Accuracy Of Surface Registration Research Articles

Deep Learning-Based Fine-Tuning Approach of Coarse Registration for Ear-Nose-Throat (ENT) Surgical Navigation Systems.

Accurate registration between medical images and patient anatomy is crucial for surgical navigation systems in minimally invasive surgeries. This study introduces a novel deep learning-based refinement step to enhance the accuracy of surface registration without disrupting established workflows. The proposed method integrates a machine learning model between conventional coarse registration and ICP fine registration. A deep-learning model was trained using simulated anatomical landmarks with introduced localization errors. The model architecture features global feature-based learning, an iterative prediction structure, and independent processing of rotational and translational components. Validation with silicon-masked head phantoms and CT imaging compared the proposed method to both conventional registration and a recent deep-learning approach. The results demonstrated significant improvements in target registration error (TRE) across different facial regions and depths. The average TRE for the proposed method (1.58 ± 0.52 mm) was significantly lower than that of the conventional (2.37 ± 1.14 mm) and previous deep-learning (2.29 ± 0.95 mm) approaches (p < 0.01). The method showed a consistent performance across various facial regions and enhanced registration accuracy for deeper areas. This advancement could significantly enhance precision and safety in minimally invasive surgical procedures.

High-Accuracy Neuro-Navigation with Computer Vision for Frameless Registration and Real-Time Tracking.

For the past three decades, neurosurgeons have utilized cranial neuro-navigation systems, bringing millimetric accuracy to operating rooms worldwide. These systems require an operating room team, anesthesia, and, most critically, cranial fixation. As a result, treatments for acute neurosurgical conditions, performed urgently in emergency rooms or intensive care units on awake and non-immobilized patients, have not benefited from traditional neuro-navigation. These emergent procedures are performed freehand, guided only by anatomical landmarks with no navigation, resulting in inaccurate catheter placement and neurological deficits. A rapidly deployable image-guidance technology that offers highly accurate, real-time registration and is capable of tracking awake, moving patients is needed to improve patient safety. The Zeta Cranial Navigation System is currently the only non-fiducial-based, FDA-approved neuro-navigation device that performs real-time registration and continuous patient tracking. To assess this system's performance, we performed registration and tracking of phantoms and human cadaver heads during controlled motions and various adverse surgical test conditions. As a result, we obtained millimetric or sub-millimetric target and surface registration accuracy. This rapid and accurate frameless neuro-navigation system for mobile subjects can enhance bedside procedure safety and expand the range of interventions performed with high levels of accuracy outside of an operating room.

Evaluation of DIR schemes on tumor/organ with progressive shrinkage: Accuracy of tumor/organ internal tissue tracking during the radiation treatment

PurposeAccuracy of intratumoral treatment dose accumulation and response assessment highly depends on the accuracy of a DIR method. However, achievable accuracy of the existing DIR methods for tumor/organ with large and progressive shrinkage during the radiotherapy course have not been explored. This study aimed to use a bio-tissue phantom to quantify the achievable accuracy of different DIR schemes. Materials /methodsA fresh porcine liver was used for phantom material. Sixty gold markers were implanted on the surface and inside of the liver. To simulate the progressive radiation-induced tumor/organ shrinkage, the phantom was heated using a microwave oven incrementally from 30 s to 200 s in 8 phases. For each phase, the phantom was scanned by CT. Two extra image sets were generated from the original images: 1) the image set with overriding the high-density gold markers (feature image); 2) the image set with overriding the entire phantom to the mean soft tissue intensity (featureless image). Ten DIR schemes were evaluated to mimic clinical treatment situations of tumor/critical organ with respect to their surface and internal condition of image features, availability of intermediate feedback images and DIR methods. The internal marker's positions were utilized to evaluate DIR accuracy quantified by target registration error (TRE). ResultsVolume reduction was about 20 % ∼ 40 % of the initial volume after 90 s ∼ 200 s of the heating. Without image features on the surface and inside of the phantom, the hybrid-DIR (image-based DIR followed by biomechanical model-based refinement) with the surface constraint achieved the registration TRE from 2.6 ± 1.2 mm to 5.3 ± 2.6 mm proportional to the %volume shrinkage. Meanwhile, the hybrid-DIR with the surface-marker constraint achieved the TRE from 2.4 ± 1.2 mm to 2.6 ± 1.0 mm. If both the surface and internal image features would be viable on the feedback images, the achievable accuracy could be minimal with the TRE from 1.6 ± 0.9 mm to 1.9 ± 1.2 mm. ConclusionsStandard DIR methods cannot guarantee intratumoral tissue registration accuracy for tumor/organ with large progressive shrinkage. Achievable accuracy with using the hybrid DIR method is highly dependent on the surface registration accuracy. If the surface registration mean TRE can be controlled within 2 mm, the mean TRE of internal tissue can be controlled within 3 mm.

