Global attention based GNN with Bayesian collaborative learning for glomerular lesion recognition.

BACKGROUND: Glomerular lesions reflect the onset and progression of renal disease. Pathological diagnoses are widely regarded as the definitive method for recognizing these lesions, as the deviations in histopathological structures closely correlate with impairments in renal function. METHODS: Deep learning plays a crucial role in streamlining the laborious, challenging, and subjective task of recognizing glomerular lesions by pathologists. However, the current methods treat pathology images as data in regular Euclidean space, limiting their ability to efficiently represent the complex local features and global connections. In response to this challenge, this paper proposes a graph neural network (GNN) that utilizes global attention pooling (GAP) to more effectively extract high-level semantic features from glomerular images. The model incorporates Bayesian collaborative learning (BCL), enhancing node feature fine-tuning and fusion during training. In addition, this paper adds a soft classification head to mitigate the semantic ambiguity associated with a purely hard classification. RESULTS: This paper conducted extensive experiments on four glomerular datasets, comprising a total of 491 whole slide images (WSIs) and 9030 images. The results demonstrate that the proposed model achieves impressive F1 scores of 81.37%, 90.12%, 87.72%, and 98.68% on four private datasets for glomerular lesion recognition. These scores surpass the performance of the other models used for comparison. Furthermore, this paper employed a publicly available BReAst Carcinoma Subtyping (BRACS) dataset with an 85.61% F1 score to further prove the superiority of the proposed model. CONCLUSION: The proposed model not only facilitates precise recognition of glomerular lesions but also serves as a potent tool for diagnosing kidney diseases effectively. Furthermore, the framework and training methodology of the GNN can be adeptly applied to address various pathology image classification challenges.

The Use of QR Codes to Engage Nursing Students With Next Generation NCLEX.

BACKGROUND: The introduction of the Next Generation NCLEX (NGN) necessitates the need to prepare students to demonstrate appropriate clinical judgment and reasoning. Innovative teaching strategies, such as the use of QR codes, may help to engage learners and promote the transition to the new NCLEX testing format. METHOD: A three-phase pediatric-based case study was used to introduce NGN style questions. The QR codes provided resources, answers, and rationales for the case study questions. RESULTS: The students reported an increase in perceived confidence with a case study-based NGN testing item. Student feedback was positive for this active and collaborative learning experience. CONCLUSION: Nurse educators are challenged with finding methods to engage learners and prepare students for practice. Using QR codes in the classroom is an innovative approach to expose students to NGN questions and may help increase students confidence as they prepare for the new NGN. [J Nurs Educ. 2024;63(3):186-187.].

Investigating the conception of collaborative learning (CL) and student engagement in the acquisition of practical skills (SEPSA) among prospective physical education and sports students.

The study aims to explore the association between collaborative learning and practical skills acquisition (SEPSA) among 310 students from second-year, third-year, and fourth-year (First stage of higher education) from the Institute of Arts, Culture, and Sports- Abai Kazakh National Pedagogical University. The data was collected using the time-lag approach at three intervals; 3rd week, 7th week, and 14th week. The mediation analysis suggests that collaborative learning (CL) has a positive mediating association with self-efficacy, and student engagement in practical skills acquisition (SEPSA). Additionally, collaborative learning (CL) has a positive mediating association with value-benefits, and practical skills acquisition (SEPSA). Furthermore, Collaborative learning (CL) has a positive significant association with practical skills acquisition (SEPSA). Our findings highlight the important potential of CL for increasing SEPSA. The finding of the study has implications for higher education teachers, students, administrators, and policymakers for developing more effective teaching and learning approaches using the concept of sharing and discussion with a specific focus on students' engagement.

Separated collaborative learning for semi-supervised prostate segmentation with multi-site heterogeneous unlabeled MRI data.

