Role of Diffusion MRI Tractography in Endoscopic Endonasal Skull Base Surgery

Summary

We present a protocol to integrate diffusion MRI tractography in patient work-up to endoscopic endonasal surgery for a skull base tumor. The methods for adopting these neuroimaging studies in the pre- and intra-operative phases are described.

Abstract

Endoscopic endonasal surgery has gained a prominent role in the management of complex skull base tumors. It allows the resection of a large group of benign and malignant lesions through a natural anatomical extra-cranial pathway, represented by the nasal cavities, avoiding brain retraction and neurovascular manipulation. This is reflected by the patients' prompt clinical recovery and the low risk of permanent neurological sequelae, representing the main caveat of conventional skull base surgery. This surgery must be tailored to each specific case, considering its features and relationship with surrounding neural structures, mostly based on preoperative neuroimaging. Advanced MRI techniques, such as tractography, have been rarely adopted in skull base surgery due to technical issues: lengthy and complicated processes to generate reliable reconstructions for inclusion in the neuronavigation system.

This paper aims to present the protocol implemented in the institution and highlights the synergistic collaboration and teamwork between neurosurgeons and the neuroimaging team (neurologists, neuroradiologists, neuropsychologists, physicists, and bioengineers) with the final goal of selecting the optimal treatment for each patient, improving the surgical results and pursuing the advancement of personalized medicine in this field.

Introduction

The possibility to approach the skull base midline and paramedian regions through an anterior route, adopting the nasal fossae as natural cavities, has a long history, dating back more than one century¹. However, in the last 20 years, the visualization and operative technologies have improved enough to expand their possibility of including the treatment of the most complex tumors such as meningiomas, chordomas, chondrosarcomas, and craniopharyngiomas¹ due to the (1) introduction of the endoscope, which gives a panoramic and detailed 2D/3D view of these regions to the surgeon, (2) the development of intraoperative neuronavigation systems, and (3) the implementation of dedicated surgical instruments. As painstakingly demonstrated by Kassam et al. and confirmed by multiple reviews and meta-analyses, the advantages of this surgical approach are mainly represented by its chances to resect challenging skull base tumors, avoiding any direct brain retraction or nerve manipulation, thus reducing the risk of surgical complications and long-term neurological and visual sequelae^{2,3,4,5,6,7,8,9,10,11,12}.

For multiple skull base and pituitary-diencephalic tumors, the ideal surgical goal has changed in the last years from the most extensive tumor removal possible to the safest removal with preservation of the neurological functions to preserve the patient's quality of life³. This limitation could be compensated by innovative and effective adjuvant treatments, such as radiation therapy (adopting massive particles such as proton or carbon ions when appropriate) and, for selected neoplasms, by chemotherapy as inhibitors of the BRAF/MEK pathway for the craniopharyngiomas^13,14,15.

However, to pursue these goals, a careful preoperative assessment is crucial, to tailor the surgical strategy to each case's specific feature². In most centers, the MRI preoperative protocol is usually performed only with standard structural sequences, which provide the morphological characterization of the lesion. However, with these techniques it is not always possible to assess the anatomical relationship of the tumor with adjacent structures reliably³. Moreover, each patient may present different pathology-induced functional reorganization profiles detectable only with diffusion MRI tractography and functional MRI (fMRI), which can be used to provide guidance both in the surgery planning and in the intraoperative steps^16,17.

Currently, fMRI is the most commonly used neuroimaging modality for mapping brain functional activity and connectivity, as guidance for surgical planning^18,19 and to improve the patients' outcome²⁰. Task-based fMRI is the modality of choice to identify "eloquent" brain regions that are functionally involved in specific task performance (e.g., finger tapping, phonemic fluency), but is not applicable for the study of skull base tumors.

