Virtual urgent care is here to stay: driving toward safe, equitable, and sustainable integration within emergency medicine.

BACKGROUND: Virtual care in Canada rapidly expanded during the COVID-19 pandemic in a low-rules environment in response to pressing needs for ongoing access to care amid public health restrictions. Emergency medicine specialists now face the challenge of advising on which virtual urgent care services ought to remain as part of comprehensive emergency care. Consideration must be given to safe, quality, and appropriate care as well as issues of equitable access, public demand, and sustainability (financial and otherwise). The aim of this project was to summarize current literature and expert opinion and formulate recommendations on the path forward for virtual care in emergency medicine. METHODS: We formed a working group of emergency medicine physicians from across Canada working in a variety of practice settings. The virtual care working group conducted a scoping review of the literature and met monthly to discuss themes and develop recommendations. The final recommendations were circulated to stakeholders for input and subsequently presented at the 2023 Canadian Association of Emergency Physicians (CAEP) Academic Symposium for discussion, feedback, and refinement. RESULTS: The working group developed and reached unanimity on nine recommendations addressing the themes of system design, equity and accessibility, quality and patient safety, education and curriculum, financial models, and sustainability of virtual urgent care services in Canada. CONCLUSION: Virtual urgent care has become an established service in the Canadian health care system. Emergency medicine specialists are uniquely suited to provide leadership and guidance on the optimal delivery of these services to enhance and complement emergency care in Canada.

RéSUMé: CONTEXTE: Les soins virtuels au Canada ont rapidement pris de l'ampleur pendant la pandémie de COVID-19 dans un environnement où les règles sont peu strictes, en réponse aux besoins urgents d'accès continu aux soins dans un contexte de restrictions en santé publique. Les spécialistes de la médecine d'urgence sont maintenant confrontés au défi de conseiller sur les services de soins d'urgence virtuels qui devraient rester dans le cadre des soins d'urgence complets. Il faut tenir compte des soins sécuritaires, de qualité et appropriés, ainsi que des questions d'accès équitable, de la demande publique et de la durabilité (financière et autre). L'objectif de ce projet était de résumer la littérature actuelle et l'opinion d'experts et de formuler des recommandations sur la voie à suivre pour les soins virtuels en médecine d'urgence. MéTHODES: Nous avons formé un groupe de travail composé de médecins urgentistes de partout au Canada qui travaillent dans divers milieux de pratique. Le groupe de travail sur les soins virtuels a effectué un examen de la portée de la documentation et s'est réuni chaque mois pour discuter des thèmes et formuler des recommandations. Les recommandations finales ont été distribuées aux intervenants pour obtenir leurs commentaires, puis présentées au symposium universitaire 2023 de l'Association canadienne des médecins d'urgence (ACMU) pour discussion, rétroaction et perfectionnement. RéSULTATS: Le groupe de travail a élaboré et atteint l'unanimité sur neuf recommandations portant sur les thèmes de la conception du système, de l'équité et de l'accessibilité, de la qualité et de la sécurité des patients, de l'éducation et des programmes, des modèles financiers et de la viabilité des services virtuels de soins d'urgence au Canada. CONCLUSION : Les soins d'urgence virtuels sont devenus un service établi dans le système de santé canadien. Les spécialistes en médecine d'urgence sont particulièrement bien placés pour fournir un leadership et des conseils sur la prestation optimale de ces services afin d'améliorer et de compléter les soins d'urgence au Canada.

The association between paramedic service system hospital offload time and response time.