Segmenting prostate from magnetic resonance imaging (MRI) is a critical procedure in prostate cancer staging and treatment planning. Considering the nature of labeled data scarcity for medical images, semi-supervised learning (SSL) becomes an appealing solution since it can simultaneously exploit limited labeled data and a large amount of unlabeled data. However, SSL relies on the assumption that the unlabeled images are abundant, which may not be satisfied when the local institute has limited image collection capabilities. An intuitive solution is to seek support from other centers to enrich the unlabeled image pool. However, this further introduces data heterogeneity, which can impede SSL that works under identical data distribution with certain model assumptions. Aiming at this under-explored yet valuable scenario, in this work, we propose a separated collaborative learning (SCL) framework for semi-supervised prostate segmentation with multi-site unlabeled MRI data. Specifically, on top of the teacher-student framework, SCL exploits multi-site unlabeled data by: (i) Local learning, which advocates local distribution fitting, including the pseudo label learning that reinforces confirmation of low-entropy easy regions and the cyclic propagated real label learning that leverages class prototypes to regularize the distribution of intra-class features; (ii) External multi-site learning, which aims to robustly mine informative clues from external data, mainly including the local-support category mutual dependence learning, which takes the spirit that mutual information can effectively measure the amount of information shared by two variables even from different domains, and the stability learning under strong adversarial perturbations to enhance robustness to heterogeneity. Extensive experiments on prostate MRI data from six different clinical centers show that our method can effectively generalize SSL on multi-site unlabeled data and significantly outperform other semi-supervised segmentation methods. Besides, we validate the extensibility of our method on the multi-class cardiac MRI segmentation task with data from four different clinical centers.

A novel collaborative learning model for mixed dentition and fillings segmentation in panoramic radiographs.

INTRODUCTION: It is critical for dentists to identify and differentiate primary and permanent teeth, fillings, dental restorations and areas with pathological findings when reviewing dental radiographs to ensure that an accurate diagnosis is made and the optimal treatment can be planned. Unfortunately, dental radiographs are sometimes read incorrectly due to human error or low-quality images. While secondary or group review can help catch errors, many dentists work in practice alone and/or do not have time to review all of their patients' radiographs with another dentist. Artificial intelligence may facilitate the accurate interpretation of radiographs. To help support the review of panoramic radiographs, we developed a novel collaborative learning model that simultaneously identifies and differentiates primary and permanent teeth and detects fillings. METHODS: We used publicly accessible dental panoramic radiographic images and images obtained from the University of Missouri-Kansas City, School of Dentistry to develop and optimize two high-performance classifiers: (1) a system for tooth segmentation that can differentiate primary and permanent teeth and (2) a system to detect dental fillings. RESULTS: By utilizing these high-performance classifiers, we created models that can identify primary and permanent teeth (mean average precision [mAP] 95.32 % and performance [F-1] 92.50 %), as well as their associated dental fillings (mAP 91.53 % and F-1 91.00 %). We also designed a novel method for collaborative learning that utilizes these two classifiers to enhance recognition performance (mAP 94.09 % and F-1 93.41 %). CONCLUSIONS: Our model improves upon the existing machine learning models to simultaneously identify and differentiate primary and permanent teeth, and to identify any associated fillings. CLINICAL SIGNIFICANCE: Human error can lead to incorrect readings of panoramic radiographs. By developing artificial intelligence and machine learning methods to analyze panoramic radiographs, dentists can use this information to support their radiograph interpretations, help communicate the information to patients, and assist dental students learning to read radiographs.

Enhancing collaborative learning in health management education: an investigation of Padlet-mediated interventions and the influence of flexible thinking.

