Probe motion during mid-trimester fetal anomaly scan in the clinical setting: A prospective observational study.

INTRODUCTION: This study aims to investigate probe motion during full mid-trimester anomaly scans. METHODS: We undertook a prospective, observational study of obstetric sonographers at a UK University Teaching Hospital. We collected prospectively full-length video recordings of routine second-trimester anomaly scans synchronized with probe trajectory tracking data during the scan. Videos were reviewed and trajectories analyzed using duration, path metrics (path length, velocity, acceleration, jerk, and volume) and angular metrics (spectral arc, angular area, angular velocity, angular acceleration, and angular jerk). These trajectories were then compared according to the participant level of expertise, fetal presentation, and patient BMI. RESULTS: A total of 17 anomaly scans were recorded. The average velocity of the probe was 12.9 ± 3.4 mm/s for the consultants versus 24.6 ± 5.7 mm/s for the fellows (p = 0.02), the average acceleration 170.4 ± 26.3 mm/s2 versus 328.9 ± 62.7 mm/s2 (p = 0.02), and the average jerk 7491.7 ± 1056.1 mm/s3 versus 14944.1 ± 3146.3 mm/s3 (p = 0.02), the working volume 9.106 ± 4.106 mm3 versus 29.106 ± 11.106 mm3 (p = 0.03), respectively. The angular metrics were not significantly different according to the participant level of expertise, the fetal presentation, or to patients BMI. CONCLUSION: Some differences in the probe path metrics (velocity, acceleration, jerk and working volume) were noticed according to operator's level.

Understanding metric-related pitfalls in image analysis validation.

Validation metrics are key for tracking scientific progress and bridging the current chasm between artificial intelligence research and its translation into practice. However, increasing evidence shows that, particularly in image analysis, metrics are often chosen inadequately. Although taking into account the individual strengths, weaknesses and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multistage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides a reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Although focused on biomedical image analysis, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. The work serves to enhance global comprehension of a key topic in image analysis validation.

Metrics reloaded: recommendations for image analysis validation.

Increasing evidence shows that flaws in machine learning (ML) algorithm validation are an underestimated global problem. In biomedical image analysis, chosen performance metrics often do not reflect the domain interest, and thus fail to adequately measure scientific progress and hinder translation of ML techniques into practice. To overcome this, we created Metrics Reloaded, a comprehensive framework guiding researchers in the problem-aware selection of metrics. Developed by a large international consortium in a multistage Delphi process, it is based on the novel concept of a problem fingerprint-a structured representation of the given problem that captures all aspects that are relevant for metric selection, from the domain interest to the properties of the target structure(s), dataset and algorithm output. On the basis of the problem fingerprint, users are guided through the process of choosing and applying appropriate validation metrics while being made aware of potential pitfalls. Metrics Reloaded targets image analysis problems that can be interpreted as classification tasks at image, object or pixel level, namely image-level classification, object detection, semantic segmentation and instance segmentation tasks. To improve the user experience, we implemented the framework in the Metrics Reloaded online tool. Following the convergence of ML methodology across application domains, Metrics Reloaded fosters the convergence of validation methodology. Its applicability is demonstrated for various biomedical use cases.

A preliminary exploration into top-down and bottom-up deep-learning approaches to localising neuro-interventional point targets in volumetric MRI.

PURPOSE: Point localisation is a critical aspect of many interventional planning procedures, specifically representing anatomical regions of interest or landmarks as individual points. This could be seen as analogous to the problem of visual search in cognitive psychology, in which this search is performed either: bottom-up, constructing increasingly abstract and coarse-resolution features over the entire image; or top-down, using contextual cues from the entire image to refine the scope of the region being investigated. Traditional convolutional neural networks use the former, but it is not clear if this is optimal. This article is a preliminary investigation as to how this motivation affects 3D point localisation in neuro-interventional planning. METHODS: Two neuro-imaging datasets were collected: one for cortical point localisation for repetitive transcranial magnetic stimulation and the other for sub-cortical anatomy localisation for deep brain stimulation. Four different frameworks were developed using top-down versus bottom-up paradigms as well as representing points as co-ordinates or heatmaps. These networks were applied to point localisation for transcranial magnetic stimulation and subcortical anatomy localisation. These networks were evaluated using cross-validation and a varying number of training datasets to analyse their sensitivity to quantity of training data. RESULTS: Each network shows increasing performance as the amount of available training data increases, with the co-ordinate-based top-down network consistently outperforming the others. Specifically, the top-down architectures tend to outperform the bottom-up ones. An analysis of their memory consumption also encourages the top-down co-ordinate based architecture as it requires significantly less memory than either bottom-up architectures or those representing their predictions via heatmaps. CONCLUSION: This paper is a preliminary foray into a fundamental aspect of machine learning architectural design: that of the top-down/bottom-up divide from cognitive psychology. Although there are additional considerations within the particular architectures investigated that could affect these results and the number of architectures investigated is limited, our results do indicate that the less commonly used top-down paradigm could lead to more efficient and effective architectures in the future.

