Explainable deep learning in plant phenotyping.

The increasing human population and variable weather conditions, due to climate change, pose a threat to the world's food security. To improve global food security, we need to provide breeders with tools to develop crop cultivars that are more resilient to extreme weather conditions and provide growers with tools to more effectively manage biotic and abiotic stresses in their crops. Plant phenotyping, the measurement of a plant's structural and functional characteristics, has the potential to inform, improve and accelerate both breeders' selections and growers' management decisions. To improve the speed, reliability and scale of plant phenotyping procedures, many researchers have adopted deep learning methods to estimate phenotypic information from images of plants and crops. Despite the successful results of these image-based phenotyping studies, the representations learned by deep learning models remain difficult to interpret, understand, and explain. For this reason, deep learning models are still considered to be black boxes. Explainable AI (XAI) is a promising approach for opening the deep learning model's black box and providing plant scientists with image-based phenotypic information that is interpretable and trustworthy. Although various fields of study have adopted XAI to advance their understanding of deep learning models, it has yet to be well-studied in the context of plant phenotyping research. In this review article, we reviewed existing XAI studies in plant shoot phenotyping, as well as related domains, to help plant researchers understand the benefits of XAI and make it easier for them to integrate XAI into their future studies. An elucidation of the representations within a deep learning model can help researchers explain the model's decisions, relate the features detected by the model to the underlying plant physiology, and enhance the trustworthiness of image-based phenotypic information used in food production systems.

Transcriptional re-programming of insulin B-chain epitope-specific T-follicular helper cells into anti-diabetogenic T-regulatory type-1 cells.

Systemic delivery of nanoparticles (NPs) coated with mono-specific autoimmune disease-relevant peptide-major histocompatibility complex class II (pMHCII) molecules can resolve organ inflammation in various disease models in a disease-specific manner without impairing normal immunity. These compounds invariably trigger the formation and systemic expansion of cognate pMHCII-specific T-regulatory type 1 (TR1) cells. By focusing on type 1 diabetes (T1D)-relevant pMHCII-NP types that display an epitope from the insulin B-chain bound to the same MHCII molecule (IAg7) on three different registers, we show that pMHCII-NP-induced TR1 cells invariably co-exist with cognate T-Follicular Helper (TFH)-like cells of quasi-identical clonotypic composition and are oligoclonal, yet transcriptionally homogeneous. Furthermore, these three different TR1 specificities have similar diabetes reversal properties in vivo despite being uniquely reactive against the peptide MHCII-binding register displayed on the NPs. Thus, pMHCII-NP treatment using nanomedicines displaying different epitope specificities results in the simultaneous differentiation of multiple antigen-specific TFH-like cell clones into TR1-like cells that inherit the fine antigenic specificity of their precursors while acquiring a defined transcriptional immunoregulatory program.

A T follicular helper cell origin for T regulatory type 1 cells.

Chronic antigenic stimulation can trigger the differentiation of antigen-experienced CD4+ T cells into T regulatory type 1 (TR1) cells, a subset of interleukin-10-producing Treg cells that do not express FOXP3. The identities of the progenitor(s) and transcriptional regulators of this T-cell subset remain unclear. Here, we show that the peptide-major histocompatibility complex class II (pMHCII) monospecific immunoregulatory T-cell pools that arise in vivo in different genetic backgrounds in response to pMHCII-coated nanoparticles (pMHCII-NPs) are invariably comprised of oligoclonal subpools of T follicular helper (TFH) and TR1 cells with a nearly identical clonotypic composition but different functional properties and transcription factor expression profiles. Pseudotime analyses of scRNAseq data and multidimensional mass cytometry revealed progressive downregulation and upregulation of TFH and TR1 markers, respectively. Furthermore, pMHCII-NPs trigger cognate TR1 cell formation in TFH cell-transfused immunodeficient hosts, and T-cell-specific deletion of Bcl6 or Irf4 blunts both the TFH expansion and TR1 formation induced by pMHCII-NPs. In contrast, deletion of Prdm1 selectively abrogates the TFH-to-TR1 conversion. Bcl6 and Prdm1 are also necessary for anti-CD3 mAb-induced TR1 formation. Thus, TFH cells can differentiate into TR1 cells in vivo, and BLIMP1 is a gatekeeper of this cellular reprogramming event.