This study had three primary goals. First, it aimed to craft an intervention program centered around collaborative learning enabled by Padlet. Second, it aimed to gauge the perceptions of health management students regarding this intervention and how it affected their collaborative learning experiences. Additionally, the third objective of the study aimed to investigate how students' flexible thinking within the learning process might shape their perceptions of the advantages derived from this instructional activity within the domain of online collaborative learning. Data for the analysis were gathered from 100 Israeli undergraduate students by two measurements: Flexible thinking in learning and Student perceptions of collaborative learning via Padlet. The intervention program included several stages. First, the students discussed the pedagogic objective of using Padlet. In the second stage, the students were presented with ill-structured problems related to the course content. Each group had to choose one problem and analyze it from three perspectives discussed in the course-healthcare provider, patient, and organization. Next, the students presented and explained their solutions employing the shared knowledge base. The final work was presented in different formats using various technologies. The PLS-SEM analysis has corroborated our hypothesis that students' flexible thinking might positively contribute to their perception of Padlet utilization. According to the empirical model, in general, students who perceived themselves as more flexible were found more receptive to utilizing the proposed technological tool (Padlet) and hence tended to appreciate its function as a collaborative learning platform enabler. This study mainly underscores the important role flexible thinking plays in motivating managers and medical professionals to embrace innovative technologies or methods for teamwork, that could enable them to weigh arguments, seek alternative solutions to authentic problems, and adjust their approaches effectively and collaboratively as new challenges emerge.

SCREENER: Streamlined collaborative learning of NER and RE model for discovering gene-disease relations.

Finding relations between genes and diseases is essential in developing a clinical diagnosis, treatment, and drug design for diseases. One successful approach for mining the literature is the document-based relation extraction method. Despite recent advances in document-level extraction of entity-entity, there remains a difficulty in understanding the relations between distant words in a document. To overcome the above limitations, we propose an AI-based text-mining model that learns the document-level relations between genes and diseases using an attention mechanism. Furthermore, we show that including a direct edge (DE) and indirect edges between genetic targets and diseases when training improves the model's performance. Such relation edges can be visualized as graphs, enhancing the interpretability of the model. For the performance, we achieved an F1-score of 0.875, outperforming state-of-the-art document-level extraction models. In summary, the SCREENER identifies biological connections between target genes and diseases with superior performance by leveraging direct and indirect target-disease relations. Furthermore, we developed a web service platform named SCREENER (Streamlined CollaboRativE lEarning of NEr and Re), which extracts the gene-disease relations from the biomedical literature in real-time. We believe this interactive platform will be useful for users to uncover unknown gene-disease relations in the world of fast-paced literature publications, with sufficient interpretation supported by graph visualizations. The interactive website is available at: https://ican.standigm.com.

Improving autonomy support: a collaborative learning experience / Améliorer les pratiques d'accompagnement à l'autonomie en santé : effets d'un dispositif d'apprentissage coopératif

Introduction: The ACESO project, which was part of the Autonomy support in health national experimentation, brought together 21 partners from Ile-de-France. Among these partners, 14 had practices similar to autonomy support. Partners' presupposition was that experimenting a cooperative approach would encourage the empowerment of participants, improve their autonomy support and put into place the conditions necessary for the empowerment of people who would be supported. To help participants to meet this goal, the project leader took on a role as third party whose function was to facilitate the cooperative approach by proposing a framework and a method. Purpose of research: The study aimed to report the effects of this approach on the participants' practices as well as to identify the process for achieving this. Results: The participants' learning enabled them to align themselves with the good practice guidelines collectively constructed within the project. With the project leader's support, they initiated a transformative learning process that allowed them to develop their reflexivity and empowerment. These transformations had repercussions on their teams and structures, through a halo effect. The halo effect varied, in each partner structure, according to the participation and involvement of the project referent and the other members of the structure, in particular managers. Conclusions: This study has highlighted the value of a cooperative approach to facilitate the learning necessary for sustainable practices transformations and the improvement partners autonomy supports. This resulted in gains in autonomy for the autonomy support practitioners and the people they supported.

