Developing an Evidence-Based Surgical Curriculum: Learning from a Randomized Controlled Trial of Surgical Rehearsal in Virtual Reality.

BACKGROUND: Surgical rehearsal - patient-specific preoperative surgical practice - can be provided by virtual reality simulation. This study investigated the effect of surgical rehearsal on cortical mastoidectomy performance and procedure duration. METHODS: University students (n=40) were randomized evenly into a rehearsal and control group. After watching a video tutorial on cortical mastoidectomy, participants completed the procedure on a virtual reality simulator as a pre-test. Participants completed a further 8 cortical mastoidectomies on the virtual reality simulator as training before drilling two 3-dimensional (3D) printed temporal bones. The rehearsal group received 3D printed bones they had previously operated on in virtual reality, while the control group received 2 new bones. Cortical mastoidectomy was assessed by 3 blinded graders using the Melbourne Mastoidectomy Scale. RESULTS: There was high interrater reliability between the 3 graders (intraclass correlation coefficient, r=0.8533, P < .0001). There was no difference in the mean surgical performance on the two 3D printed bones between the control and rehearsal groups (P=.2791). There was no significant difference in the mean procedure duration between the control and rehearsal groups for both 3D printed bones (P=.8709). However, there was a significant decrease in procedure duration between the first and second 3D printed bones (P < .0001). CONCLUSION: In this study, patient-specific virtual reality rehearsal provided no additional advantage to cortical mastoidectomy performance by novice operators compared to generic practice on a virtual reality simulator. Further, virtual reality training did not improve cortical mastoidectomy performance on 3D printed bones, highlighting the impact of anatomical diversity and changing operating modalities on the acquisition of new surgical skills.

Automatic analysis of cochlear response using electrocochleography signals during cochlear implant surgery.

Cochlear implants (CIs) provide an opportunity for the hearing impaired to perceive sound through electrical stimulation of the hearing (cochlear) nerve. However, there is a high risk of losing a patient's natural hearing during CI surgery, which has been shown to reduce speech perception in noisy environments as well as music appreciation. This is a major barrier to the adoption of CIs by the hearing impaired. Electrocochleography (ECochG) has been used to detect intra-operative trauma that may lead to loss of natural hearing. There is early evidence that ECochG can enable early intervention to save natural hearing of the patient. However, detection of trauma by observing changes in the ECochG response is typically carried out by a human expert. Here, we discuss a method of automating the analysis of cochlear responses during CI surgery. We establish, using historical patient data, that the proposed method is highly accurate (∼94% and ∼95% for sensitivity and specificity respectively) when compared to a human expert. The automation of real-time cochlear response analysis is expected to improve the scalability of ECochG and improve patient safety.

A Deep Learning Framework for Segmenting Brain Tumors Using MRI and Synthetically Generated CT Images.

Multi-modal three-dimensional (3-D) image segmentation is used in many medical applications, such as disease diagnosis, treatment planning, and image-guided surgery. Although multi-modal images provide information that no single image modality alone can provide, integrating such information to be used in segmentation is a challenging task. Numerous methods have been introduced to solve the problem of multi-modal medical image segmentation in recent years. In this paper, we propose a solution for the task of brain tumor segmentation. To this end, we first introduce a method of enhancing an existing magnetic resonance imaging (MRI) dataset by generating synthetic computed tomography (CT) images. Then, we discuss a process of systematic optimization of a convolutional neural network (CNN) architecture that uses this enhanced dataset, in order to customize it for our task. Using publicly available datasets, we show that the proposed method outperforms similar existing methods.

Surgical approach to the facial recess influences the acceptable trajectory of cochlear implantation electrodes.