OBJECTIVE: To address an important care issue in Canada, we tested the association between paramedic system hospital offload and response time, while considering the impact of other system-level factors. METHODS: Data from Calgary, Alberta (2014-2017), included median offload (exposure) and response (outcome) time aggregated by hour, with covariates paramedic system episodes of care-dispatch and arrival of a response unit-and hospital transport arrivals (collectively called volume), time of day, and season. Analyses used linear regression and modified Poisson models. RESULTS: 301,105 EMS episodes of care over 26,193 1-h periods were included. For any given 1-h period, the median (IQR) across all episodes of care for offload time, response time, episodes of care, and hospital transport arrivals were 55.3 (45.7, 66.3) min, 8.6 (7.6, 9.8) min, 12 (8, 16) episodes, and 8 (5, 10) hospital arrivals, respectively. Multivariable modelling revealed a complex association differing over levels of exposure and covariates, requiring description using "light stress" and "heavy stress" system scenarios. The light scenario was defined as median offload of 30 min and volume < 10th percentile (six episodes and four hospital arrivals), in the summer, and the heavy scenario as median offload of 90 min and volume > 90th percentile (17 episodes and 13 hospital arrivals), in the winter. An increase is reported in minutes:seconds for median hourly response time between scenarios by time of day: 1:04-4:16 (0000-0559 h.), 0:42-2:05 (0600-1159 h.), 0:57-3:01 (1200-1759 h.), and 0:18-2:21 (1800-2359 h.). CONCLUSIONS: Increasing offload is associated with increased response time; however the relationship is complex, with a greater impact on response time noted in select situations such as high volume in the winter. These observations illustrate the interdependence of paramedic, ED, and inpatient systems and provide high-yield targets for polices to mitigate the risk to community availability of paramedic resources at times of high offload delay/system stress.

ABSTRAIT: OBJECTIF: Afin de régler un problème important de soins au Canada, nous avons testé l'association entre le déchargement du système paramédical et le temps de réponse, tout en tenant compte de l'incidence d'autres facteurs au niveau du système. MéTHODES: Les données de Calgary, en Alberta (2014-2017) incluent le temps médian de déchargement (exposition) et de réponse (résultat) agrégé par heure, qui s'agit co-variables épisodes de soins du système paramédical - répartition et arrivée d'une unité d'intervention - et arrivées de transport hospitalier (collectivement appelé volume), l'heure et la saison. Les analyses ont utilisé la régression linéaire et des modèles de Poisson modifiés. RéSULTATS: 301105 épisodes de soins médicaux d'urgence sur 26193 périodes d'une heure ont été inclus. Pour une période d'une heure donnée, la médiane (QRI) pour tous les épisodes de soins pour le temps de déchargement, le temps de réponse, les épisodes de soins et les arrivées par transport à l'hôpital était de 55,3 (45,7, 66,3) minutes, 8,6 (7,6, 9,8) minutes, 12 (8, 16) épisodes et 8 (5, 10) arrivées à l'hôpital, respectivement. La modélisation multi-variable a révélé une association complexe qui varie selon les niveaux d'exposition et les co-variables, et qui nécessite une description à l'aide de scénarios de systèmes de « stress léger ¼ et de « stress lourd ¼. Le scénario léger a été défini comme un déchargement médian de 30 minutes, volume inférieur au 10e percentile (six épisodes et quatre arrivées à l'hôpital), pendant l'été. Le scénario lourd comme déchargement médian de 90 minutes, volume > 90e percentile (17 épisodes et 13 arrivées à l'hôpital), en hiver. Une augmentation est rapportée en minutes: secondes pour le temps de réponse horaire médian entre des scénarios par heure du jour : 1:04-4:16 (0000-0559 h.), 0:42-2:05 (0600-1159 h.), 0:57-3:01 (1200-1759 h.), et 0:18-2:21 (1800-2359 h.). CONCLUSIONS: L'augmentation du déchargement est associée à une augmentation du temps de réponse, mais la relation est complexe, avec un impact plus important sur le temps de réponse noté dans certaines situations, comme un volume élevé en hiver. Ces observations illustrent l'interdépendance des systèmes paramédicaux, des services d'urgence et des services aux patients hospitalisés et fournissent des cibles à haut rendement pour les politiques afin d'atténuer le risque pour la disponibilité des ressources paramédicales dans la collectivité en période de retard élevé ou de stress systémique.

Comparing ensemble methods combined with different aggregating models using micrograph cell segmentation as an initial application example.