Introduction: Le projet ACESO, participant à l'expérimentation nationale des dispositifs d'accompagnement à l'autonomie en santé (AAS), a rassemblé 21 partenaires franciliens parmi lesquels 14 portaient des pratiques qui empruntent à l'accompagnement. Son présupposé était qu'en expérimentant une démarche coopérative favorisant l'empowerment des partenaires, ceux-ci amélioreraient leurs pratiques d'accompagnement et mettraient notamment en place les conditions nécessaires à l'empowerment des personnes accompagnées. Pour les y aider, le porteur de projet a tenu un rôle de tiers dont la fonction était de faciliter la démarche coopérative en proposant un cadre et une méthode. But de l'étude: L'étude visait à rendre compte des effets de cette démarche sur les pratiques des partenaires, ainsi qu'à identifier le processus pour y parvenir. Résultats: Les apprentissages réalisés ont permis aux partenaires de se donner des balises de bonnes pratiques construites collectivement au sein du projet (valeurs, principes et postures). Avec le soutien du tiers, ils ont initié un processus d'apprentissage transformationnel développant leur réflexivité et leur empowerment. Ces transformations ont eu des répercussions sur leurs équipes et structures, par effet de halo. Ce dernier a varié, dans chaque structure partenaire, en fonction de la participation et de l'implication du référent-projet et des membres de la structure, en particulier la direction. Conclusion: Cette étude met en évidence l'intérêt d'une démarche coopérative pour faciliter l'apprentissage nécessaire aux transformations durables des pratiques et l'amélioration des pratiques de partenaires d'un collectif apprenant. Dans le cas de l'AAS, ceci s'est traduit par des gains d'autonomie pour les accompagnants et les personnes accompagnées.

Effects of a Moodle-based E-learning environment on E-collaborative learning, perceived satisfaction, and study achievement among nursing students: A cross-sectional study.

BACKGROUND: Studies suggest that e-learning environments (ELEMs) in nursing education may be more effective than traditional face-to-face teaching, as they lead to learning outcomes that equal or exceed those of face-to-face teaching. OBJECTIVES: To determine whether using ELEM for educational applications can significantly improve e-collaborative learning, perceived satisfaction, and study achievement among nursing students in a paediatric nursing course. DESIGN: Nonrandomized pretest-posttest quasi-experimental research design. SETTINGS: A medical college in northern Taiwan. PARTICIPANTS: Eighty-four students (52 in the non-ELEM group and 32 in the ELEM group) completed both the pretest and posttest. METHODS: Third-year nursing students were recruited and nonrandomly assigned to an experimental group (ELEM) and a nonexperimental group (non-ELEM) of their choice. Students in the former group received traditional classroom teaching without the use of Moodle-based ELEMs, while those in the latter completed the course through Moodle-based ELEMs and classroom lectures. RESULTS: Regarding perceived satisfaction, e-collaborative learning, and study achievement, the overall test results indicated a significant difference in the posttest between the two groups (F (1,82) = 10.622, P = 0.002), (F (1,82) = 9.481, P = 0.003), (F (1,82) = 59.301, P < 0.001, respectively). The explanatory power η2 reached 11.5 %, 10.4 %, and 42.0 %, respectively. CONCLUSION: The students who used Moodle-based ELEMs combined with classroom teaching showed significantly higher levels of e-collaborative learning, perceived satisfaction, and study achievement in the paediatric nursing course. ELEMs for educational purposes can serve as effective complementary learning tools for paediatric nursing courses.

A Federated Attention-Based Multimodal Biometric Recognition Approach in IoT.

The rise of artificial intelligence applications has led to a surge in Internet of Things (IoT) research. Biometric recognition methods are extensively used in IoT access control due to their convenience. To address the limitations of unimodal biometric recognition systems, we propose an attention-based multimodal biometric recognition (AMBR) network that incorporates attention mechanisms to extract biometric features and fuse the modalities effectively. Additionally, to overcome issues of data privacy and regulation associated with collecting training data in IoT systems, we utilize Federated Learning (FL) to train our model This collaborative machine-learning approach enables data parties to train models while preserving data privacy. Our proposed approach achieves 0.68%, 0.47%, and 0.80% Equal Error Rate (EER) on the three VoxCeleb1 official trial lists, performs favorably against the current methods, and the experimental results in FL settings illustrate the potential of AMBR with an FL approach in the multimodal biometric recognition scenario.