Developing a dashboard for program evaluation in competency-based training programs: a design-based research project.

Background: Canadian specialist residency training programs are implementing a form of competency-based medical education (CBME) that requires the assessment of entrustable professional activities (EPAs). Dashboards could be used to track the completion of EPAs to support program evaluation. Methods: Using a design-based research process, we identified program evaluation needs related to CBME assessments and designed a dashboard containing elements (data, analytics, and visualizations) meeting these needs. We interviewed leaders from the emergency medicine program and postgraduate medical education office at the University of Saskatchewan. Two investigators thematically analyzed interview transcripts to identify program evaluation needs that were audited by two additional investigators. Identified needs were described using quotes, analytics, and visualizations. Results: Between July 1, 2019 and April 6, 2021 we conducted 17 interviews with six participants (two program leaders and four institutional leaders). Four needs emerged as themes: tracking changes in overall assessment metrics, comparing metrics to the assessment plan, evaluating rotation performance, and engagement with the assessment metrics. We addressed these needs by presenting analytics and visualizations within a dashboard. Conclusions: We identified program evaluation needs related to EPA assessments and designed dashboard elements to meet them. This work will inform the development of other CBME assessment dashboards designed to support program evaluation.

Contexte: Les programmes de résidence dans une spécialité au Canada offrent une formation médicale fondée sur les compétences (FMFC) qui exige l'évaluation des activités professionnelles confiables (APC). Des tableaux de bord pourraient être utilisés pour suivre la complétion des APC afin de faciliter l'évaluation des programmes. Méthodes: Par un processus de recherche basé sur la conception, nous avons ciblé les besoins des programmes liés aux évaluations de la FMFC et conçu un tableau de bord qui comprend les éléments (données, analyses et visualisations) nécessaires pour répondre à ces besoins. Nous avons eu recours à des entretiens auprès des responsables du programme de médecine d'urgence et du bureau d'éducation médicale postdoctorale de l'Université de Saskatchewan. Deux enquêteurs ont effectué une analyse thématique des transcriptions des entretiens afin de recenser les besoins d'évaluation du programme, vérifiés par la suite par deux autres enquêteurs. Les besoins recensés ont été décrits à l'aide de citations, d'analyses et de visualisations. Résultats: Entre le 1er juillet 2019 et le 6 avril 2021, nous avons mené 17 entretiens avec six participants (deux responsables de programmes et quatre responsables de l'établissement). Quatre besoins sont ressortis en tant que thèmes : le suivi des changements dans les mesures d'évaluation globales, la comparaison des mesures avec le plan d'évaluation, l'évaluation de l'efficacité du stage et l'engagement face à des mesures d'évaluation. Nous avons répondu à ces besoins en présentant des analyses et des visualisations dans un tableau de bord. Conclusions: Nous avons identifié les besoins d'évaluation du programme liés aux évaluations des APC et conçu des éléments de tableau de bord pour y répondre. Ce travail guidera la conception d'autres tableaux de bord d'évaluation de la FMFC en vue de faciliter l'évaluation des programmes.

Leveraging Guided Backpropagation to Select Convolutional Neural Networks for Plant Classification.