Strategies such as ensemble learning and averaging techniques try to reduce the variance of single deep neural networks. The focus of this study is on ensemble averaging techniques, fusing the results of differently initialized and trained networks. Thereby, using micrograph cell segmentation as an application example, various ensembles have been initialized and formed during network training, whereby the following methods have been applied: (a) random seeds, (b) L 1-norm pruning, (c) variable numbers of training examples, and (d) a combination of the latter 2 items. Furthermore, different averaging methods are in common use and were evaluated in this study. As averaging methods, the mean, the median, and the location parameter of an alpha-stable distribution, fit to the histograms of class membership probabilities (CMPs), as well as a majority vote of the members of an ensemble were considered. The performance of these methods is demonstrated and evaluated on a micrograph cell segmentation use case, employing a common state-of-the art deep convolutional neural network (DCNN) architecture exploiting the principle of the common VGG-architecture. The study demonstrates that for this data set, the choice of the ensemble averaging method only has a marginal influence on the evaluation metrics (accuracy and Dice coefficient) used to measure the segmentation performance. Nevertheless, for practical applications, a simple and fast estimate of the mean of the distribution is highly competitive with respect to the most sophisticated representation of the CMP distributions by an alpha-stable distribution, and hence seems the most proper ensemble averaging method to be used for this application.

Number of necessary training examples for Neural Networks with different number of trainable parameters.

In this work, the network complexity should be reduced with a concomitant reduction in the number of necessary training examples. The focus thus was on the dependence of proper evaluation metrics on the number of adjustable parameters of the considered deep neural network. The used data set encompassed Hematoxylin and Eosin (H&E) colored cell images provided by various clinics. We used a deep convolutional neural network to get the relation between a model's complexity, its concomitant set of parameters, and the size of the training sample necessary to achieve a certain classification accuracy. The complexity of the deep neural networks was reduced by pruning a certain amount of filters in the network. As expected, the unpruned neural network showed best performance. The network with the highest number of trainable parameter achieved, within the estimated standard error of the optimized cross-entropy loss, best results up to 30% pruning. Strongly pruned networks are highly viable and the classification accuracy declines quickly with decreasing number of training patterns. However, up to a pruning ratio of 40%, we found a comparable performance of pruned and unpruned deep convolutional neural networks (DCNN) and densely connected convolutional networks (DCCN).

Forecasting brain activity based on models of spatiotemporal brain dynamics: A comparison of graph neural network architectures.

Comprehending the interplay between spatial and temporal characteristics of neural dynamics can contribute to our understanding of information processing in the human brain. Graph neural networks (GNNs) provide a new possibility to interpret graph-structured signals like those observed in complex brain networks. In our study we compare different spatiotemporal GNN architectures and study their ability to model neural activity distributions obtained in functional MRI (fMRI) studies. We evaluate the performance of the GNN models on a variety of scenarios in MRI studies and also compare it to a VAR model, which is currently often used for directed functional connectivity analysis. We show that by learning localized functional interactions on the anatomical substrate, GNN-based approaches are able to robustly scale to large network studies, even when available data are scarce. By including anatomical connectivity as the physical substrate for information propagation, such GNNs also provide a multimodal perspective on directed connectivity analysis, offering a novel possibility to investigate the spatiotemporal dynamics in brain networks.

Linkage of Emergency Medical Services and Hospital Data: A Necessary Precursor to Improve Understanding of Outcomes of Prehospital Care.