Automated machine learning (AutoML)-based surface registration methodology for image-guided surgical navigation system.

Although the surface registration technique has the advantage of being relatively safe and the operation time is short, it generally has the disadvantage of low accuracy. This research proposes automated machine learning (AutoML)-based surface registration to improve the accuracy of image-guided surgical navigation systems. The state-of-the-art surface registration concept is that first, using a neural network model, a new point-cloud that matches the facial information acquired by a passive probe of an optical tracking system (OTS) is extracted from the facial information obtained by computerized tomography. Target registration error (TRE) representing the accuracy of surface registration is then calculated by applying the iterative closest point (ICP) algorithm to the newly extracted point-cloud and OTS information. In this process, the hyperparameters used in the neural network model and ICP algorithm are automatically optimized using Bayesian optimization with expected improvement to yield improved registration accuracy. Using the proposed surface registration methodology, the average TRE for the targets located in the sinus space and nasal cavity of the soft phantoms is 0.939±0.375mm, which shows 57.8% improvement compared to the average TRE of 2.227±0.193mm calculated by the conventional surface registration method (p<0.01). The performance of the proposed methodology is evaluated, and the average TREs computed by the proposed methodology and the conventional method are 0.767±0.132 and 2.615±0.378mm, respectively. Additionally, for one healthy adult, the clinical applicability of the AutoML-based surface registration is also presented. Our findings showed that the registration accuracy could be improved while maintaining the advantages of the surface registration technique.

Accuracy of electroanatomical mapping-guided cardiac radiotherapy for ventricular tachycardia: pitfalls and solutions.

To analyse and optimize the interobserver agreement for gross target volume (GTV) delineation on cardiac computed tomography (CCT) based on electroanatomical mapping (EAM) data acquired to guide radiotherapy for ventricular tachycardia (VT). Electroanatomical mapping data were exported and merged with the segmented CCT using manual registration by two observers. A GTV was created by both observers for predefined left ventricular (LV) areas based on preselected endocardial EAM points indicating a two-dimensional (2D) surface area of interest. The influence of (interobserver) registration accuracy and availability of EAM data on the final GTV and 2D surface location within each LV area was evaluated. The median distance between the CCT and EAM after registration was 2.7 mm, 95th percentile 6.2 mm for observer #1 and 3.0 mm, 95th percentile 7.6 mm for observer #2 (P = 0.9). Created GTVs were significantly different (8 vs. 19 mL) with lowest GTV overlap (35%) for lateral wall target areas. Similarly, the highest shift between 2D surfaces was observed for the septal LV (6.4 mm). The optimal surface registration accuracy (2.6 mm) and interobserver agreement (Δ interobserver EAM surface registration 1.3 mm) was achieved if at least three cardiac chambers were mapped, including high-quality endocardial LV EAM. Detailed EAM of at least three chambers allows for accurate co-registration of EAM data with CCT and high interobserver agreement to guide radiotherapy of VT. However, the substrate location should be taken in consideration when creating a treatment volume margin.

Accuracy of electroanatomical mapping guided cardiac radiotherapy for ventricular arrhytmias: pitfalls and solutions