The development of state-of-the-art convolutional neural networks (CNN) has allowed researchers to perform plant classification tasks previously thought impossible and rely on human judgment. Researchers often develop complex CNN models to achieve better performances, introducing over-parameterization and forcing the model to overfit on a training dataset. The most popular process for evaluating overfitting in a deep learning model is using accuracy and loss curves. Train and loss curves may help understand the performance of a model but do not provide guidance on how the model could be modified to attain better performance. In this article, we analyzed the relation between the features learned by a model and its capacity and showed that a model with higher representational capacity might learn many subtle features that may negatively affect its performance. Next, we showed that the shallow layers of a deep learning model learn more diverse features than the ones learned by the deeper layers. Finally, we propose SSIM cut curve, a new way to select the depth of a CNN model by using the pairwise similarity matrix between the visualization of the features learned at different depths by using Guided Backpropagation. We showed that our proposed method could potentially pave a new way to select a better CNN model.

Re-programming mouse liver-resident invariant natural killer T cells for suppressing hepatic and diabetogenic autoimmunity.

Invariant NKT (iNKT) cells comprise a heterogeneous group of non-circulating, tissue-resident T lymphocytes that recognize glycolipids, including alpha-galactosylceramide (αGalCer), in the context of CD1d, but whether peripheral iNKT cell subsets are terminally differentiated remains unclear. Here we show that mouse and human liver-resident αGalCer/CD1d-binding iNKTs largely correspond to a novel Zbtb16+Tbx21+Gata3+MaflowRorc- subset that exhibits profound transcriptional, phenotypic and functional plasticity. Repetitive in vivo encounters of these liver iNKT (LiNKT) cells with intravenously delivered αGalCer/CD1d-coated nanoparticles (NP) trigger their differentiation into immunoregulatory, IL-10+IL-21-producing Zbtb16highMafhighTbx21+Gata3+Rorc- cells, termed LiNKTR1, expressing a T regulatory type 1 (TR1)-like transcriptional signature. This response is LiNKT-specific, since neither lung nor splenic tissue-resident iNKT cells from αGalCer/CD1d-NP-treated mice produce IL-10 or IL-21. Additionally, these LiNKTR1 cells suppress autoantigen presentation, and recognize CD1d expressed on conventional B cells to induce IL-10+IL-35-producing regulatory B (Breg) cells, leading to the suppression of liver and pancreas autoimmunity. Our results thus suggest that LiNKT cells are plastic for further functional diversification, with such plasticity potentially targetable for suppressing tissue-specific inflammatory phenomena.

Decision-making process for esthetic treatment of gummy smile: A surgical perspective.

Gummy smile is a quite frequently found esthetic alteration characterized by excessive display of gingiva during smiling. Several causes have been implicated in the literature, but a dearth of clinical decision-making process has been found in the surgical treatment of excessive gingival display. An external bevel gingivectomy with osseous correction was performed in anterior maxillary region in accordance with the proposed decision-making process. The clinical observation at 1 month postoperatively revealed restoration of natural smile with 1-2 mm of facial gingival display. The outcome seems to suggest that this proposed decision-making process can provide valid treatment options for gummy smile cases.

Developing a dashboard for faculty development in competency-based training programs: a design-based research project.

BACKGROUND: Canadian specialist residency training programs are implementing a form of competency-based medical education (CBME) that requires frequent assessments of entrustable professional activities (EPAs). Faculty struggle to provide helpful feedback and assign appropriate entrustment scores. CBME faculty development initiatives rarely incorporate teaching metrics. Dashboards could be used to visualize faculty assessment data to support faculty development. METHODS: Using a design-based research process, we identified faculty development needs related to CBME assessments and designed a dashboard containing elements (data, analytics, and visualizations) meeting these needs. Data was collected within the emergency medicine residency program at the University of Saskatchewan through interviews with program leaders, faculty development experts, and faculty participating in development sessions. Two investigators thematically analyzed interview transcripts to identify faculty needs that were audited by a third investigator. The needs were described using representative quotes and the dashboard elements designed to address them. RESULTS: Between July 1, 2019 and December 11, 2020 we conducted 15 interviews with nine participants (two program leaders, three faculty development experts, and four faculty members). Three needs emerged as themes from the analysis: analysis of assessments, contextualization of assessments, and accessible reporting. We addressed these needs by designing an accessible dashboard to present contextualized quantitative and narrative assessment data for each faculty member. CONCLUSIONS: We identified faculty development needs related to EPA assessments and designed dashboard elements to meet them. The resulting dashboard was used for faculty development sessions. This work will inform the development of CBME assessment dashboards for faculty.