PURPOSE: To provide practical guidance to the operative surgeon by mapping the location, where acceptable straight-line virtual cochlear implant electrode trajectories intersect the facial recess. In addition, to investigate the influence of facial recess preparation, virtual electrode width and surgical approach to the cochlea on these available trajectories. METHODS: The study was performed on imaging data from eight cadaveric temporal bones within the University of Melbourne Virtual Reality (VR) Temporal Bone Surgery Simulator. The facial recess was opened to varying degrees, and acceptable trajectory vectors with varying diameters were calculated for electrode insertions via cochleostomy or round window membrane (RWM). The percentage of acceptable insertion vectors through each location of the facial recess was visually represented using heatmaps. RESULTS: Seven of the eight bones allowed for acceptable vector trajectories via both cochleostomy and RWM approaches. These acceptable trajectories were more likely to lie superiorly within the facial recess for insertion via the round window, and inferiorly for insertion via cochleostomy. Cochleostomy insertions required a greater degree of preparation and skeletonisation of the junction of the facial nerve and chorda tympani within the facial recess. The width of the virtual electrode had only marginal impact on the availability of acceptable trajectories. Heatmaps emphasised the intimate relationship the acceptable trajectories have with the facial nerve and chorda tympani. CONCLUSION: These findings highlight the differences in the acceptable straight-line trajectories for electrodes when implanted via the round window or cochleostomy. There were notable exceptions to both surgical approaches, likely explained by the variation of hook region anatomy. The methodology used in this study holds promise for translation to patient specific surgical planning.

Automated assessment of cortical mastoidectomy performance in virtual reality.

INTRODUCTION: Cortical mastoidectomy is a core skill that Otolaryngology trainees must gain competency in. Automated competency assessments have the potential to reduce assessment subjectivity and bias, as well as reducing the workload for surgical trainers. OBJECTIVES: This study aimed to develop and validate an automated competency assessment system for cortical mastoidectomy. PARTICIPANTS: Data from 60 participants (Group 1) were used to develop and validate an automated competency assessment system for cortical mastoidectomy. Data from 14 other participants (Group 2) were used to test the generalisability of the automated assessment. DESIGN: Participants drilled cortical mastoidectomies on a virtual reality temporal bone simulator. Procedures were graded by a blinded expert using the previously validated Melbourne Mastoidectomy Scale: a different expert assessed procedures by Groups 1 and 2. Using data from Group 1, simulator metrics were developed to map directly to the individual items of this scale. Metric value thresholds were calculated by comparing automated simulator metric values to expert scores. Binary scores per item were allocated using these thresholds. Validation was performed using random sub-sampling. The generalisability of the method was investigated by performing the automated assessment on mastoidectomies performed by Group 2, and correlating these with scores of a second blinded expert. RESULTS: The automated binary score compared with the expert score per item had an accuracy, sensitivity and specificity of 0.9450, 0.9547 and 0.9343, respectively, for Group 1; and 0.8614, 0.8579 and 0.8654, respectively, for Group 2. There was a strong correlation between the total scores per participant assigned by the expert and calculated by the automatic assessment method for both Group 1 (r = .9144, P < .0001) and Group 2 (r = .7224, P < .0001). CONCLUSION: This study outlines a virtual reality-based method of automated assessment of competency in cortical mastoidectomy, which proved comparable to the assessment provided by human experts.

A deep learning based framework for the registration of three dimensional multi-modal medical images of the head.

Image registration is a fundamental task in image analysis in which the transform that moves the coordinate system of one image to another is calculated. Registration of multi-modal medical images has important implications for clinical diagnosis, treatment planning, and image-guided surgery as it provides the means of bringing together complimentary information obtained from different image modalities. However, since different image modalities have different properties due to their different acquisition methods, it remains a challenging task to find a fast and accurate match between multi-modal images. Furthermore, due to reasons such as ethical issues and need for human expert intervention, it is difficult to collect a large database of labelled multi-modal medical images. In addition, manual input is required to determine the fixed and moving images as input to registration algorithms. In this paper, we address these issues and introduce a registration framework that (1) creates synthetic data to augment existing datasets, (2) generates ground truth data to be used in the training and testing of algorithms, (3) registers (using a combination of deep learning and conventional machine learning methods) multi-modal images in an accurate and fast manner, and (4) automatically classifies the image modality so that the process of registration can be fully automated. We validate the performance of the proposed framework on CT and MRI images of the head obtained from a publicly available registration database.

Development of a virtual reality clinically oriented temporal bone anatomy module with randomised control study of three-dimensional display technology.