Diffusion MRI tractography permits in vivo and noninvasive reconstruction of white matter brain connections as well as cranial nerves, investigating the brain hodological structure²¹. Different tractography algorithms have been developed to reconstruct axonal pathways by linking water molecule diffusivity profiles, evaluated within each brain voxel. Deterministic tractography follows the dominant diffusivity direction, whereas probabilistic tractography evaluates possible pathways' connectivity distribution. Additionally, different models can be applied to evaluate diffusivity within each voxel, and it is possible to define two main categories: single fiber models, such as the diffusion tensor model, where a single fiber orientation is evaluated, and multiple-fiber models, such as spherical deconvolution, where several crossing-fiber orientations are reconstructed^22,23. Despite the methodological debate about diffusion MRI tractography, its utility in the neurosurgical workflow is currently established. It is possible to evaluate white matter tract dislocation and distance to the tumor, preserving specific white matter connections. Moreover, diffusion tensor imaging (DTI) maps, especially fractional anisotropy (FA) and mean diffusivity (MD), can be applied to assess microstructural white matter alterations related to possible tumor infiltration and for longitudinal tract monitoring. All these features make diffusion MRI tractography a powerful tool both for pre-surgical planning and intra-operative decision making through neuronavigation systems²⁴.

However, the application of tractography techniques to skull base surgery has been limited by the need for specialized technical knowledge and the time-consuming work-up to optimize diffusion MRI sequence acquisition, analysis protocol, and incorporating tractography results in neuronavigation systems²⁵.Finally, further limitations are due to the technical difficulties extending these analyses from intraparenchymal to extra-parenchymal white matter structures, as cranial nerves. Indeed, only recent studies presented preliminary results attempting to integrate advanced MRI and skull base surgery^26,27,28.

The present paper presents a protocol for the multidisciplinary management of pituitary-diencephalic and skull base tumors using diffusion MRI tractography. The implementation of this protocol in the institution resulted from the collaboration between neurosurgeons, neuro-endocrinologists and the neuroimaging team (including clinical and bioinformatics expertise) to offer an effective integrated multi-axial approach to these patients.

In the center, we have integrated multidisciplinary protocols for managing patients with skull base tumors, to provide the most informative description possible, and to tailor and personalize the surgical plan. We show that this protocol can be adopted both in the clinical and the research setting for any patient with a skull base tumor to guide the treatment strategy and to improve the knowledge on the brain modifications induced by these lesions.

Protocol

The protocol is following the Local Research Committee's ethical standards and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

1. Selection of the patients

1. Adopt the following inclusion criteria: patients older than 18 years old, fully collaborating, presenting a tumor of the skull base, or pituitary-diencephalic region.
2. Exclude patients with contraindication to MRI (i.e., a pacemaker or ferromagnetic material) or presenting with emergent clinical conditions (i.e., intracranial hypertension, the acute visual loss that requires immediate surgery), or pregnant women, or patients with mental illness, or those who explicitly refuse to participate in this protocol.

2. Preparation for the MRI exam

1. Before the MRI exam, administer the safety form to exclude significant contraindication to the exam and contrast agent injection: no ferromagnetic materials in the body, evaluation of MRI devices, safe or conditional, no pacemaker, no eye contact lenses on.
2. If the scanner used for the MRI acquisition is a high field (e.g., 3 T, see Table of Materials), consider any potential additional contraindications related, for example, to neurostimulation devices.
3. Check whether the patient has claustrophobia.
4. Ensure that the patient has read and signed the MRI consent form to acknowledge the imaging exam's risks and benefits.
5. Have a neuropsychologist perform a general evaluation and a targeted neurocognitive assessment based on the tumor location.
6. Administer the Edinburgh inventory to evaluate handedness dominance²⁹.

3. Positioning of the patient in the scanner

1. Give earplugs to the patient to reduce MRI noise.
2. Head movements can affect imaging quality; thus, use foam pads to reduce head movements, immobilizing the head inside the MRI coil.
3. Provide an emergency alarm button to the patient in case of need to interrupt the exam.
4. Switch on the camera and microphone inside the scanner to monitor, speak, and listen to the patient from the MRI acquisition room outside the scanner.