Shared learning experiences: Pilot study of an online exchange project between pharmacy students in South Africa and the United States.

INTRODUCTION: Internationalisation enhances students' understanding of social, cultural, and ethical differences, preparing them to be global-minded, socially accountable healthcare practitioners. Traditionally, internationalisation of education involves international student travel. Online tools provide opportunities for international, peer-driven learning and collaboration without costly travel. This research described the experiences of pharmacy students from South Africa (SA) and the United States (US) that participated in a virtual peer exchange project during the COVID-19 pandemic. METHODS: The virtual peer exchange project allowed students in SA and the US to establish connections within the global pharmacy community and compare healthcare, pharmacy education, and pharmacy practice between the countries. Students engaged in facilitated dialogue through video recordings, video conferencing, and a group discussion board. Student introduction video comments and discussion board posts were thematically analysed. RESULTS: Twenty-one students participated in the pilot project that met some of the intentions and goals of internationalisation via a virtual platform. Two over-arching themes of Practice of Pharmacy and Pharmacy Education emerged from both the introduction video and discussion board comments. Students described lessons learned about similarities and differences in socioeconomic determinants of health as well as structure, functioning, and financing of the different healthcare systems. CONCLUSIONS: This project was a unique way to conduct exchange programmes via a virtual platform, and bypassed challenges of traditional exchange programmes. Through technology, more students in diverse geographic locations can be exposed to various perspectives and healthcare experiences with international students.

Kickstart in nursing home-Nursing students experiences of a model for active and collaborative learning in clinical placement.

AIMS AND OBJECTIVES: To explore students' experiences from a pilot project testing out a model for active, collaborative learning in first-year placement at a nursing home. BACKGROUND: There is a need for innovative learning activities and projects to improve clinical education in nursing homes. Active, collaborative approaches in placement learning may enhance students learning outcome. DESIGN: The study had a qualitative and explorative design, in which the experiences of students participating in the pilot were investigated through paired interviews at the end of their placement. METHODS: Twenty-two students participated in the study, and data from paired interviews were analysed using qualitative content analysis. COREQ reporting guidelines were used. RESULTS: Three themes emerged from the analysis: (1) The learning cell as facilitator for learning; (2) Discovering learning possibilities in nursing homes and (3) Applying tools and resources for learning. CONCLUSIONS: The model could reduce tension and anxiety while helping the students focus on learning options and use their environment more actively for learning. Working with a learning partner seems to increase student learning through common planning, feedback and reflection. The study emphasises the importance of facilitating active learning through the scaffolding structures and configuration of the students' learning space. RELEVANCE TO CLINICAL PRACTICE: This study indicates the potential for introducing active and collaborative pedagogical models in clinical placement. The model can promote nursing homes as a conducive learning arena for nursing students and help prepare students for a future work role in a rapidly changing health care field. PATIENT OR PUBLIC CONTRIBUTION: The result of the research is shared and discussed with stakeholders prior to finalising the article.

EDC-DTI: An end-to-end deep collaborative learning model based on multiple information for drug-target interactions prediction.