Objective: To investigate the effectiveness of a virtual reality (VR), three-dimensional (3D) clinically orientated temporal bone anatomy module, including an assessment of different display technologies. Methods: A clinically orientated, procedural and interactive anatomy module was generated from a micro-CT of a cadaveric temporal bone. The module was given in three different display technologies; 2D, 3D with monoscopic vision, and 3D with stereoscopic vision. A randomised control trial assessed the knowledge acquisition and attitudes of 47 medical students though a pretutorial and post-tutorial questionnaire. The questionnaire included questions identifying anatomic structures as well as understanding structural relations and clinical relevance. Furthermore, a five-point Likert scale assessed the students' attitudes to the module and alternative learning outcomes, such as interest in otology and preparedness for clinical rotations. Results: As a whole cohort, the total test score improved significantly, with a large effect size (p≤0.005, Cohen's d=1.41). The 23 students who returned the retention questionnaire had a significant improvement in total test score compared with their pretutorial score, with a large effect size (p≤0.005, Cohen's d=0.83). Display technology did not influence the majority of learning outcomes, with the exception of 3D technologies, showing a significantly improvement in understanding of clinical relevance and structural relations (p=0.034). Students preferred 3D technologies for ease of use, perceived effectiveness and willingness to use again. Conclusions: The developed VR temporal bone anatomy tutor was an effective self-directed education tool. 3D technology remains valuable in facilitating spatial learning and superior user satisfaction.

A Vision-Based Machine Learning Method for Barrier Access Control Using Vehicle License Plate Authentication.

Automatic vehicle license plate recognition is an essential part of intelligent vehicle access control and monitoring systems. With the increasing number of vehicles, it is important that an effective real-time system for automated license plate recognition is developed. Computer vision techniques are typically used for this task. However, it remains a challenging problem, as both high accuracy and low processing time are required in such a system. Here, we propose a method for license plate recognition that seeks to find a balance between these two requirements. The proposed method consists of two stages: detection and recognition. In the detection stage, the image is processed so that a region of interest is identified. In the recognition stage, features are extracted from the region of interest using the histogram of oriented gradients method. These features are then used to train an artificial neural network to identify characters in the license plate. Experimental results show that the proposed method achieves a high level of accuracy as well as low processing time when compared to existing methods, indicating that it is suitable for real-time applications.

The Melbourne Mastoidectomy Scale: Validation of an end-product dissection scale for cortical mastoidectomy.

INTRODUCTION: Cortical mastoidectomy is a common otolaryngology procedure and represents a compulsory part of otolaryngology training. As such, a specific validated assessment score is needed for the progression of competency-based training in this procedure. Although multiple temporal bone dissection scales have been developed, they have all been validated for advanced temporal bone dissection including posterior tympanotomy, rather than the task of cortical mastoidectomy. METHODS: The Melbourne Mastoidectomy Scale, a 20-item end-product dissection scale to assess cortical mastoidectomy, was developed. The scale was validated using dissections by 30 participants (10 novice, 10 intermediate and 10 expert) on a virtual reality temporal bone simulator. All dissections were assessed independently by three blinded graders. Additionally, all procedures were graded with an abbreviated Welling Scale by one grader. RESULTS: There was high inter-rater reliability between the three graders (r = .9210, P < .0001). There was a significant difference in scores between the three groups (P < .0001). Additionally, there was a large effect size between all three groups: the differences between the novice group and both the intermediate group (P = .0119, η2  = 0.2482) and expert group (P < .001, η2  = 0.6356) were significant. The difference between the intermediate group and expert group again had a large effect size (η2  = 0.3217), but was not significant. The Melbourne Mastoidectomy Scale correlated well with an abbreviated Welling Scale (r = .8485, P < .0001). CONCLUSION: The Melbourne Mastoidectomy Scale offers a validated score for use in the assessment of cortical mastoidectomy.

A rotation and translation invariant method for 3D organ image classification using deep convolutional neural networks.