Objective: Linking emergency medical services (EMS) data to hospital outcomes is important for quality assurance and research initiatives. However, non-linkage due to missing or incomplete patient information may increase the risk of bias and distort findings. The purpose of this study was to explore if an optimization strategy, in addition to an existing linkage process, improved the linkage rate and reduced selection and information bias. Methods: 4,150 transported patients in a metropolitan EMS system in Alberta, Canada from 2016/17 were linked to two Emergency Department (ED) databases by a standard strategy using a unique health care number, date/time of ED arrival, and hospital name. An optimized strategy added additional linkage steps incorporating last name, year of birth, and a manual search. The strategies were compared to assess the rate of linkage, and to describe event and patient-level characteristics of unlinked records. Results: The standard strategy resulted in 3,650 out of 4,150 (88.0%) linked records (95% CI 86.9%-88.9%). Of the 500 non-linked records, an additional 381 were linked by the optimized strategy (n = 4,031/4,150 [97.1%; 95% CI: 96.6%-97.6%]). There were no false positive linkages. The highest linkage failure was in 25 to 34 year-old patients (n = 93/478, 19.5%), males (n = 236/1975, 12.0%), Echo level events (n = 15/77, 19.5%), and emergency transport (45/231, 19.5%). The optimized strategy improved linkage in these groups by 68.8% (64/93), 79.2% (187/236), 40.0% (6/15), and 51.1% (23/45) respectively. For dispatch card, the highest linkage failure occurred in Card 24-Pregnancy/Childbirth/Miscarriage (n = 30/44, 68.2%), Card 27-Stab/Gunshot/Penetrating Trauma (n = 6/17, 35.3%), and Card 9-Cardiac/Respiratory Arrest/Death (n = 12/46, 26.1%). The optimized strategy improved linkage by 10.0% (3/30), 83.3% (5/6), and 41.7% (5/12) respectively. For the 119 unlinked records, 71 (59.7%) had sufficient information for linkage, but no appropriately matching records could be found. Conclusion: An optimized sequential deterministic strategy linking EMS data to ED outcomes improved the linkage rate without increasing the number of false positive links, and reduced the potential for bias. Even with adequate information, some records were not linked to their ED visit. This study underscores the importance of understanding how data are linked to hospital outcomes in EMS research and the potential for bias.

Safety of elective paediatric surgery during the coronavirus disease 2019 pandemic.

INTRODUCTION: Corona-virus Disease 2019 (COVID-19) has had a huge impact on the delivery of healthcare worldwide, particularly elective surgery. There is a lack of data regarding risk of postoperative COVID-19 infection in children undergoing elective surgery, and regarding the utility of pre-operative COVID-19 testing, and preoperative "cocooning" or restriction of movements. The purpose of this present study was to examine the safety of elective paediatric Otolaryngology surgery during the COVID-19 pandemic with respect to incidence of postoperative symptomatic COVID-19 infection or major respiratory complications. MATERIALS AND METHODS: Prospective cohort study of paediatric patients undergoing elective Otolaryngology surgery between September and December 2020. Primary outcome measure was incidence of symptomatic COVID-19 or major respiratory complications within the 14 days after surgery. Parents of prospectively enrolled patients were contacted 14 days after surgery and enquiry made regarding development of postoperative symptoms, COVID-19 testing, or diagnosis of COVID-19. RESULTS: 302 patients were recruited. 125 (41.4%) underwent preoperative COVID-19 RT-PCR testing. 66 (21.8%) restricted movements prior to surgery. The peak 14-day COVID-19 incidence during the study was 302.9 cases per 100,000 population. No COVID-19 infections or major respiratory complications were reported in the 14 day follow-up period. CONCLUSION: The results of our study support the safety of elective paediatric Otolaryngology surgery during the pandemic, in the setting of community incidence not exceeding that observed during the study period.

Influence of concomitant injuries on post-concussion symptoms after a mild traumatic brain injury - a prospective multicentre cohort study.

Objectives: To compare post-concussion symptoms (PCS) and return to normal activities between mild Traumatic Brain Injury (mTBI) patients with or without concomitant injuries at 7-and 90 days post-mTBI.Methods: Design: Sub-analysis of a multicentre prospective cohort study. PARTICIPANTS AND SETTING: patients with mTBI from 7 Canadian Emergency Departments. PROCEDURE: Research assistants conducted telephone follow-ups using the Rivermead Postconcussion Symptoms Questionnaire (RPQ) at 7-, 30- and 90 days post-mTBI. MAIN OUTCOME: Presence of PCS (RPQ: ≥3 symptoms) at 90 days. SECONDARY OUTCOMES: RPQ score ≥21, prevalence of individual RPQ symptoms and patients' return to normal activities, at 7- and 90-days. Adjusted risk ratios (RR) were calculated.Results: 1725 mTBI patients were included and 1055 (61.1%) had concomitant injuries. Patients with concomitant injuries were at higher risk of having ≥3 symptoms on the RPQ (RR:1.26 [95% CI 1.01-1.58]) at 90 days. They were also at higher risk of experiencing specific symptoms (dizziness, fatigue, headaches and taking longer to think) and of non-return to their normal activities (RR:2.11 [95% CI 1.30-3.45]).Conclusion: Patients with concomitant injuries have slightly more PCS and seemed to be at higher risk of non-return to their normal activities 90 days, compared to patients without concomitant injuries.