Innovations in drug-target interactions (DTIs) prediction accelerate the progression of drug development. The introduction of deep learning models has a dramatic impact on DTIs prediction, with a distinct influence on saving time and money in drug discovery. This study develops an end-to-end deep collaborative learning model for DTIs prediction, called EDC-DTI, to identify new targets for existing drugs based on multiple drug-target-related information including homogeneous information and heterogeneous information by the way of deep learning. Our end-to-end model is composed of a feature builder and a classifier. Feature builder consists of two collaborative feature construction algorithms that extract the molecular properties and the topology property of networks, and the classifier consists of a feature encoder and a feature decoder which are designed for feature integration and DTIs prediction, respectively. The feature encoder, mainly based on the improved graph attention network, incorporates heterogeneous information into drug features and target features separately. The feature decoder is composed of multiple neural networks for predictions. Compared with six popular baseline models, EDC-DTI achieves highest predictive performance in the case of low computational costs. Robustness tests demonstrate that EDC-DTI is able to maintain strong predictive performance on sparse datasets. As well, we use the model to predict the most likely targets to interact with Simvastatin (DB00641), Nifedipine (DB01115) and Afatinib (DB08916) as examples. Results show that most of the predictions can be confirmed by literature with clear evidence.

Effect of task-based group experience on collaborative learning: Exploring the transaction activities.

BACKGROUND: Collaborative learning is a widely used approach where students gather in small groups to solve problems and develop skills. However, grouping students is not always effective, and it may be necessary to provide task-specific collaborative experiences to optimize their interactions for subsequent learning tasks. AIMS: To test this hypothesis, we conducted an experiment with 90 Ecuadorian students in their mathematics class. SAMPLE: Participants were 90 Ecuadorian students (average age = 13.80 years, SD = .70; 48.89% female) from a private school in Sangolquí, who participated as part of their mathematics class. METHOD: The experiment consisted of four phases: preparation, learning, retention one-day testing, and delayed seven-day testing. In the preparation phase, 15 triads received guidance on working collaboratively with quadratic equations (i.e., experienced groups), while 45 other individual learners worked independently. In the learning phase, 15 experienced triads and 45 individual learners (who were later divided into 15 non-experienced triads) received a new learning task in the domain of economics, precisely the break-even point. RESULTS: The experienced group outperformed the non-experienced group in the retention one-day test, investing less mental effort and demonstrating greater efficiency. However, there was no significant difference in the delayed one-week test. We analysed the interactions of the groups and found that experienced groups exhibited more cognitive, fewer regulatory, an equal number of emotional interactions, and fewer task-unrelated interactions than the non-experienced groups. CONCLUSIONS: Providing task-specific collaborative experiences can reduce the cognitive load associated with transactional activities and increase learning in new tasks.

Shared learning from a physiologically based pharmacokinetic modeling strategy for human pharmacokinetics prediction through retrospective analysis of Genentech compounds.

The quantitative prediction of human pharmacokinetics (PK) including the PK profile and key PK parameters are critical for early drug development decisions, successful phase I clinical trials, and the establishment of a range of doses to enable phase II clinical dose selection. Here, we describe an approach employing physiologically based pharmacokinetic (PBPK) modeling (Simcyp) to predict human PK and to validate its performance through retrospective analysis of 18 Genentech compounds for which clinical data are available. In short, physicochemical parameters and in vitro data for preclinical species were integrated using PBPK modeling to predict the in vivo PK observed in mouse, rat, dog, and cynomolgus monkey. Through this process, the in vitro to in vivo extrapolation (IVIVE) was determined and then incorporated into PBPK modeling in order to predict human PK. Overall, the prediction obtained using this PBPK-IVIVE approach captured the observed human PK profiles of the compounds from the dataset well. The predicted Cmax was within 2-fold of the observed Cmax for 94% of the compounds while the predicted area under the curve (AUC) was within 2-fold of the observed AUC for 72% of the compounds. Additionally, important IVIVE trends were revealed through this investigation, including application of scaling factors determined from preclinical IVIVE to human PK prediction for each molecule. Based upon the analysis, this PBPK-based approach now serves as a practical strategy for human PK prediction at the candidate selection stage at Genentech.

Addressing multi-site functional MRI heterogeneity through dual-expert collaborative learning for brain disease identification.

