International audience

International audience

International audience

Decision trees are commonly used for representing medical reasoning, both in structured representations for automatic reasoning and in visual presentations for clinicians. However, in the latter purpose, decision trees often suffer from two problems: (a) some reasoning processes are not well suited for trees, e.g. when considering several independent followup elements, (b) the size of the tree is limited by the screen. Here, we propose the dynamic and interactive visualization of decision trees. To solve problem ($a$), we used multi-path decision trees, allowing the selection of multiple choices at some nodes, which greatly reduces the size of the tree in some situations. We provide algorithms for one-click navigation in a multi-path tree, i.e. the user may click on any node of the tree at any time. For problem (b), we used the fisheye technique for reducing the visual space devoted to the inaccessible part of the tree. In addition, we represented the tree internally in a formal ontology, making it machine-interpretable and permitting the automatic execution of parts of the tree. We applied this approach in a clinical decision support system related to the management of Covid-19, and we showed the resulting user interface to 6 clinicians.

Polypharmacy, the consuming of more than five drugs, is a public health problem. It can lead to many interactions and adverse drug reactions and is very expensive. Therapeutic guidelines for managing polypharmacy in the elderly have been issued, but are highly complex, limiting their use. Decision-support systems have therefore been developed to automate the execution of these guidelines, or to provide information about drugs adapted to the context of polypharmacy. These systems differ widely in terms of their technical design, knowledge sources and evaluation methods. We present here a scoping review of electronic systems for supporting the management, by healthcare providers, of polypharmacy in elderly patients. Most existing reviews have focused mainly on evaluation results, whereas the present review also describes the technical design of these systems and the methodologies for developing and evaluating them. A systematic bibliographic search identified 19 systems differing considerably in terms of their technical design (rule-based systems, documentary approach, mixed); outputs (textual report, alerts and/or visual approaches); and evaluations (impact on clinical practices, impact on patient outcomes, efficiency and/or user satisfaction). The evaluations performed are minimal (among all the systems identified, only one system has been evaluated according to all the criteria mentioned above) and no machine learning systems and/or conflict management systems were retrieved. This review highlights the need to develop new methodologies, combining various approaches for decision support system in polypharmacy.

Prioritized Qualitative Choice Logic (PQCL) is obtained from the propositional logic by adding a new logical connective, called ordered disjunction (→×) to express user preferences. Intuitively, if x and y are two outcomes then x⃗×y means: “if possible x, but if x is impossible then at least y”. In this paper, we propose a method for learning PQCL user's preferences. We evaluate the efficiency of the proposed method on real medical datasets.

Drugs are evaluated and compared during clinical trials, and the observed adverse drug events are recorded. Many trial results are available online in trial registries, such as ClinicalTrials.gov. However, this data remains little used currently, due to the volume and the tabular presentation.In this paper, we focus on the visualization of adverse drug event rates, aggregated in 13 anatomical categories and 2 levels of seriousness. We designed four interactive visual analytics: horizontal stacked bar chart with labels, vertical stacked bar chart, area-proportional flower glyphs and star glyphs. We compare the four approaches, analytically and through a preliminary user study aimed at determining the user preferred visual analytics, and we identify their advantages and disadvantages. We conclude that the best options are the horizontal stacked bar chart, which allows labels to help identify the categories of adverse drug events, and the area-proportional flower glyph, which permits a better visualization of small values when very high values are present.

Set visualization is a well-known task in information visualization. In biology, it is used for comparing visually sets of genes or proteins, typically using Venn diagrams. However, limitations of the Venn diagram are well-known: they are limited to 6 sets and difficult to read above 4. Many other set visualization techniques have been proposed, but they have never been widely used in biology. In this paper, we introduce RainBio, a technique for visualizing sets in biology and aimed at providing a global overview showing the size of the main intersections, in a proportional way, and the similarities between sets. We adapt rainbow boxes, a technique for visualizing small datasets, to the visualization of larger sets, using element aggregation and intersection clustering. We present the application of RainBio to three datasets, with 5, 6 and 12 sets. We also describe a small user study comparing RainBio with Venn diagrams, involving 30 students in biology. Results showed that RainBio led to significantly fewer errors on 6-set dataset, and that the majority of students preferred RainBio. RainBio is proposed as a web-based tool for up to 15 sets.