Three-dimensional (3D) medical image classification is useful in applications such as disease diagnosis and content-based medical image retrieval. It is a challenging task due to several reasons. First, image intensity values are vastly different depending on the image modality. Second, intensity values within the same image modality may vary depending on the imaging machine and artifacts may also be introduced in the imaging process. Third, processing 3D data requires high computational power. In recent years, significant research has been conducted in the field of 3D medical image classification. However, most of these make assumptions about patient orientation and imaging direction to simplify the problem and/or work with the full 3D images. As such, they perform poorly when these assumptions are not met. In this paper, we propose a method of classification for 3D organ images that is rotation and translation invariant. To this end, we extract a representative two-dimensional (2D) slice along the plane of best symmetry from the 3D image. We then use this slice to represent the 3D image and use a 20-layer deep convolutional neural network (DCNN) to perform the classification task. We show experimentally, using multi-modal data, that our method is comparable to existing methods when the assumptions of patient orientation and viewing direction are met. Notably, it shows similarly high accuracy even when these assumptions are violated, where other methods fail. We also explore how this method can be used with other DCNN models as well as conventional classification approaches.

Street Sign Recognition Using Histogram of Oriented Gradients and Artificial Neural Networks.

Street sign identification is an important problem in applications such as autonomous vehicle navigation and aids for individuals with vision impairments. It can be especially useful in instances where navigation techniques such as global positioning system (GPS) are not available. In this paper, we present a method of detection and interpretation of Malaysian street signs using image processing and machine learning techniques. First, we eliminate the background from an image to segment the region of interest (i.e., the street sign). Then, we extract the text from the segmented image and classify it. Finally, we present the identified text to the user as a voice notification. We also show through experimental results that the system performs well in real-time with a high level of accuracy. To this end, we use a database of Malaysian street sign images captured through an on-board camera.

Do experts practice what they profess?

We investigated the variation of drilled regions of expert and trainee surgeons performing virtual temporal bone surgery to identify their compliance with standard drilling procedures. To this end, we recruited seven expert and six trainee ENT surgeons, who were asked to perform the surgical preparations for cochlear implantation on a virtual temporal bone. The temporal bone was divided into six regions using a semi-automated approach. The drilled area in each region was compared between groups using a sign test. Similarity within groups was calculated as a ratio of voxels (3D points) drilled by at least 75% of surgeons and at least 25% of surgeons. We observed a significant difference between groups when performing critical tasks such as exposing the facial nerve, opening the facial recess, and finding the round window. In these regions, experts' practice is more similar to each other than that between trainees. Consistent with models of skills development, expertise and expert-performance, the outcome of the analysis shows that experts perform similarly in critical parts of the procedure, and do indeed practice what they profess.

Defining the Hook Region Anatomy of the Guinea Pig Cochlea for Modeling of Inner Ear Surgery.

HYPOTHESIS: The aim of this study was to describe the hook region anatomy of the guinea pig cochlea to identify the optimal surgical approach for cochlear implantation and to determine what anatomical structures are at risk. BACKGROUND: Animal studies investigating hearing loss after cochlear implantation surgery are currently constrained by the lack of a reproducible implantation model. METHODS: Guinea pig cochleae were imaged using thin-sheet laser imaging microscopy. Images were stitched, reconstructed, and segmented for analysis. Insertion vectors were determined by tracing their paths to the outer wall and converting to Cartesian coordinates. Spherical surface and multiplane views were generated to analyze outer wall and radial forces of the insertion vector. RESULTS: Thin-sheet laser imaging microscopy enabled quantitative, whole specimen analysis of the soft and bony tissue relationships of the complex cochlear hook region in any desired plane without loss of image quality. Round window or cochleostomy approaches in the anteroinferior plane avoided direct damage to cochlear structures. Cochleostomy approach had large interindividual variability of angular depth and outer wall forces but predictable radial force. CONCLUSION: The guinea pig hook region and lower basal turn have similar structural relationships to humans. Careful cochleostomy placement is essentially for minimizing cochlear trauma and for ensuring a straight insertion vector that successfully advances around the outer wall. Experiments with guinea pigs that control for the surgical approach are likely to provide useful insights into the aetiology and the development of therapies directed at postimplantation hearing loss.

Comparison of Experts and Residents Performing a Complex Procedure in a Temporal Bone Surgery Simulator.