Several studies employ multi-site rs-fMRI data for major depressive disorder (MDD) identification, with a specific site as the to-be-analyzed target domain and other site(s) as the source domain. But they usually suffer from significant inter-site heterogeneity caused by the use of different scanners and/or scanning protocols and fail to build generalizable models that can well adapt to multiple target domains. In this article, we propose a dual-expert fMRI harmonization (DFH) framework for automated MDD diagnosis. Our DFH is designed to simultaneously exploit data from a single labeled source domain/site and two unlabeled target domains for mitigating data distribution differences across domains. Specifically, the DFH consists of a domain-generic student model and two domain-specific teacher/expert models that are jointly trained to perform knowledge distillation through a deep collaborative learning module. A student model with strong generalizability is finally derived, which can be well adapted to unseen target domains and analysis of other brain diseases. To the best of our knowledge, this is among the first attempts to investigate multi-target fMRI harmonization for MDD diagnosis. Comprehensive experiments on 836 subjects with rs-fMRI data from 3 different sites show the superiority of our method. The discriminative brain functional connectivities identified by our method could be regarded as potential biomarkers for fMRI-related MDD diagnosis.

Bayesian Collaborative Learning for Whole-Slide Image Classification.

Whole-slide image (WSI) classification is fundamental to computational pathology, which is challenging in extra-high resolution, expensive manual annotation, data heterogeneity, etc. Multiple instance learning (MIL) provides a promising way towards WSI classification, which nevertheless suffers from the memory bottleneck issue inherently, due to the gigapixel high resolution. To avoid this issue, the overwhelming majority of existing approaches have to decouple the feature encoder and the MIL aggregator in MIL networks, which may largely degrade the performance. Towards this end, this paper presents a Bayesian Collaborative Learning (BCL) framework to address the memory bottleneck issue with WSI classification. Our basic idea is to introduce an auxiliary patch classifier to interact with the target MIL classifier to be learned, so that the feature encoder and the MIL aggregator in the MIL classifier can be learned collaboratively while preventing the memory bottleneck issue. Such a collaborative learning procedure is formulated under a unified Bayesian probabilistic framework and a principled Expectation-Maximization algorithm is developed to infer the optimal model parameters iteratively. As an implementation of the E-step, an effective quality-aware pseudo labeling strategy is also suggested. The proposed BCL is extensively evaluated on three publicly available WSI datasets, i.e., CAMELYON16, TCGA-NSCLC and TCGA-RCC, achieving an AUC of 95.6%, 96.0% and 97.5% respectively, which consistently outperforms all the methods compared. Comprehensive analysis and discussion will also be presented for in-depth understanding of the method. To promote future work, our source code is released at: https://github.com/Zero-We/BCL.

Heterogeneous Collaborative Learning for Personalized Healthcare Analytics via Messenger Distillation.

The Healthcare Internet-of-Things (IoT) framework aims to provide personalized medical services with edge devices. Due to the inevitable data sparsity on an individual device, cross-device collaboration is introduced to enhance the power of distributed artificial intelligence. Conventional collaborative learning protocols (e.g., sharing model parameters or gradients) strictly require the homogeneity of all participant models. However, real-life end devices have various hardware configurations (e.g., compute resources), leading to heterogeneous on-device models with different architectures. Moreover, clients (i.e., end devices) may participate in the collaborative learning process at different times. In this paper, we propose a Similarity-Quality-based Messenger Distillation (SQMD) framework for heterogeneous asynchronous on-device healthcare analytics. By introducing a preloaded reference dataset, SQMD enables all participant devices to distill knowledge from peers via messengers (i.e., the soft labels of the reference dataset generated by clients) without assuming the same model architecture. Furthermore, the messengers also carry important auxiliary information to calculate the similarity between clients and evaluate the quality of each client model, based on which the central server creates and maintains a dynamic collaboration graph (communication graph) to improve the personalization and reliability of SQMD under asynchronous conditions. Extensive experiments on three real-life datasets show that SQMD achieves superior performance.