4. Brain MRI protocol setting and acquisition parameters

1. Acquire a standardized multimodal MRI protocol high-field scanner (1.5 T or 3T). The following sequence parameters refer to a 3 T MRI, using a head-neck high-density array coil (64 channels).
2. Acquire high-resolution and volumetric anatomical sequences: T1-weighted pre- and post-gadolinium contrast agent administration and FLAIR T2-weighted.
3. For T1 and T2 weighted images acquire continuous sagittal slices providing isotropic resolution of 1x1x1 mm³ scanning time of about 5 min per sequence.
4. Acquire a high-resolution T2-weighted sequence and localize the tumor area for cranial nerve visualization: a volumetric CISS (Constructive Interference in Steady State) with voxel dimension of 0.5x0.5x0.5 mm³ (scanning time of about 9 minutes).
5. Acquire diffusion-weighted sequences using single-shot echo-planar images (EPI), voxel dimension of 2x2x2 mm³, 64 magnetic gradient directions with b-value of 2000 s/mm², echo time of 98 ms, and relaxation time of 4300 ms.
6. Acquire five volumes with null b-value at the beginning of the diffusion-weighted acquisition with phase encoding direction set to anterior-posterior (for diffusion weighted images total scanning time of 5 minutes).
7. Additionally, acquire three volumes with null b-value but reversed phase encoding direction, posterior-anterior, to correct imaging distortions due to the EPI acquisition (scanning time of 42 seconds). Continuous near-axial slices are acquired.
8. Acquire additional sequences to investigate specific tumor features, such as multi- or single-voxel MRI-spectroscopy localized in the tumor area.
   ​NOTE: The total scanning time duration is about 30 minutes, excluding patient preparation for the MRI exam.

5. Brain MR images pre-processing

1. Convert the MRI data from the imaging format adopted by MRI acquisition consoles, DICOM (.dcm), to the NIFTI format (.nii) used in advanced imaging analyses.
2. Run the dcm2niix function (https://github.com/rordenlab/dcm2niix). Set as input files dicom images and as output the corresponding .nii files: T1.nii, Flair.nii, T1_contrast.nii, DTI_b2000.nii and DTI_b0_flip.nii.
3. Install the FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki) and MRtrix3 (https://www.mrtrix.org) software needed for the advanced imaging analyses.
4. Register the Flair.nii and T1_contrast.nii to the T1.nii image by running the FSL-flirt function, which performs a linear image registration.
5. Register the DTI_b2000.nii image to the T1.nii by running the FSL-epi_reg function, which takes into account EPI imaging distortion artifacts.
6. Run the FSL-topup function to correct phase encoding direction artifacts presenting the DTI_b2000.nii image. Set the DTI_b0_flip.nii inverse phase encoding acquisition as the "in_main" input file.
7. Run the MRtrix3-dwidenoise function for imaging denoising with a principal component noise modeling.
8. To correct eddy current and signal drop-out artifact, run the FSL-eddy function, and for MRI coil-induced signal inhomogeneities, the MRtrix3-dwibias correct function.
9. Run the FSL-bet function to remove the scalp signal presenting the T1.nii image and rename the output file by using the "_brain" suffix: T1_brain.nii.