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): This work was partially supported by the grant AZV NU20-02-00244 from the Ministry of Health of the Czech Republic OnBehalf WECAM Background One of the main limitations of Radiofrequency catheter ablation (RFCA) is the difficulty in targeting specific VT substrate locations when they are deep intramural, covered by epicardial fat or in the proximity of a coronary artery. Stereotactic body radiotherapy (SBRT) has emerged as an alternative treatment for VT after RFCA failure. However, the accuracy of substrate volume delineation, the gross target volume (GTV), based on electroanatomical mapping (EAM) and interobserver agreement are still unknown. Aim To analyze and optimize the interobserver agreement for GTV delineation on cardiac CT (CCT) based on EAM data acquired to guide radiotherapy for ventricular arrhythmias. Method Available EAM data was exported and merged with the segmented CCT using manual registration by two observers. A GTV was created by the observers for predefined left ventricular (LV) areas based on preselected endocardial EAM points, indicating the 2D-surface area of interest. The interobserver (non)overlapping GTV volumes were analyzed. The influence of (interobserver) registration accuracy and availability of EAM data on the final GTV and 2D-surface location within each LV area was evaluated. Results The median distance between the CCT and EAM after registration was 2.7 mm, 95%th percentile 6.2 mm for observer #1 and 3.0 mm, 95%th percentile 7.6 mm for observer #2 (p = 0.9). Created GTVs were significantly different (8 vs 19 ml) with lowest GTV overlap (35%) for lateral wall target areas. Similar, the highest shift between 2D-surfaces was observed for the lateral LV (6.4mm (3.0-9.9)). The same trend is seen in the (non) overlapping volume for lateral area (P = 0.004). The optimal surface registration accuracy (2.6 mm) and interobserver agreement (Δ interobserver EAM surface registration accuracy 1.32 mm) on final GTVs could be achieved if at least three cardiac chambers were mapped, including high-density LV EAM. Conclusion Detailed EAM of at least three chambers allows for accurate co-registration of EAM data with CCT and high interobserver agreement on GTVs to guide radiotherapy of ventricular arrythmias. However, the substrate location should be taken in consideration when creating a treatment volume margin. Gross target volume (GTV) and InterobserGTV volume Observer #1 (cc)GTV volume Observer #2 (cc)Surface shift (mm)Basal anterior, (IQR)7.6 (6.2-11.5)9.3 (5.3-10)3.8 (2.2-4.5)Lateral, (IQR)8.0 (6.9-12.7)19.1 (15.7-21.4)6.2 (5.3-6.4)Septal, (IQR)10.1 (6.2-12.8)13.9 (7.7-14.6)6.4 (3.0-9.9)Abstract Figure. Graphical abstract

A Novel Surface Registration for Image-Guided Neurosurgery: Effects of Intervals of Points in Patient Space on Registration Accuracy

Surface registration is an important factor in surgical navigation in determining the success rate and stability of surgery by providing the operator with the exact location of a lesion. The problem of surface registration is that point cloud in the patient space is acquired at irregular intervals due to the operator’s tracking speed and method. The purpose of this study is to analyze the effect of irregular intervals of point cloud caused by tracking speed and method on the registration accuracy. For this study, we created the head phantom to obtain a point cloud in the patient space with a similar object to that of a patient and acquired a point cloud in a total of ten times. In order to analyze the accuracy of registration according to the interval, cubic spline interpolation was applied to the existing point cloud. Additionally, irregular intervals of the point cloud were regenerated into regular intervals. As a result of applying the regenerated point cloud to the surface registration, the surface registration error was not statistically different from the existing point cloud. However, the target registration error significantly lower (p &lt; 0.01). These results indicate that the intervals of point cloud affect the accuracy of registration, and that point cloud with regular intervals can improve the surface registration accuracy.

Acquisition of point cloud in CT image space to improve accuracy of surface registration: Application to neurosurgical navigation system

One important technique of surgical navigation is surface registration that matches coordinates in two different spaces. We proposed a novel method to extract an optimal point cloud in image space corresponding to the point cloud in patient space to improve the registration accuracy. In the hemispherical study, our proposed method demonstrated a reduced surface registration error (SRE) and target registration error (TRE) compared to the conventional method in all cases (number of points, distribution, noise). In the phantom study, the SRE and TRE were low and stable in the proposed method, with SRE reduced by 22 % and TRE by 18 % (p < 0.05). The proposed method was not affected by the number of points, distribution, or noise in the point cloud, and it could improve the registration accuracy without the aid of additional equipment.

Surface Registration Accuracy of Clinically Obtained Intraoral Optical Scans with Manually Threshold Segmented CBCT Data

Combining CBCT imaging and intraoral scans with an intention to prepare comprehensive treatment plans is common in clinical practice. Segmentation and superimposition of models is indispensable when planning complicated prosthetic reconstructions. The authors of this publication evaluated registration discrepancies of virtual dental arch models obtained by means of CBCT imaging at different segmentation thresholds with intraoral dental scans. For comparisons, intraoral digital scans and volumetric CBCT scans of the upper and lower jaw were used in randomly selected patients. The mean distance, Gaussian mean and standard deviation from the Gaussian mean as registration inconsistencies between the combined models were measured and subjected to a statistical analysis. The results showed that the registration of superimposed models may be affected by errors of up to 300 microns in the case of full dental arches. The statistical analysis proved that there was no correlation between the segmentation threshold and the quantitative variables studied (e.g. mean distance, Gaussian mean and standard deviation from the Gaussian mean). The results of this study indicate that CBCT data and full arch optical scans can be superimposed and successfully applied under clinical conditions within the accepted error.