Understanding metric-related pitfalls in image analysis validation.

Validation metrics are key for the reliable tracking of scientific progress and for bridging the current chasm between artificial intelligence (AI) research and its translation into practice. However, increasing evidence shows that particularly in image analysis, metrics are often chosen inadequately in relation to the underlying research problem. This could be attributed to a lack of accessibility of metric-related knowledge: While taking into account the individual strengths, weaknesses, and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multi-stage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides the first reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Focusing on biomedical image analysis but with the potential of transfer to other fields, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. To facilitate comprehension, illustrations and specific examples accompany each pitfall. As a structured body of information accessible to researchers of all levels of expertise, this work enhances global comprehension of a key topic in image analysis validation.

Virtual reality simulation training improve diagnostic knee arthroscopy and meniscectomy skills: a prospective transfer validity study.

PURPOSE: Limited data exist on the actual transfer of skills learned using a virtual reality (VR) simulator for arthroscopy training because studies mainly focused on VR performance improvement and not on transfer to real word (transfer validity). The purpose of this single-blinded, controlled trial was to objectively investigate transfer validity in the context of initial knee arthroscopy training. METHODS: For this study, 36 junior resident orthopaedic surgeons (postgraduate year one and year two) without prior experience in arthroscopic surgery were enrolled to receive standard knee arthroscopy surgery training (NON-VR group) or standard training plus training on a hybrid virtual reality knee arthroscopy simulator (1 h/month) (VR group). At inclusion, all participants completed a questionnaire on their current arthroscopic technical skills. After 6 months of training, both groups performed three exercises that were evaluated independently by two blinded trainers: i) arthroscopic partial meniscectomy on a bench-top knee simulator; ii) supervised diagnostic knee arthroscopy on a cadaveric knee; and iii) supervised knee partial meniscectomy on a cadaveric knee. Training level was determined with the Arthroscopic Surgical Skill Evaluation Tool (ASSET) score. RESULTS: Overall, performance (ASSET scores) was better in the VR group than NON-VR group (difference in the global scores: p < 0.001, in bench-top meniscectomy scores: p = 0.03, in diagnostic knee arthroscopy on a cadaveric knee scores: p = 0.04, and in partial meniscectomy on a cadaveric knee scores: p = 0.02). Subgroup analysis by postgraduate year showed that the year-one NON-VR subgroup performed worse than the other subgroups, regardless of the exercise. CONCLUSION: This study showed the transferability of the technical skills acquired by novice residents on a hybrid virtual reality simulator to the bench-top and cadaveric models. Surgical skill acquired with a VR arthroscopy surgical simulator might safely improve arthroscopy competences in the operating room, also helping to standardise resident training and follow their progress.

sEMG-based Motion Recognition for Robotic Surgery Training - A Preliminary Study.

Robotic surgery represents a major breakthrough in the evolution of medical technology. Accordingly, efficient skill training and assessment methods should be developed to meet the surgeon's need of acquiring such robotic skills over a relatively short learning curve in a safe manner. Different from conventional training and assessment methods, we aim to explore the surface electromyography (sEMG) signal during the training process in order to obtain semantic and interpretable information to help the trainee better understand and improve his/her training performance. As a preliminary study, motion primitive recognition based on sEMG signal is studied in this work. Using machine learning (ML) technique, it is shown that the sEMG-based motion recognition method is feasible and promising for hand motions along 3 Cartesian axes in the virtual reality (VR) environment of a commercial robotic surgery training platform, which will hence serve as the basis for new robotic surgical skill assessment criterion and training guidance based on muscle activity information. Considering certain motion patterns were less accurately recognized than others, more data collection and deep learning-based analysis will be carried out to further improve the recognition accuracy in future research.

Physical environment of the operating room during cesarean section: A systematic review.