6. Tumor segmentation

1. Install the itk-snap software (http://www.itksnap.org) ³⁰.
2. Once the itk-snap software is installed, press File - Open Main Image and select the T1.nii image, then press File - Add Another Imager and upload the Flair.nii and T1_contrast.nii images, setting the semi-transparent overlay option.
3. Inspect the tumor in the T1.nii, Flair.nii, and T1_contrast.nii images. Choose the anatomical plane to follow when drawing the lesion, e.g., axial.
4. Place the pointer in one axial slice to start. In the Main Toolbar, select the Polygon Inspector icon and start drawing tumor boundaries by using the Freehand Drawing Style - Smooth curve or Polygon.
5. Once finished drawing the tumor perimeter, close the curve linking the first and last dots, press Accept, and continue drawing in the next slice. For large tumor lesions, to accelerate the drawing process, skip some axial slices (e.g., three), and draw the lesion perimeter in interleaved slices.
6. At the end of the lesion perimeter drawing, select Tools - Interpolate Labels, set the Label to/with interpolate as the tumor lesion and the Interpolate along a single axis as the axis orientation followed in drawing the tumor boundaries.
7. Select Segmentation - Save Segmentation Image and name the tumor segmentation as Tumor_mask.nii by selecting the Nifti format option to save.

7. Tractography analysis

1. Run the FSL-dtifit function to model diffusivity and the different spatial directions and obtain the following diffusion tensor maps: FA.nii, MD.nii, and V1.nii. Evaluate these DTI maps to access abnormal diffusivity values that may occur in the presence of tumor edema or infiltration.
2. Run the MRtrix3-tckgen function with the default setting "ifod2" to perform probabilistic tractography and reconstruct the white matter pathways by modeling crossing-fibers issues³¹.
3. Adopt a seed-target approach by setting the "-seed_image" and "-include" options based on a priori anatomical knowledge.
4. Manually draw regions of interest (ROIs) set as seed or target for tractography. Alternatively, use atlas-based ROIs. See Mormina et at.³² for the optic radiation tractography, Hales et al.³³ for the optic chiasm and optic cranial nerves, and Testa et al.³⁴ for the pyramidal tracts.
5. Launch the FSL-fsleyes image viewer, select Open, and choose images to inspect visually.
6. In the FSL- fsleyes viewer, go to Setting - Ortho View 1 and activate the Edit Mode tool.
7. Click the FSL-fsleyes pencil icon and draw the tractography ROIs.
8. Install the Freesurfer (https://surfer.nmr.mgh.harvard.edu) software.
9. Run the Freesurfer-Recon-all function on the T1.nii image to obtain the automatic cortical region segmentation to use as tractography ROIs.
10. Run the FSL-epi_regregistration function, setting as input image the T1.nii, and reference image the DTI_b2000.nii, save the registration output matrix (T1_onto_DTI.mat).
11. Use the obtained T1_onto_DTI.mat matrix to register the segmented ROIs to the DTI_b2000.nii image.
12. Run the tractography using the MRtrix3-tckgen function.
13. Run the MRtrix3-tckmap function to convert the ".tck" streamlines tractography output in the "-template FA.nii" image.
14. Run the FSL-flirt function to linearly register the T1.nii image to the MNI152_T1_2mm_brain.nii template.
15. Save the output matrix as T1_onto_MNI.mat. Run the FSL-convert_xfm function setting the "-concat" option as T1_onto_MNI.mat and T1_onto_DTI.mat, save the output matrix as DTI_onto_MNI.mat.

8. Tractography: along-tract analysis

1. For an accurate description of DTI parameters, use along-tract algorithms, such as the Matlab-based algorithm that models the surface tract geometry with the Laplacian operator properties³⁵.
2. Install the Matlab software (https://matlab.mathworks.com) and request the along-tract code to the developing authors³⁵.
3. Alternatively, use the MRtrix3-tcksample function for along-tract analysis as Matlab requires a license.

9. 3D-rendering visualization

1. Install the Surf Ice software (https://www.nitrc.org/plugins/mwiki/index.php/surfice:MainPage).
2. In the Surf Ice command panel, click on Advanced - Convert voxelwise to mesh, select the nifti image to convert, save the resulting .obj file.
3. In the Surf Ice command panel, click File - Open, and select the .obj file to visualize the 3D volume rendering.