CONTEXTE: Les programmes de résidence de spécialité au Canada mettent en œuvre une forme d'éducation axée sur les compétences (EASC) qui exige des évaluations formatives fréquentes des activités professionnelles confiables (APC). Les enseignants ont du mal à fournir une rétroaction utile et à attribuer des notes appropriées au niveau de confiance. Les initiatives de formation professorale des enseignants qui interviennent dans la EASC intègrent rarement leurs données psychométriques. Des tableaux de bord pourraient être utilisés pour visualiser les données d'évaluation du corps professoral afin de soutenir leur perfectionnement. MÉTHODES: En utilisant un processus de recherche orientée par la conception, nous avons déterminé les besoins de formation professorale liés aux évaluations dans la EASC et nous avons conçu un tableau de bord contenant des éléments (données, analyses et éléments visuels) pour répondre à ces besoins. Les données ont été recueillies dans le cadre du programme de résidence en médecine d'urgence de l'Université de Saskatchewan par le biais d'entretiens avec les responsables du programme, des experts en formation professorale et les enseignants participant aux séances de formation. Deux chercheurs ont procédé à une analyse thématique des transcriptions d'entrevues afin d'identifier les besoins des enseignants, et un troisième chercheur les a vérifiées. Les besoins ont été décrits à l'aide de citations représentatives et des éléments du tableau de bord conçus pour y répondre. RÉSULTATS: Entre le 1er juillet 2019 et le 11 décembre 2020, nous avons mené 15 entretiens avec neuf participants (deux responsables de programme, trois experts en formation professorale et quatre membres du corps professoral). Trois besoins sont ressortis comme thèmes de l'analyse : l'analyse des évaluations formatives, la contextualisation des évaluations formatives et l'accessibilité des rapports. Pour répondre à ces besoins, nous avons conçu un tableau de bord accessible présentant des données d'évaluation quantitatives et narratives contextualisées pour chaque membre du corps professoral. CONCLUSIONS: Nous avons identifié les besoins de formation professorale liés aux évaluations des APC et conçu les éléments d'un tableau de bord permettant d'y répondre. Le tableau de bord a été utilisé dans des séances de formation professorale. Ce travail orientera la réalisation de tableaux de bord afin de faciliter l'évaluation pour les enseignants dans le cadre de la EASC.

Extremely short bioavailability and fast pharmacodynamic effects of pMHC-based nanomedicines.

Nanoparticles (NPs) coated with autoimmune disease-relevant peptide-major histocompatibility complexes (pMHCs) can blunt autoimmune diseases by re-programming cognate effector T-lymphocytes into disease-suppressing regulatory T-cells, followed by massive expansion. Here, a method to quantify the absolute amounts of the active drug product is developed, to understand the relationship between bioavailability and pharmacodynamics. Incubation with plasma results in the formation of a protein corona that stabilizes the directional pMHC coat, shielding it from proteolysis or anti-drug antibody recognition, without any appreciable loss in biological potency. A quantitative method that harnesses these features indicates that the half-life of these compounds in the circulation and organs is an order of magnitude shorter (minutes vs. hours) than that measured using commonly-used semi-quantitative methods. Extensive transmission electron microscopy-based organ scanning and flow cytometry-based enumeration of pMHCII-NP capturing cells confirmed that these compounds are rapidly captured (within 1 min) by liver sinusoidal endothelial cells, Kupffer cells, splenic phagocytes and cognate T-cells, leading to a fast decline in the circulation. Therefore, the powerful pharmacodynamic effects of these compounds are dissociated from long bioavailability, implying a hit-and-run event. Collectively, these data provide a detailed view of the life-cycle of a nanoimmunomedicine, and suggest that the real half-lives of intact nanomedicines may be much shorter than those estimated using indirect approaches.