OBJECTIVE: To investigate the use of automated metrics from a virtual reality (VR) temporal bone surgery simulator to determine how the performance of experts and trainees differs when performing a complex otological procedure (mastoidectomy with posterior tympanotomy and cochleostomy). STUDY DESIGN: Cohort study. METHODS: Using the University of Melbourne VR temporal bone surgery simulator, seven ENT consultants and seven ENT residents performed two trials of the surgical approach to cochlear implantation on a virtual temporal bone. Simulator recordings were used to calculate a range of automated metrics for each stage of the procedure, capturing efficiency, technique characteristics, drilled bone regions, and damage to vital anatomical structures. RESULTS: Results confirm that experts drilled more efficiently than residents. Experts generally used larger burrs and applied higher forces, resulting in faster material removal. However, they exercised more caution when drilling close to anatomical structures. Residents opened the temporal bone more widely, but neglected important steps in obtaining a clear view toward the round window, such as thinning the external ear canal wall and skeletonizing the medial aspect of the facial nerve. Residents used higher magnification and reoriented the temporal bone more often than experts. CONCLUSION: VR simulation provides metrics that allow the objective analysis of surgical technique, and identification of differences between the performance of surgical residents and their senior colleagues. The performance of residents could be improved with more guidance regarding how much force they should apply, what burr size they should use, how they should orient the bone, and for cochlear implant surgery guidance regarding anatomical regions requiring particular attention, to visualize the round window.

Supporting skill acquisition in cochlear implant surgery through virtual reality simulation.

OBJECTIVES: To evaluate the effectiveness of a virtual reality (VR) temporal bone simulator in training cochlear implant surgery. METHODS: We compared the performance of 12 otolaryngology registrars conducting simulated cochlear implant surgery before (pre-test) and after (post-tests) receiving training on a VR temporal bone surgery simulator with automated performance feedback. The post-test tasks were two temporal bones, one that was a mirror image of the temporal bone used as a pre-test and the other, a novel temporal bone. Participant performances were assessed by an otologist with a validated cochlear implant competency assessment tool. Structural damage was derived from an automatically generated simulator metric and compared between time points. RESULTS: Wilcoxon signed-rank test showed that there was a significant improvement with a large effect size in the total performance scores between the pre-test (PT) and both the first and second post-tests (PT1, PT2) (PT-PT1: P = 0.007, r = 0.78, PT-PT2: P = 0.005, r = 0.82). CONCLUSION: The results of the study indicate that VR simulation with automated guidance can effectively be used to train surgeons in training complex temporal bone surgeries such as cochlear implantation.

Correlations of External Landmarks With Internal Structures of the Temporal Bone.

HYPOTHESIS: The internal anatomy of a temporal bone could be inferred from external landmarks. BACKGROUND: Mastoid surgery is an important skill that ENT surgeons need to acquire. Surgeons commonly use CT scans as a guide to understanding anatomical variations before surgery. Conversely, in cases where CT scans are not available, or in the temporal bone laboratory where residents are usually not provided with CT scans, it would be beneficial if the internal anatomy of a temporal bone could be inferred from external landmarks. METHODS: We explored correlations between internal anatomical variations and metrics established to quantify the position of external landmarks that are commonly exposed in the operating room, or the temporal bone laboratory, before commencement of drilling. Mathematical models were developed to predict internal anatomy based on external structures. RESULTS: From an operating room view, the distances between the following external landmarks were observed to have statistically significant correlations with the internal anatomy of a temporal bone: temporal line, external auditory canal, mastoid tip, occipitomastoid suture, and Henle's spine. These structures can be used to infer a low lying dura mater (p = 0.002), an anteriorly located sigmoid sinus (p = 0.006), and a more lateral course of the facial nerve (p < 0.001). In the temporal bone laboratory view, the mastoid tegmen and sigmoid sinus were also regarded as external landmarks. The distances between these two landmarks and the operating view external structures were able to further infer the laterality of the facial nerve (p < 0.001) and a sclerotic mastoid (p < 0.001). Two nonlinear models were developed that predicted the distances between the following internal structures with a high level of accuracy: the distance from the sigmoid sinus to the posterior external auditory canal (p < 0.001) and the diameter of the round window niche (p < 0.001). CONCLUSION: The prospect of encountering some of the more technically challenging anatomical variants encountered in temporal bone dissection can be inferred from the distance between external landmarks found on the temporal bone. These relationships could be used as a guideline to predict challenges during drilling and choosing appropriate temporal bones for dissection.