10. Preoperative clinical examinations

1. Perform bio-humoral endocrinological assessment, consisting of prolactin, TSH, freeT4, ACTH, cortisol, GH, LH, FSH, and serum tests total testosterone/estradiol, respectively in men and women.
2. Analyze the 24-hour urine volume and serum and urine osmolality and sodium levels to determine the presence of diabetes insipidus.
3. Perform an ophthalmological evaluation, including visual acuity measurement, computerized visual field assessment, and retinal optical coherence tomography (OCT).
4. Perform a neurological physical examination, with a collection of anamnestic information about weight gain, the sensation of hunger, continuously monitoring rectal temperature every 2 min for 24 h using a portable device to evaluate the circadian temperature rhythm, and 24 h sleep-wake cycle recording (including an electroencephalogram, right and left electro-oculogram, electrocardiogram, and electromyogram of mylohyoid and left and right tibialis muscles)^36,37,38.

11. Surgical planning

1. Discuss in a collegial team meeting each patient candidate to surgery, based on the results of tumor segmentation and relationship with the functional eloquent neural structures (optic nerves and chiasm, pituitary stalk, third ventricle, internal carotid artery, anterior cerebral artery-anterior communicating artery (ACA-ACoA) complex, basilar artery, cranial nerves III, IV, VI, mammillary bodies, white matter tracts, and functional cortical areas) to determine the most appropriate surgical approach.
2. Select the surgical corridor with the minimal risk of injuries of neural structures³⁹.
3. Define the safe resection area for each case, localizing the critical neural structure (such as chiasm, mammillary body) under whose proximity the resection must be arrested to avoid permanent damage³⁹.
4. Merge the most relevant MRI sequences and import them into the operative phase's neuronavigation system.

12. Surgery preparation

1. Induce general anesthesia adopting total intravenous anesthesia with propofol and remifentanil (it has been demonstrated that the other anesthetic agents are among the most critical factors affecting intraoperative monitoring reliability, increasing the false-negative rate), avoiding myorelaxant⁴⁰.
2. Perform oro-tracheal intubation with gauzes in the oropharynx to prevent blood or fluid leakage in the stomach or airways⁴¹.
3. Set up the neurophysiological monitoring, with continuous recording of motor evoked potentials (MEPs) and somatosensory evoked potentials (SEPs) and free-running electromyography (EMG) for cranial nerves.42
4. Import the MRI data, including the tractography reconstructions, in the neuronavigation system (Table of Materials).
5. Select the brain surgery electromagnetic registration modality on the neuronavigation system.
6. Register the neuronavigation system on the patient, adopting a free-tracking technique or external markers.
7. Control the accuracy of the achieved registration, checking the position of external markers (i.e., ear or nose) on the imported MRI; if the result is not acceptable, repeat the registration.
8. Place the patient in a semi-sitting position; Mayfield's use to fix the head is not needed⁴³.
9. Administer corticosteroid (endovenous flebocortid, dosage depending on the patient's weight) and antibiotics (2 g of amoxicillin-clavulanic acid)⁴⁴.

13. Endoscopic endonasal surgery

1. Start with a 0° endoscope (Table of Materials).
2. Harvest the naso-septal flap⁴⁵.
3. Perform an anterior sphenoidotomy, followed by posterior septostomy and ethmoidectomy with preservation of the middle turbinate, when possible⁴³.
4. Open the sellar and tuberculum bone⁴¹.
5. Incise the dura layer with an H-shape, after coagulation of the superior intercavernous sinus⁴¹.
6. Cleave the tumor by the arachnoidal plane⁴³.
7. Centrally debulk the tumor⁴³.
8. Remove its capsule from the surrounding diencephalic neural structures, arresting the resection in case of tumor adhesion to eloquent structures visualized under neuronavigation guidance⁴³.
9. Explore the surgical cavity with angled optics (Table of Materials)⁴⁶.
10. Ensure hemostasis with bipolar coagulation or hemostatic agents.
11. Close the osteo-meningeal opening with an intradural intracranial layer of dural substitute⁴³.
12. Place an extradural intracranial layer of dural substitute, scaffolded with abdominal fat and eventually bone (Table of Materials)⁴³.
13. Cover the closure with the naso-septal flap⁴³.