Liver-specific T regulatory type-1 cells program local neutrophils to suppress hepatic autoimmunity via CRAMP.

Neutrophils with immunoregulatory properties, also referred to as type-2 neutrophils (N2), myeloid-derived suppressor cells (MDSCs), or tumor-associated neutrophils (TANs), comprise a heterogeneous subset of cells that arise from unknown precursors in response to poorly understood cues. Here, we find that, in several models of liver autoimmunity, pharmacologically induced, autoantigen-specific T regulatory type-1 (TR1) cells and TR1-cell-induced B regulatory (Breg) cells use five immunoregulatory cytokines to coordinately recruit neutrophils into the liver and program their transcriptome to generate regulatory neutrophils. The liver-associated neutrophils from the treated mice, unlike their circulating counterparts or the liver neutrophils of sick mice lacking antigen-specific TR1 cells, are proliferative, can transfer disease protection to immunocompromised hosts engrafted with pathogenic effectors, and blunt antigen-presentation and local autoimmune responses via cathelin-related anti-microbial peptide (CRAMP), a cathelicidin, in a CRAMP-receptor-dependent manner. These results, thus, identify antigen-specific regulatory T cells as drivers of tissue-restricted regulatory neutrophil formation and CRAMP as an effector of regulatory neutrophil-mediated immunoregulation.

Developing a dashboard to meet the needs of residents in a competency-based training program: A design-based research project.

BACKGROUND: Canadian specialty programs are implementing Competence By Design, a competency-based medical education (CBME) program which requires frequent assessments of entrustable professional activities. To be used for learning, the large amount of assessment data needs to be interpreted by residents, but little work has been done to determine how visualizing and interacting with this data can be supported. Within the University of Saskatchewan emergency medicine residency program, we sought to determine how our residents' CBME assessment data should be presented to support their learning and to develop a dashboard that meets our residents' needs. METHODS: We utilized a design-based research process to identify and address resident needs surrounding the presentation of their assessment data. Data was collected within the emergency medicine residency program at the University of Saskatchewan via four resident focus groups held over 10 months. Focus group discussions were analyzed using a grounded theory approach to identify resident needs. This guided the development of a dashboard which contained elements (data, analytics, and visualizations) that support their interpretation of the data. The identified needs are described using quotes from the focus groups as well as visualizations of the dashboard elements. RESULTS: Resident needs were classified under three themes: (1) Provide guidance through the assessment program, (2) Present workplace-based assessment data, and (3) Present other assessment data. Seventeen dashboard elements were designed to address these needs. CONCLUSIONS: Our design-based research process identified resident needs and developed dashboard elements to meet them. This work will inform the creation and evolution of CBME assessment dashboards designed to support resident learning.