INTRODUCTION: Environmental factors in the operating room during cesarean sections are likely important for both women/birthing people and their babies but there is currently a lack of rigorous literature about their evaluation. The principal aim of this study was to systematically examine studies published on the physical environment in the obstetrical operating room during c-sections and its impact on mother and neonate outcomes. The secondary objective was to identify the sensors used to investigate the operating room environment during cesarean sections. METHODS: In this literature review, we searched MEDLINE a database using the following keywords: Cesarean section AND (operating room environment OR Noise OR Music OR Video recording OR Light level OR Gentle OR Temperature OR Motion Data). Eligible studies had to be published in English or French within the past 10 years and had to investigate the operating room environment during cesarean sections in women. For each study we reported which aspects of the physical environment were investigated in the OR (i.e., noise, music, movement, light or temperature) and the involved sensors. RESULTS: Of a total of 105 studies screened, we selected 8 articles from title and abstract in PubMed. This small number shows that the field is poorly investigated. The most evaluated environment factors to date are operating room noise and temperature, and the presence of music. Few studies used advanced sensors in the operating room to evaluate environmental factors in a more nuanced and complete way. Two studies concern the sound level, four concern music, one concerns temperature and one analyzed the number of entrances/exits into the OR. No study analyzed light level or more fine-grained movement data. CONCLUSIONS: Main findings include increase of noise and motion at specific time-points, for example during delivery or anaesthesia; the positive impact of music on parents and staff alike; and that a warmer theatre is better for babies but more uncomfortable for surgeons.

SAGES consensus recommendations on surgical video data use, structure, and exploration (for research in artificial intelligence, clinical quality improvement, and surgical education).

BACKGROUND: Surgery generates a vast amount of data from each procedure. Particularly video data provides significant value for surgical research, clinical outcome assessment, quality control, and education. The data lifecycle is influenced by various factors, including data structure, acquisition, storage, and sharing; data use and exploration, and finally data governance, which encompasses all ethical and legal regulations associated with the data. There is a universal need among stakeholders in surgical data science to establish standardized frameworks that address all aspects of this lifecycle to ensure data quality and purpose. METHODS: Working groups were formed, among 48 representatives from academia and industry, including clinicians, computer scientists and industry representatives. These working groups focused on: Data Use, Data Structure, Data Exploration, and Data Governance. After working group and panel discussions, a modified Delphi process was conducted. RESULTS: The resulting Delphi consensus provides conceptualized and structured recommendations for each domain related to surgical video data. We identified the key stakeholders within the data lifecycle and formulated comprehensive, easily understandable, and widely applicable guidelines for data utilization. Standardization of data structure should encompass format and quality, data sources, documentation, metadata, and account for biases within the data. To foster scientific data exploration, datasets should reflect diversity and remain adaptable to future applications. Data governance must be transparent to all stakeholders, addressing legal and ethical considerations surrounding the data. CONCLUSION: This consensus presents essential recommendations around the generation of standardized and diverse surgical video databanks, accounting for multiple stakeholders involved in data generation and use throughout its lifecycle. Following the SAGES annotation framework, we lay the foundation for standardization of data use, structure, and exploration. A detailed exploration of requirements for adequate data governance will follow.

Gaze behavior is related to objective technical skills assessment during virtual reality simulator-based surgical training: a proof of concept.

PURPOSE: Simulation-based training allows surgical skills to be learned safely. Most virtual reality-based surgical simulators address technical skills without considering non-technical skills, such as gaze use. In this study, we investigated surgeons' visual behavior during virtual reality-based surgical training where visual guidance is provided. Our hypothesis was that the gaze distribution in the environment is correlated with the simulator's technical skills assessment. METHODS: We recorded 25 surgical training sessions on an arthroscopic simulator. Trainees were equipped with a head-mounted eye-tracking device. A U-net was trained on two sessions to segment three simulator-specific areas of interest (AoI) and the background, to quantify gaze distribution. We tested whether the percentage of gazes in those areas was correlated with the simulator's scores. RESULTS: The neural network was able to segment all AoI with a mean Intersection over Union superior to 94% for each area. The gaze percentage in the AoI differed among trainees. Despite several sources of data loss, we found significant correlations between gaze position and the simulator scores. For instance, trainees obtained better procedural scores when their gaze focused on the virtual assistance (Spearman correlation test, N = 7, r = 0.800, p = 0.031). CONCLUSION: Our findings suggest that visual behavior should be quantified for assessing surgical expertise in simulation-based training environments, especially when visual guidance is provided. Ultimately visual behavior could be used to quantitatively assess surgeons' learning curve and expertise while training on VR simulators, in a way that complements existing metrics.