Developing effective automated feedback in temporal bone surgery simulation.

OBJECTIVE: We aim to test the effectiveness, accuracy, and usefulness of an automated feedback system in facilitating skill acquisition in virtual reality surgery. STUDY DESIGN: We evaluate the performance of the feedback system through a randomized controlled trial of 24 students allocated to feedback and nonfeedback groups. SETTING: The feedback system was based on the Melbourne University temporal bone surgery simulator. The study was conducted at the simulation laboratory of the Royal Victorian Eye and Ear Hospital, Melbourne. SUBJECTS AND METHODS: The study participants were medical students from the University of Melbourne, who were asked to perform virtual cortical mastoidectomy on the simulator. The extent to which the drilling behavior of the feedback and nonfeedback groups differed was used to evaluate the effectiveness of the system. Its accuracy was determined through a postexperiment observational assessment of recordings made during the experiment by an expert surgeon. Its usability was evaluated using students' self-reports of their impressions of the system. RESULTS: A Friedman's test showed that there was a significant improvement in the drilling performance of the feedback group, χ(2)(1) = 14.450, P < .001. The postexperiment assessment demonstrated that the system provided timely feedback (when trainee behavior was detected) 88.6% of the time and appropriate feedback (accurate advice) 84.2% of the time. Participants' opinions about the usefulness of the system were highly positive. CONCLUSION: The automated feedback system was observed to be effective in improving surgical technique, and the provided feedback was found to be accurate and useful.

The construct validity and reliability of an assessment tool for competency in cochlear implant surgery.

INTRODUCTION: We introduce a rating tool that objectively evaluates the skills of surgical trainees performing cochlear implant surgery. METHODS: Seven residents and seven experts performed cochlear implant surgery sessions from mastoidectomy to cochleostomy on a standardized virtual reality temporal bone. A total of twenty-eight assessment videos were recorded and two consultant otolaryngologists evaluated the performance of each participant using these videos. RESULTS: Interrater reliability was calculated using the intraclass correlation coefficient for both the global and checklist components of the assessment instrument. The overall agreement was high. The construct validity of this instrument was strongly supported by the significantly higher scores in the expert group for both components. CONCLUSION: Our results indicate that the proposed assessment tool for cochlear implant surgery is reliable, accurate, and easy to use. This instrument can thus be used to provide objective feedback on overall and task-specific competency in cochlear implantation.

A temporal bone surgery simulator with real-time feedback for surgical training.

Timely feedback on surgical technique is an important aspect of surgical skill training in any learning environment, be it virtual or otherwise. Feedback on technique should be provided in real-time to allow trainees to recognize and amend their errors as they occur. Expert surgeons have typically carried out this task, but they have limited time available to spend with trainees. Virtual reality surgical simulators offer effective, repeatable training at relatively low cost, but their benefits may not be fully realized while they still require the presence of experts to provide feedback. We attempt to overcome this limitation by introducing a real-time feedback system for surgical technique within a temporal bone surgical simulator. Our evaluation study shows that this feedback system performs exceptionally well with respect to accuracy and effectiveness.

Constructive real time feedback for a temporal bone simulator.

As demands on surgical training efficiency increase, there is a stronger need for computer assisted surgical training systems. The ability to provide automated performance feedback and assessment is a critical aspect of such systems. The development of feedback and assessment models will allow the use of surgical simulators as self-guided training systems that act like expert trainers and guide trainees towards improved performance. This paper presents an approach based on Random Forest models to analyse data recorded during surgery using a virtual reality temporal bone simulator and generate meaningful automated real-time performance feedback. The training dataset consisted of 27 temporal bone simulation runs composed of 16 expert runs provided by 7 different experts and 11 trainee runs provided by 6 trainees. We demonstrate how Random Forest models can be used to predict surgical expertise and deliver feedback that improves trainees' surgical technique. We illustrate the potential of the approach through a feasibility study.