CONTEXTE: Les programmes canadiens de spécialité sont à implanter la compétence par conception (CPC), un programme d'éducation médicale par compétences qui nécessite des évaluations fréquentes des activités professionnelles confiables. Pour servir aux fins d'apprentissage, la grande quantité de données d'évaluation doit être interprétée par les résidents, mais peu de travaux ont été réalisés pour déterminer comment la visualisation et l'interaction avec ces données peuvent être soutenues. Dans le cadre du programme de résidence en médecine d'urgence de l'Université de Saskatchewan, nous avons cherché à déterminer comment les données d'évaluation de la CPC de nos résidents devraient être présentées pour soutenir leur apprentissage et pour développer un tableau de bord qui réponde aux besoins de nos résidents. MÉTHODOLOGIE: Nous avons utilisé un processus de recherche orientée par la conception pour cerner les besoins des résidents en lien avec la présentation de leurs données d'évaluation. Les données ont été recueillies au cours du programme de résidence en médecine d'urgence de l'Université de Saskatchewan grâce à quatre groupes de discussion de résidents qui se sont tenus sur une période de 10 mois. Les groupes de discussion ont été analysés en utilisant l'approche de la théorisation ancrée (Grounded Theory) pour cerner les besoins des résidents, pour guider le développement d'un tableau de bord contenant des éléments (données, analyses et visualisations) qui soutiennent leur interprétation de leurs propres données. Les besoins identifiés sont décrits à l'aide de citations des groupes de discussion ainsi que de visualisations des éléments du tableau de bord. RÉSULTATS: Les besoins des résidents ont été classés sous trois thèmes : 1. être guidés quant au programme d'évaluation, 2. présenter des données d'évaluation en milieu de travail, et 3. présenter d'autres données d'évaluation. Dix-sept éléments du tableau de bord ont été conçus pour répondre à ces besoins. CONCLUSIONS: Notre méthode de recherche orientée par conception a permis de cerner les besoins des résidents et d'élaborer les éléments d'un tableau de bord pour y répondre. Ce travail servira de base à la création et à l'évolution des tableaux de bord d'évaluation en CPC conçus pour soutenir l'apprentissage des résidents.

Developing a dashboard to meet Competence Committee needs: a design-based research project.

BACKGROUND: Competency-based programs are being adopted in medical education around the world. Competence Committees must visualize learner assessment data effectively to support their decision-making. Dashboards play an integral role in decision support systems in other fields. Design-based research allows the simultaneous development and study of educational environments. METHODS: We utilized a design-based research process within the emergency medicine residency program at the University of Saskatchewan to identify the data, analytics, and visualizations needed by its Competence Committee, and developed a dashboard incorporating these elements. Narrative data were collected from two focus groups, five interviews, and the observation of two Competence Committee meetings. Data were qualitatively analyzed to develop a thematic framework outlining the needs of the Competence Committee and to inform the development of the dashboard. RESULTS: The qualitative analysis identified four Competence Committee needs (Explore Workplace-Based Assessment Data, Explore Other Assessment Data, Understand the Data in Context, and Ensure the Security of the Data). These needs were described with narratives and represented through visualizations of the dashboard elements. CONCLUSIONS: This work addresses the practical challenges of supporting data-driven decision making by Competence Committees and will inform the development of dashboards for programs, institutions, and learner management systems.

CONTEXTE: Les programmes fondés sur la compétence sont adoptés dans la formation médicale à travers le monde. Les comités des compétences doivent visualiser efficacement les données d'évaluation des apprenants pour soutenir leurs prises de décision. Les tableaux de bord jouent un rôle essentiel dans les systèmes d'aide à la décision dans d'autres disciplines. La recherche orientée par la conception permet le développement et l'étude simultanés des environnements éducatifs. MÉTHODES: Nous avons utilisé un processus de recherche orienté par la conception au sein du programme de résidence en médecine d'urgence à l'Université de la Saskatchewan pour déterminer les données, les analyses et les visuels dont a besoin son comité des compétences, et avons développé un tableau de bord intégrant ces éléments. Les données narratives ont été recueillies auprès de deux groupes de discussion, lors de cinq entrevues et par l'observation de deux réunions du comité des compétences. Les données ont été analysées de manière qualitative pour élaborer un cadre thématique soulignant les besoins du comité des compétences et orienter le développement du tableau de bord. RÉSULTATS: L'analyse qualitative a dégagé quatre besoins du comité des compétences (explorer les données d'évaluation en milieu de travail, explorer d'autres données d'évaluation, comprendre les données dans leur contexte et s'assurer la sécurité des données). Ces besoins ont étédécrits avec des récits et représentés par des visuelsdes éléments du tableau de bord. CONCLUSIONS: Le présent travail aborde les difficultés pratiques de soutenir une prise de décision fondée sur des données par les comités des compétences et oriente le développement des tableaux de bord pour les programmes, les établissements et les systèmes de gestion des apprenants.