A systematic review of annotation for surgical process model analysis in minimally invasive surgery based on video.

BACKGROUND: Annotated data are foundational to applications of supervised machine learning. However, there seems to be a lack of common language used in the field of surgical data science. The aim of this study is to review the process of annotation and semantics used in the creation of SPM for minimally invasive surgery videos. METHODS: For this systematic review, we reviewed articles indexed in the MEDLINE database from January 2000 until March 2022. We selected articles using surgical video annotations to describe a surgical process model in the field of minimally invasive surgery. We excluded studies focusing on instrument detection or recognition of anatomical areas only. The risk of bias was evaluated with the Newcastle Ottawa Quality assessment tool. Data from the studies were visually presented in table using the SPIDER tool. RESULTS: Of the 2806 articles identified, 34 were selected for review. Twenty-two were in the field of digestive surgery, six in ophthalmologic surgery only, one in neurosurgery, three in gynecologic surgery, and two in mixed fields. Thirty-one studies (88.2%) were dedicated to phase, step, or action recognition and mainly relied on a very simple formalization (29, 85.2%). Clinical information in the datasets was lacking for studies using available public datasets. The process of annotation for surgical process model was lacking and poorly described, and description of the surgical procedures was highly variable between studies. CONCLUSION: Surgical video annotation lacks a rigorous and reproducible framework. This leads to difficulties in sharing videos between institutions and hospitals because of the different languages used. There is a need to develop and use common ontology to improve libraries of annotated surgical videos.

Reference-free Bayesian model for pointing errors of typein neurosurgical planning.

PURPOSE: Many neurosurgical planning tasks rely on identifying points of interest in volumetric images. Often, these points require significant expertise to identify correctly as, in some cases, they are not visible but instead inferred by the clinician. This leads to a high degree of variability between annotators selecting these points. In particular, errors of type are when the experts fundamentally select different points rather than the same point with some inaccuracy. This complicates research as their mean may not reflect any of the experts' intentions nor the ground truth. METHODS: We present a regularised Bayesian model for measuring errors of type in pointing tasks. This model is reference-free; in that it does not require a priori knowledge of the ground truth point but instead works on the basis of the level of consensus between multiple annotators. We apply this model to simulated data and clinical data from transcranial magnetic stimulation for chronic pain. RESULTS: Our model estimates the probabilities of selecting the correct point in the range of 82.6[Formula: see text]88.6% with uncertainties in the range of 2.8[Formula: see text]4.0%. This agrees with the literature where ground truth points are known. The uncertainty has not previously been explored in the literature and gives an indication of the dataset's strength. CONCLUSIONS: Our reference-free Bayesian framework easily models errors of type in pointing tasks. It allows for clinical studies to be performed with a limited number of annotators where the ground truth is not immediately known, which can be applied widely for better understanding human errors in neurosurgical planning.

PEg TRAnsfer Workflow recognition challenge report: Do multimodal data improve recognition?

BACKGROUND AND OBJECTIVE: In order to be context-aware, computer-assisted surgical systems require accurate, real-time automatic surgical workflow recognition. In the past several years, surgical video has been the most commonly-used modality for surgical workflow recognition. But with the democratization of robot-assisted surgery, new modalities, such as kinematics, are now accessible. Some previous methods use these new modalities as input for their models, but their added value has rarely been studied. This paper presents the design and results of the "PEg TRAnsfer Workflow recognition" (PETRAW) challenge with the objective of developing surgical workflow recognition methods based on one or more modalities and studying their added value. METHODS: The PETRAW challenge included a data set of 150 peg transfer sequences performed on a virtual simulator. This data set included videos, kinematic data, semantic segmentation data, and annotations, which described the workflow at three levels of granularity: phase, step, and activity. Five tasks were proposed to the participants: three were related to the recognition at all granularities simultaneously using a single modality, and two addressed the recognition using multiple modalities. The mean application-dependent balanced accuracy (AD-Accuracy) was used as an evaluation metric to take into account class balance and is more clinically relevant than a frame-by-frame score. RESULTS: Seven teams participated in at least one task with four participating in every task. The best results were obtained by combining video and kinematic data (AD-Accuracy of between 93% and 90% for the four teams that participated in all tasks). CONCLUSION: The improvement of surgical workflow recognition methods using multiple modalities compared with unimodal methods was significant for all teams. However, the longer execution time required for video/kinematic-based methods(compared to only kinematic-based methods) must be considered. Indeed, one must ask if it is wise to increase computing time by 2000 to 20,000% only to increase accuracy by 3%. The PETRAW data set is publicly available at www.synapse.org/PETRAW to encourage further research in surgical workflow recognition.