A graph neural network framework for causal inference in brain networks.

A central question in neuroscience is how self-organizing dynamic interactions in the brain emerge on their relatively static structural backbone. Due to the complexity of spatial and temporal dependencies between different brain areas, fully comprehending the interplay between structure and function is still challenging and an area of intense research. In this paper we present a graph neural network (GNN) framework, to describe functional interactions based on the structural anatomical layout. A GNN allows us to process graph-structured spatio-temporal signals, providing a possibility to combine structural information derived from diffusion tensor imaging (DTI) with temporal neural activity profiles, like that observed in functional magnetic resonance imaging (fMRI). Moreover, dynamic interactions between different brain regions discovered by this data-driven approach can provide a multi-modal measure of causal connectivity strength. We assess the proposed model's accuracy by evaluating its capabilities to replicate empirically observed neural activation profiles, and compare the performance to those of a vector auto regression (VAR), like that typically used in Granger causality. We show that GNNs are able to capture long-term dependencies in data and also computationally scale up to the analysis of large-scale networks. Finally we confirm that features learned by a GNN can generalize across MRI scanner types and acquisition protocols, by demonstrating that the performance on small datasets can be improved by pre-training the GNN on data from an earlier study. We conclude that the proposed multi-modal GNN framework can provide a novel perspective on the structure-function relationship in the brain. Accordingly this approach appears to be promising for the characterization of the information flow in brain networks.

Risk of disease transmission from flexible nasoendoscopy during the coronavirus disease 2019 pandemic.

BACKGROUND: Concerns have emerged regarding infection transmission during flexible nasoendoscopy. METHODS: Information was gathered prospectively on flexible nasoendoscopy procedures performed between March and June 2020. Patients and healthcare workers were followed up to assess for coronavirus disease 2019 development. One-sided 97.5 per cent Poisson confidence intervals were calculated for upper limits of risk where zero events were observed. RESULTS: A total of 286 patients were recruited. The most common indication for flexible nasoendoscopy was investigation of 'red flag' symptoms (67 per cent). Forty-seven patients (16 per cent, 95 per cent confidence interval = 13-21 per cent) had suspicious findings on flexible nasoendoscopy requiring further investigation. Twenty patients (7.1 per cent, 95 per cent confidence interval = 4.4-11 per cent) had new cancer diagnoses. Zero coronavirus disease 2019 infections were recorded in the 273 patients. No. 27 endoscopists (the doctors and nurses who carried out the procedures) were followed up.The risk of developing coronavirus disease 2019 after flexible nasoendoscopy was determined to be 0-1.3 per cent. CONCLUSION: The risk of coronavirus disease 2019 transmission associated with performing flexible nasoendoscopy in asymptomatic patients, while using appropriate personal protective equipment, is very low. Additional data are required to confirm these findings in the setting of further disease surges.

The International Federation for Emergency Medicine report on emergency department crowding and access block: a brief summary.

OBJECTIVE: To develop comprehensive guidance that captures international impacts, causes, and solutions related to emergency department crowding and access block METHODS: Emergency physicians representing 15 countries from all IFEM regions composed the Task Force. Monthly meetings were held via video-conferencing software to achieve consensus for report content. The report was submitted and approved by the IFEM Board on June 1, 2020. RESULTS: A total of 14 topic dossiers, each relating to an aspect of ED crowding, were researched and completed collaboratively by members of the Task Force. CONCLUSIONS: The IFEM report is a comprehensive document intended to be used in whole or by section to inform and address aspects of ED crowding and access block. Overall, ED crowding is a multifactorial issue requiring systems-wide solutions applied at local, regional, and national levels. Access block is the predominant contributor of ED crowding in most parts of the world.

Designing a flexible tool for rapid implementation of brain-computer interfaces (BCI) in game development.