14. Histological examination

1. Fix tumor samples with 10% formalin and embed them in paraffin immediately after surgery.
2. Cut tissue into sections of 4 µm thickness and stain with hematoxylin and eosin. The histological diagnosis must be based on the most recent version of the WHO classification of brain tumors (2016)⁴⁷.
3. Perform specimen immunohistochemical staining by an automated immunohistochemical staining instrument, using avidin-biotin labeling and diaminobenzidine as a detection reagent. For craniopharyngiomas, adopt anti-beta-catenin, anti-BRAF v600E mutant epitope, and anti-Ki67 antibodies for immunohistochemical staining (Table of Materials).
4. Evaluate the Ki-67 index through the manual count of positive tumor cells⁴⁸.

15. Post-surgical patient management

1. Wake the patient immediately after surgery.
2. Restore spontaneous breathing from the mouth by filling nasal cavities with absorbable and non-absorbable material.
3. Monitor vital parameters (blood pressure, heart rate, oxygen saturation and consciousness state) for the following 6-12 hours in ICU.
4. Restore oral feeding after 12 hours.
5. Perform a CT scan after 6-9 hours.
6. Maintain bed rest for three days with heparin treatment.
7. Control fluid balance every 12 hours and assess serum electrolytes every 24 hours.
8. Administer corticosteroid therapy (endovenous flebocortid in the first 24 hours, and then oral cortone acetate 30 +15 mg/day).
9. Perform an MRI with/without gadolinium within 72 hours after surgery.
10. Discharge the patient on the 4^th day.

16. Early follow-up

1. Repeat the complete endocrinological assessment 30 days after surgery⁴³.
2. Repeat the ophthalmological assessment three months after surgery⁴³.
3. Repeat the neurological physical examination and temperature and sleep-wake rhythms function investigations three months after surgery⁴⁶.
4. Perform the MRI with/without gadolinium three months after surgery⁴⁶.

17. Adjuvant therapy

1. Evaluate the presence of early tumor progression, and if it is indicated, refer the patient to radiation therapy⁴³.

18. Long-term follow-up

1. Repeat the clinical, endocrinological, and ophthalmological assessments annually⁴³.
2. Perform yearly MRI with/without gadolinium: in case of recurrence, the patient can be re-operated on and then referred to radiation therapy or directly referred to radiotherapy⁴³.

Results

A 55-year-old woman presented with progressive visual deficits. Her medical history was unremarkable. On ophthalmological evaluation, bilateral reduction of visual acuity (6/10 in the right eye and 8/10 in the left eye) was revealed, and the computerized visual field showed complete bitemporal hemianopia. No further deficits were evident on neurological examination, but the patient reported persistent asthenia and an increase in hunger and thirst sensation in the previous 2-3 months, with...

Discussion

The application of the presented protocol resulted in a safe and effective treatment of one of the most challenging intracranial tumors such as a craniopharyngioma invading the 3^rd ventricle, possibly opening up a new horizon for a lesion that was defined by H. Cushing about a century ago as the most baffling intracranial neoplasm¹. The combination of accurate preoperative planning, integrating advanced MRI techniques, and multidisciplinary clinical assessments have permitted us to tail...

Materials

Name	Company	Catalog Number	Comments
BRAF V600E-specific clone VE1	Ventana
Dural Substitute	Biodesign, Cook Medical
Endoscope	Karl Storz, 4mm in diameter, 18 cm in length, Hopkins II – Karl Storz Endoscopy
Immunohistochemical staining instrument	Ventana Benchmark, Ventana Medical Systems
MRI	3T Magnetom Skyra, Siemens Health Care
Neuronavigator	Stealth Station S8 Surgical Navigation System, MEDTRONIC