Virtual reality environments to train soft skills in medical and nursing education: a technical feasibility study between France and Japan.

PURPOSE: To meet the urgent and massive training needs of healthcare professionals, the use of digital technologies is proving increasingly relevant, and the rise of digital training platforms shows their usefulness and possibilities. However, despite the impact of these platforms on the medical skills learning, cultural differences are rarely factored in the implementation of these training environments. METHODS: By using the Scrub Nurse Non-Technical Skills Training System (SunSet), we developed a methodology enabling the adaptation of a virtual reality-based environment and scenarios from French to Japanese cultural and medical practices. We then conducted a technical feasibility study between France and Japan to assess virtual reality simulations acceptance among scrub nurses. RESULTS: Results in term of acceptance do not reveal major disparity between both populations, and the only emerging significant difference between both groups is on the Behavioral Intention, which is significantly higher for the French scrub nurses. In both cases, participants had a positive outlook. CONCLUSION: The findings suggest that the methodology we have implemented can be further used in the context of cultural adaptation of non-technical skills learning scenarios in virtual environments for the training and assessment of health care personnel.

Behavioural disinhibition in frontotemporal dementia investigated within an ecological framework.

Disinhibition is a core symptom in behavioural variant frontotemporal dementia (bvFTD) particularly affecting the daily lives of both patients and caregivers. Yet, characterisation of inhibition disorders is still unclear and management options of these disorders are limited. Questionnaires currently used to investigate behavioural disinhibition do not differentiate between several subtypes of disinhibition, encompass observation biases and lack of ecological validity. In the present work, we explored disinhibition in an original semi-ecological situation, by distinguishing three categories of disinhibition: compulsivity, impulsivity and social disinhibition. First, we measured prevalence and frequency of these disorders in 23 bvFTD patients and 24 healthy controls (HC) in order to identify the phenotypical heterogeneity of disinhibition. Then, we examined the relationships between these metrics, the neuropsychological scores and the behavioural states to propose a more comprehensive view of these neuropsychiatric manifestations. Finally, we studied the context of occurrence of these disorders by investigating environmental factors potentially promoting or reducing them. As expected, we found that patients were more compulsive, impulsive and socially disinhibited than HC. We found that 48% of patients presented compulsivity (e.g., repetitive actions), 48% impulsivity (e.g., oral production) and 100% of the patients group showed social disinhibition (e.g., disregards for rules or investigator). Compulsivity was negatively related with emotions recognition. BvFTD patients were less active if not encouraged in an activity, and their social disinhibition decreased as activity increased. Finally, impulsivity and social disinhibition decreased when patients were asked to focus on a task. Summarising, this study underlines the importance to differentiate subtypes of disinhibition as well as the setting in which they are exhibited, and points to stimulating area for non-pharmacological management.

Deep neural network architecture for automated soft surgical skills evaluation using objective structured assessment of technical skills criteria.

PURPOSE: Classic methods of surgery skills evaluation tend to classify the surgeon performance in multi-categorical discrete classes. If this classification scheme has proven to be effective, it does not provide in-between evaluation levels. If these intermediate scoring levels were available, they would provide more accurate evaluation of the surgeon trainee. METHODS: We propose a novel approach to assess surgery skills on a continuous scale ranging from 1 to 5. We show that the proposed approach is flexible enough to be used either for scores of global performance or several sub-scores based on a surgical criteria set called Objective Structured Assessment of Technical Skills (OSATS). We established a combined CNN+BiLSTM architecture to take advantage of both temporal and spatial features of kinematic data. Our experimental validation relies on real-world data obtained from JIGSAWS database. The surgeons are evaluated on three tasks: Knot-Tying, Needle-Passing and Suturing. The proposed framework of neural networks takes as inputs a sequence of 76 kinematic variables and produces an output float score ranging from 1 to 5, reflecting the quality of the performed surgical task. RESULTS: Our proposed model achieves high-quality OSATS scores predictions with means of Spearman correlation coefficients between the predicted outputs and the ground-truth outputs of 0.82, 0.60 and 0.65 for Knot-Tying, Needle-Passing and Suturing, respectively. To our knowledge, we are the first to achieve this regression performance using the OSATS criteria and the JIGSAWS kinematic data. CONCLUSION: An effective deep learning tool was created for the purpose of surgical skills assessment. It was shown that our method could be a promising surgical skills evaluation tool for surgical training programs.