Early neurological injury or disease can lead to severe life-long physical impairments, despite normal cognitive function. For such individuals, brain-computer interfaces (BCI) may provide a means to regain access to the world by offering control of systems through directly processing brain patterns. However, current BCI applications are often research driven and consequently seen as uninteresting, particularly for prolonged use and younger BCI-users. To help mitigate this concern, this paper establishes a tool for researchers and game developers alike to rapidly incorporate a BCI control scheme (the P300 oddball response) into a gaming environment. Preliminary results indicate the proposed P300 Dynamic Cube (PDC) asset works in online BCI environments (n=20, healthy adult participants), resulting in median classification accuracy of 75 ± 3.28%. Additionally, the PDC tool can be rapidly adapted for a variety of game designs, evidenced by its incorporation into submissions to the Brain-Computer Interface (BCI) Game Jam 2019 competition. These findings support the PDC as a useful asset in the design and development of BCI-based games.

A Child's Right to Play: Results from the Brain-Computer Interface Game Jam 2019 (Calgary Competition).

Children with severe neurological disabilities may be unable to communicate or interact with their environments, depriving them of their right to play. Brain-computer interfaces (BCI) offer a means for such children to control external devices using only their brain signals, thereby introducing new opportunities for interaction. We organized the first North American BCI Game Jam to incite the development of BCI-compatible games for children. Nine games were submitted by 30 participants across North America. Games were judged by researchers and disabled children currently using BCI. Preliminary results demonstrate variety in game criteria preferences amongst the children who judged the games. The BCI Game Jam demonstrated promising potential for the creation of enjoyable games to suit the individual needs and preferences of children with severe neurological disabilities.

Three-dimensional Monte Carlo-based voxel-wise tumor dosimetry in patients with neuroendocrine tumors who underwent 177Lu-DOTATOC therapy.

BACKGROUND: Patients with advanced neuroendocrine tumors (NETs) of the midgut are suitable candidates for 177Lu-DOTATOC therapy. Integrated SPECT/CT systems have the potential to help improve the accuracy of patient-specific tumor dosimetry. Dose estimations to target organs are generally performed using the Medical Internal Radiation Dose scheme. We present a novel Monte Carlo-based voxel-wise dosimetry approach to determine organ- and tumor-specific total tumor doses (TTD). METHODS: A cohort of 14 patients with histologically confirmed metastasized NETs of the midgut (11 men, 3 women, 62.3 ± 11.0 years of age) underwent a total of 39 cycles of 177Lu-DOTATOC therapy (mean 2.8 cycles, SD ± 1 cycle). After the first cycle of therapy, regions of interest were defined manually on the SPECT/CT images for the kidneys, the spleen, and all 198 tracer-positive tumor lesions in the field of view. Four SPECT images, taken at 4 h, 24 h, 48 h and 72 h after injection of the radiopharmaceutical, were used to determine their effective half-lives in the structures of interest. The absorbed doses were calculated by a three-dimensional dosimetry method based on Monte Carlo simulations. TTD was calculated as the sum of all products of single tumor doses with single tumor volumes divided by the sum of all tumor volumes. RESULTS: The average dose values per cycle were 3.41 ± 1.28 Gy (1.91-6.22 Gy) for the kidneys, 4.40 ± 2.90 Gy (1.14-11.22 Gy) for the spleen, and 9.70 ± 8.96 Gy (1.47-39.49 Gy) for all 177Lu-DOTATOC-positive tumor lesions. Low- and intermediate-grade tumors (G 1-2) absorbed a higher TTD compared to high-grade tumors (G 3) (signed-rank test, p = < 0.05). The pre-therapeutic chromogranin A (CgA) value and the TTD correlated significantly (Pearson correlation: = 0.67, p = 0.01). Higher TTD resulted in a significant decrease of CgA after therapy. CONCLUSION: These results suggest that Monte Carlo-based voxel-wise dosimetry is a very promising tool for predicting the absorbed TTD based on histological and clinical parameters.

A deep learning approach to radiation dose estimation.