Accuracy of surface registration in cranial electromagnetic navigation

Accuracy of surface registration in cranial electromagnetic navigation

Three-dimensional analysis of the surface registration accuracy of electromagnetic navigation systems in live endoscopic sinus surgery

Background: Studies on the performance of surface registration with electromagnetic tracking systems are lacking in both live surgery and the laboratory setting. This study presents the efficiency in time of the system preparation as well as the navigational accuracy of surface registration using electromagnetic tracking systems. Methodology: Forty patients with bilateral chronic paranasal pansinusitis underwent endoscopic sinus surgery after undergoing sinus computed tomography scans. The surgeries were performed under electromagnetic navigation guidance after the surface registration had been carried out on all of the patients. The intraoperative measurements indicate the time taken for equipment set-up, surface registration and surgical procedure, as well as the degree of navigation error along 3 axes. Results: The time taken for equipment set-up, surface registration and the surgical procedure was 179 +- 23 seconds, 39 +- 4.8 seconds and 114 +- 36 minutes, respectively. A comparison of the navigation error along the 3 axes showed that the deviation in the medial-lateral direction was significantly less than that in the anterior-posterior and cranial-caudal directions. Conclusion: The procedures of equipment set-up and surface registration in electromagnetic navigation tracking are efficient, convenient and easy to manipulate. The system accuracy is within the acceptable ranges, especially on the medial-lateral axis.

Three-dimensional analysis of the surface registration accuracy of electromagnetic navigation systems in live endoscopic sinus surgery

Studies on the performance of surface registration with electromagnetic tracking systems are lacking in both live surgery and the laboratory setting. This study presents the efficiency in time of the system preparation as well as the navigational accuracy of surface registration using electromagnetic tracking systems. Forty patients with bilateral chronic paranasal pansinusitis underwent endoscopic sinus surgery after undergoing sinus computed tomography scans. The surgeries were performed under electromagnetic navigation guidance after the surface registration had been carried out on all of the patients. The intraoperative measurements indicate the time taken for equipment set-up, surface registration and surgical procedure, as well as the degree of navigation error along 3 axes. The time taken for equipment set-up, surface registration and the surgical procedure was 179 +- 23 seconds, 39 +- 4.8 seconds and 114 +- 36 minutes, respectively. A comparison of the navigation error along the 3 axes showed that the deviation in the medial-lateral direction was significantly less than that in the anterior-posterior and cranial-caudal directions. The procedures of equipment set-up and surface registration in electromagnetic navigation tracking are efficient, convenient and easy to manipulate. The system accuracy is within the acceptable ranges, especially on the medial-lateral axis.

Three-dimensional cardiac image integration of electroanatomical mapping of only left atrial posterior wall with CT image to guide circumferential pulmonary vein ablation

CARTO merge™ is used in integrating left atrial (LA) CARTO and computed tomography (CT) images, but the integration method is not established. The relative anatomic position and configuration of the LA posterior wall (LAPW) between the right and left pulmonary veins (PVs) are not greatly affected by respiration, LA contraction or hydration state. LAPW and adjacent roof area has an anatomically curved structure which is geometrically amenable to integration using a surface registration software. We examined the accuracy of surface registration using only the LAPW CARTO image constructed by mapping of a mean of 101 ± 34 points in 108 consecutive AF patients before PV isolation. After visual alignment of CARTO LAPW and LA CT images using one anatomically defined position in each image, the two images were integrated with an installed surface registration program. Points with differences ≥ 4.0mm between the two images were deleted (mean, 17 points/patient) and a second surface registration was performed. The mean distance between CARTO and CT images was 1.37 ± 0.23mm, with mean minimum and maximum values of 0.03 and 3.99mm, respectively. The accuracy of integration was verified in 34 patients by measuring the gaps between the catheter tip on the LA wall and the design line delineated for PV isolation on the integrated image. The gaps (mm) at the superior, inferior, anterior and posterior sites on the right PV side were 0.8 ± 0.5, 0.9 ± 0.7, 1.3 ± 1.0, and 0.8 ± 0.7, respectively, and those on the left side were 0.8 ± 0.5, 0.9 ± 0.7, 1.0 ± 0.5, and 1.0 ± 0.6, respectively. Thus, the gaps were all <1.0mm, except for the right anterior aspect. These results show that surface registration using only the LAPW image can accurately integrate CARTO and CT images.