"Stand-up straight!": human pose estimation to evaluate postural skills during orthopedic surgery simulations.

PURPOSE: Surgery simulators can be used to learn technical and non-technical skills and, to analyse posture. Ergonomic skill can be automatically detected with a Human Pose Estimation algorithm to help improve the surgeon's work quality. The objective of this study was to analyse the postural behaviour of surgeons and identify expertise-dependent movements. Our hypothesis was that hesitation and the occurrence of surgical instruments interfering with movement (defined as interfering movements) decrease with expertise. MATERIAL AND METHODS: Sixty surgeons with three expertise levels (novice, intermediate, and expert) were recruited. During a training session using an arthroscopic simulator, each participant's movements were video-recorded with an RGB camera. A modified OpenPose algorithm was used to detect the surgeon's joints. The detection frequency of each joint in a specific area was visualized with a heatmap-like approach and used to calculate a mobility score. RESULTS: This analysis allowed quantifying surgical movements. Overall, the mean mobility score was 0.823, 0.816, and 0.820 for novice, intermediate and expert surgeons, respectively. The mobility score alone was not enough to identify postural behaviour differences. A visual analysis of each participants' movements highlighted expertise-dependent interfering movements. CONCLUSION: Video-recording and analysis of surgeon's movements are a non-invasive approach to obtain quantitative and qualitative ergonomic information in order to provide feedback during training. Our findings suggest that the interfering movements do not decrease with expertise but differ in function of the surgeon's level.

Combining simple interactivity and machine learning: a separable deep learning approach to subthalamic nucleus localization and segmentation in MRI for deep brain stimulation surgical planning.

Purpose: Deep brain stimulation (DBS) is an interventional treatment for some neurological and neurodegenerative diseases. For example, in Parkinson's disease, DBS electrodes are positioned at particular locations within the basal ganglia to alleviate the patient's motor symptoms. These interventions depend greatly on a preoperative planning stage in which potential targets and electrode trajectories are identified in a preoperative MRI. Due to the small size and low contrast of targets such as the subthalamic nucleus (STN), their segmentation is a difficult task. Machine learning provides a potential avenue for development, but it has difficulty in segmenting such small structures in volumetric images due to additional problems such as segmentation class imbalance. Approach: We present a two-stage separable learning workflow for STN segmentation consisting of a localization step that detects the STN and crops the image to a small region and a segmentation step that delineates the structure within that region. The goal of this decoupling is to improve accuracy and efficiency and to provide an intermediate representation that can be easily corrected by a clinical user. This correction capability was then studied through a human-computer interaction experiment with seven novice participants and one expert neurosurgeon. Results: Our two-step segmentation significantly outperforms the comparative registration-based method currently used in clinic and approaches the fundamental limit on variability due to the image resolution. In addition, the human-computer interaction experiment shows that the additional interaction mechanism allowed by separating STN segmentation into two steps significantly improves the users' ability to correct errors and further improves performance. Conclusions: Our method shows that separable learning not only is feasible for fully automatic STN segmentation but also leads to improved interactivity that can ease its translation into clinical use.

Voxel-based diktiometry: Combining convolutional neural networks with voxel-based analysis and its application in diffusion tensor imaging for Parkinson's disease.

Extracting population-wise information from medical images, specifically in the neurological domain, is crucial to better understanding disease processes and progression. This is frequently done in a whole-brain voxel-wise manner, in which a population of patients and healthy controls are registered to a common co-ordinate space and a statistical test is performed on the distribution of image intensities for each location. Although this method has yielded a number of scientific insights, it is further from clinical applicability as the differences are often small and altogether do not permit for a high-performing classifier. In this article, we take the opposite approach of using a high-performing classifier, specifically a traditional convolutional neural network, and then extracting insights from it which can be applied in a population-wise manner, a method we call voxel-based diktiometry. We have applied this method to diffusion tensor imaging (DTI) analysis for Parkinson's disease (PD), using the Parkinson's Progression Markers Initiative database. By using the network sensitivity information, we can decompose what elements of the DTI contribute the most to the network's performance, drawing conclusions about diffusion biomarkers for PD that are based on metrics which are not readily expressed in the voxel-wise approach.