Currently methods for predicting absorbed dose after administering a radiopharmaceutical are rather crude in daily clinical practice. Most importantly, individual tissue density distributions as well as local variations of the concentration of the radiopharmaceutical are commonly neglected. The current study proposes machine learning techniques like Green's function-based empirical mode decomposition and deep learning methods on U-net architectures in conjunction with soft tissue kernel Monte Carlo (MC) simulations to overcome current limitations in precision and reliability of dose estimations for clinical dosimetric applications. We present a hybrid method (DNN-EMD) based on deep neural networks (DNN) in combination with empirical mode decomposition (EMD) techniques. The algorithm receives x-ray computed tomography (CT) tissue density maps and dose maps, estimated according to the MIRD protocol, i.e. employing whole organ S-values and related time-integrated activities (TIAs), and from measured SPECT distributions of 177Lu radionuclei, and learns to predict individual absorbed dose distributions. In a second step, density maps are replaced by their intrinsic modes as deduced from an EMD analysis. The system is trained using individual full MC simulation results as reference. Data from a patient cohort of 26 subjects are reported in this study. The proposed methods were validated employing a leave-one-out cross-validation technique. Deviations of estimated dose from corresponding MC results corroborate a superior performance of the newly proposed hybrid DNN-EMD method compared to its related MIRD DVK dose calculation. Not only are the mean deviations much smaller with the new method, but also the related variances are much reduced. If intrinsic modes of the tissue density maps are input to the algorithm, variances become even further reduced though the mean deviations are less affected. The newly proposed hybrid DNN-EMD method for individualized radiation dose prediction outperforms the MIRD DVK dose calculation method. It is fast enough to be of use in daily clinical practice.

A comparison of methods for adapting [Formula: see text] dose-voxel-kernels to tissue inhomogeneities.

In [Formula: see text] radionuclide therapies, dosimetry is used for determining patient-individual dose burden. Standard approaches provide whole organ doses only. For assessing dose heterogeneity inside organs, voxel-wise dosimetry based on 3D SPECT/CT imaging could be applied. Often, this is achieved by convolving voxel-wise time-activity-curves with appropriate dose-voxel-kernels (DVK). The DVKs are meant to model dose deposition, and can be more accurate if modelled for the specific tissue type under consideration. In literature, DVKs are often not adapted to these inhomogeneities, or simple approximation schemes are applied. For 26 patients, which had previously undergone a [Formula: see text] -PSMA or -DOTATOC therapy, decay maps, mass-density maps as well as tissue-type maps were derived from SPECT/CT acquisitions. These were used for a voxel-based dosimetry based on convolution with DVKs (each of size [Formula: see text]) obtained by four different DVK methods proposed in literature. The simplest only considers a spatially constant soft-tissue DVK (herein named 'constant'), while others either take into account only the local density of the center voxel of the DVK (herein named 'center-voxel') or scale each voxel linearly according to the proper mass density deduced from the CT image (herein named 'density') or considered both the local mass density as well as the direct path between the center voxel and any voxel in its surrounding (herein named 'percentage'). Deviations between resulting dose values and those from full Monte-Carlo simulations (MC simulations) were compared for selected organs and tissue-types. For each DVK method, inter-patient variability was considerable showing both under- and over-estimation of energy dose compared to the MC result for all tissue densities higher than soft tissue. In kidneys and spleen, 'constant' and 'density'-scaled DVKs achieved estimated doses with smallest deviations to the full MC gold standard (∼[Formula: see text] underestimation). For low and high density tissue types such as lung and adipose or bone tissue, alternative DVK methods like 'center-voxel'- and 'percentage'- scaled achieved superior results, respectively. Concerning computational load, dose estimation with the DVK method 'constant' needs about 1.1 s per patient, center-voxel scaling amounts to 1.2 s, density scaling needs 1.4 s while percentage scaling consumes 860.3 s per patient. In this study encompassing a large patient cohort and four different DVK estimation methods, no single DVK-adaption method was consistently better than any other in case of soft tissue kernels. Hence in such cases the simplest DVK method, labeled 'constant', suffices. In case of tumors, often located in tissues of low (lung) or high (bone) density, more sophisticated DVK methods excel. The high inter-patient variability indicates that for evaluating new algorithms, a sufficiently large patient cohort needs to be involved.