magnetic resonance imaging sequence advancements in visualizing functional neurosurgery targets

OBJECTIVE

Historically, preoperative planning for functional neurosurgery has depended on the indirect localization of target brain structures using visible anatomical landmarks. However, recent technological advances in neuroimaging have permitted marked improvements in MRI-based direct target visualization, allowing for refinement of "first-pass" targeting. The authors reviewed studies relating to direct MRI visualization of the most common functional neurosurgery targets (subthalamic nucleus, globus pallidus, and thalamus) and summarize sequence specifications for the various approaches described in this literature.

METHODS

The peer-reviewed literature on MRI visualization of the subthalamic nucleus, globus pallidus, and thalamus was obtained by searching MEDLINE. Publications examining direct MRI visualization of these deep brain stimulation targets were included for review.

RESULTS

A variety of specialized sequences and postprocessing methods for enhanced MRI visualization are in current use. These include susceptibility-based techniques such as quantitative susceptibility mapping, which exploit the amount of tissue iron in target structures, and white matter attenuated inversion recovery, which suppresses the signal from white matter to improve the distinction between gray matter nuclei. However, evidence confirming the superiority of these sequences over indirect targeting with respect to clinical outcome is sparse. Future targeting may utilize information about functional and structural networks, necessitating the use of resting-state functional MRI and diffusion-weighted imaging.

CONCLUSIONS

Specialized MRI sequences have enabled considerable improvement in the visualization of common deep brain stimulation targets. With further validation of their ability to improve clinical outcomes and advances in imaging techniques, direct visualization of targets may play an increasingly important role in preoperative planning.

Deep Learning-Based Deep Brain Stimulation Targeting and Clinical Applications

Background

The purpose of the present study was to evaluate deep learning-based image-guided surgical planning for deep brain stimulation (DBS). We developed deep learning semantic segmentation-based DBS targeting and prospectively applied the method clinically.

Methods

T2^∗ fast gradient-echo images from 102 patients were used for training and validation. Manually drawn ground truth information was prepared for the subthalamic and red nuclei with an axial cut ∼4 mm below the anterior–posterior commissure line. A fully convolutional neural network (FCN-VGG-16) was used to ensure margin identification by semantic segmentation. Image contrast augmentation was performed nine times. Up to 102 original images and 918 augmented images were used for training and validation. The accuracy of semantic segmentation was measured in terms of mean accuracy and mean intersection over the union. Targets were calculated based on their relative distance from these segmented anatomical structures considering the Bejjani target.

Results

Mean accuracies and mean intersection over the union values were high: 0.904 and 0.813, respectively, for the 62 training images, and 0.911 and 0.821, respectively, for the 558 augmented training images when 360 augmented validation images were used. The Dice coefficient converted from the intersection over the union was 0.902 when 720 training and 198 validation images were used. Semantic segmentation was adaptive to high anatomical variations in size, shape, and asymmetry. For clinical application, two patients were assessed: one with essential tremor and another with bradykinesia and gait disturbance due to Parkinson’s disease. Both improved without complications after surgery, and microelectrode recordings showed subthalamic nuclei signals in the latter patient.

Conclusion

The accuracy of deep learning-based semantic segmentation may surpass that of previous methods. DBS targeting and its clinical application were made possible using accurate deep learning-based semantic segmentation, which is adaptive to anatomical variations.

Accuracy of different three-dimensional subcortical human brain atlases for DBS -lead localisation

BACKGROUND

Accurate interindividual comparability of deep brain stimulation (DBS) lead locations in relation to the surrounding anatomical structures is of eminent importance to define and understand effective stimulation areas. The objective of the current work is to compare the accuracy of the DBS lead localisation relative to the STN in native space with four recently developed three-dimensional subcortical brain atlases in the MNI template space. Accuracy is reviewed by anatomical and volumetric analysis as well as intraoperative electrophysiological data.

METHODS

Postoperative lead localisations of 10 patients (19 hemispheres) were analysed in each individual patient based on Brainlab software (native space) and after normalization into the MNI space and application of 4 different human brain atlases using Lead-DBS toolbox within Matlab (template space). Each patient's STN was manually segmented and the relation between the reconstructed lead and the STN was compared to the 4 atlas-based STN models by applying the Dice coefficient. The length of intraoperative electrophysiological STN activity along different microelectrode recording tracks was measured and compared to reconstructions in native and template space. Descriptive non-parametric statistical tests were used to calculate differences between the 4 different atlases.

RESULTS

The mean STN volume of the study cohort was 153.3 ± 40.3 mm3 (n = 19). This is similar to the STN volume of the DISTAL atlas (166 mm3; p = .22), but significantly larger compared to the other atlases tested in this study. The anatomical overlap of the lead-STN-reconstruction was highest for the DISTAL atlas (0.56 ± 0.18) and lowest for the PD25 atlas (0.34 ± 0.17). A total number of 47 MER trajectories through the STN were analysed. There was a statistically significant discrepancy of the electrophysiogical STN activity compared to the reconstructed STN of all four atlases (p < .0001).

CONCLUSION

Lead reconstruction after normalization into the MNI template space and application of four different atlases led to different results in terms of the DBS lead position relative to the STN. Based on electrophysiological and imaging data, the DISTAL atlas led to the most accurate display of the reconstructed DBS lead relative to the DISTAL-based STN.