Reliability of an efficient MRI-based method for estimation of knee cartilage volume using surface registration

To aid in detection of osteoarthritis (OA) progression in serial magnetic resonance (MR) scans, we assessed feasibility and accuracy of rapid 3D image registration of the tibial plateau in normal and arthritic subjects, and inter-scan reliability of semi-automated cartilage volume measurement from these images. Two T1 fat-suppressed knee MR scans were obtained 2 weeks apart in healthy adults (n = 9, age 23-48 years). Four scans of each of three patients with established OA were obtained over 2 years. At baseline, the tibial surface was digitized by semi-automated edge detection and medial tibial plateau cartilage volume was calculated from high-intensity voxels within a manually drawn region of interest (ROI). In subsequent scans, the digitized tibial surface was registered to the baseline location by photogrammetric 3D coordinate transformation, and cartilage volume was automatically recalculated by reuse of the ROI. We measured registration accuracy by root mean square (RMS) distance between registered tibial surfaces. In normals, RMS distance between tibial surfaces in baseline and subsequent scans was 1/3 voxel length (0.121 mm), and medial tibial plateau cartilage volumes varied by 1.4+/-3.2%. Despite change in cartilage volumes by up to 20% over 2 years in arthritic patients, surface registration accuracy was unaffected (0.122 mm). User-supervised processing time was 15 min at baseline and 7 min in subsequent scans. Tibial surfaces on magnetic resonance imaging (MRI) can be rapidly and accurately co-registered, even in arthritic knees, allowing direct visualization of changes over time. Compared to most current methods, cartilage volume measurement in registered images is faster and has equivalent inter-scan reliability in initially normal subjects.

Validation of Three‐Dimensional Cardiac Image Integration: Use of Integrated CT Image into Electroanatomic Mapping System to Perform Catheter Ablation of Atrial Fibrillation

Accurate visualization of the complex left atrial (LA) anatomy and the location of an ablation catheter within the chamber is important in the success and safety of ablation for atrial fibrillation (AF). We describe the integration of CT into an electroanatomic mapping (EAM) system and its validation in patients undergoing catheter ablation for AF. Thirty patients (59.2 +/- 8 years, 25 M) with paroxysmal (12) and persistent (18) AF underwent ablation using CT image integration into an electroanatomic mapping system. CT registration using the pulmonary veins as markers (landmark) was achieved with an error of 6.4 +/- 2.8 mm with repeat registration required in two patients. Registration of the CT by best fit to a electroanatomic geometry (surface) was achieved with an error of 2.3 +/- 0.4 mm. There was no significant difference in the regional LA registration error at superior (1.7 +/- 0.7 mm), inferior (2.2 +/- 1.4 mm), septal (1.7 +/- 0.8 mm), and lateral (1.7 +/- 0.7 mm, P = 0.13) sites. Cardiac rhythm at the time of CT did not have a significant effect on total or regional surface registration accuracy (mean total 2.5 +/- 0.3 in AF patients vs 2.3 +/- 0.5 in SR patients, P = 0.22). The integrated CT was used to guide the encirclement of the pulmonary veins (PV) in pairs with electrical isolation achieved by maintaining ablation along the ablation line in 58 of 60 PV pairs. Postprocedural PV angiography did not demonstrate significant stenosis. CT image integration into an EAM system was successfully performed in patients undergoing catheter ablation for AF. With a greater appreciation of the complex and variable nature of the PV and LA anatomy this new technology may improve the efficacy and safety of the procedure.

Effects of CT threshold value to make a surface bone model on accuracy of shape-based registration in a CT-based navigation system for hip surgery

To clarify the effects of different CT threshold values used to make computer models on the accuracy of surface registration in a CT-based navigation system, a simulation study was performed using CT data of 30 patients who underwent total hip arthroplasty with navigation guidance. Surface models of the pelvis for use in clinical applications were made by contouring the periosteal boundary at threshold ranging from 140 to 260 Hounsfield units (HU) (mean, 200 HU). In each case, the threshold was determined to be approximately 200 HU, based on the balance between soft tissue noise and surface defects. Ten pelvic surface models were made from each set of CT data, using 10 different CT threshold values ranging from 50 to 320 HU with an increment of 30 HU. The center of the acetabulum was defined as the target point in each set of CT data, so that each set of 10 surface models should have the same reference point for measuring the positional and rotational differences among the models after registration. The average residue of registration reached its nadir was minimum with the 200 HU models, and there were no significant differences in residue of registration among the models made at thresholds ranging from 110 to 320 HU. The target registration errors for position and rotation both showed strong correlation with the residue of registration ( R=0.879 and 0.880, respectively). We thus conclude that accurate surface registration can be obtained with computer models made at thresholds ranging from 110 to 320 HU by assessing the balance between soft tissue noise and bone surface defects.