Visualization tools allow visual and interactive data exploration for facilitating the interpretation of complex data sets. The Center for Research and Ambulatory Care of the Sickle Cell Disease (CERPAD), leading the neonatal screening of sickle cell program in Saint-Louis of Senegal, aims at setting up a long term database on which data analysis and reporting tools are built for epidemiological and socioanthropological studies. In this paper, we propose a tool for the proportional visualization of genotypes and phenotypes, using rainbow boxes, a recently introduced set visualization technique. We propose an improvement for rainbow boxes, featuring area-proportional boxes instead of height-proportional boxes. This new approach is integrated in the SIMENS module of the CERPAD for the visualization of sickle cell data proportion according to genotypes and phenotypes, and by ethnic groups. A qualitative evaluation is provided by two domain experts.

Artificial neural networks are machine-learning algorithms inspired by biological neural networks. Their main inconvenient is their “black-box” nature: while they are very efficient for making predictions, it is difficult to explain these predictions. In this paper, we propose a visual translation of the reasoning performed by simple neural networks, i.e. without hidden layers. This visualization relies on rainbow boxes, a recently-introduced technique for set visualization, and on three improvements we propose for rainbow boxes, including interactivity. We also present a small application of the proposed approach to decision support in antibiotherapy, for helping a physician to choose an antibiotic in urinary infections.

Case-Based Reasoning (CBR) is a form of analogical reasoning in which the solution for a (new) query case is determined using a database of previous known cases with their solutions. Cases similar to the query are retrieved from the database, and then their solutions are adapted to the query. In medicine, a case usually corresponds to a patient and the problem consists of classifying the patient in a class of diagnostic or therapy. Compared to "black box" algorithms such as deep learning, the responses of CBR systems can be justified easily using the similar cases as examples. However, this possibility is often under-exploited and the explanations provided by most CBR systems are limited to the display of the similar cases. In this paper, we propose a CBR method that can be both executed automatically as an algorithm and presented visually in a user interface for providing visual explanations or for visual reasoning. After retrieving similar cases, a visual interface displays quantitative and qualitative similarities between the query and the similar cases, so as one can easily classify the query through visual reasoning, in a fully explainable manner. It combines a quantitative approach (visualized by a scatter plot based on Multidimensional Scaling in polar coordinates, preserving distances involving the query) and a qualitative approach (set visualization using rainbow boxes). We applied this method to breast cancer management. We showed on three public datasets that our qualitative method has a classification accuracy comparable to k-Nearest Neighbors algorithms, but is better explainable. We also tested the proposed interface during a small user study. Finally, we apply the proposed approach to a real dataset in breast cancer. Medical experts found the visual approach interesting as it explains why cases are similar through the visualization of shared patient characteristics.

INTRODUCTION: Clinical decision support systems (CDSS) implementing clinical practice guidelines (CPGs) have 2 main limitations: they target only patients for whom CPGs provide explicit recommendations, and their rationale may be difficult to understand. These 2 limitations result in poor CDSS adoption. We designed AntibioHelp® as a CDSS for antibiotic treatment. It displays the recommended and nonrecommended antibiotics, together with their properties, weighted by degree of importance as outlined in the CPGs. The aim of this study was to determine whether AntibioHelp® could increase the confidence of general practitioners (GPs) in CPG recommendations and help them to extrapolate guidelines to patients for whom CPGs provide no explicit recommendations. MATERIALS AND METHODS: We carried out a 2-stage crossover study in which GPs responded to clinical cases using CPG recommendations either alone or with explanations displayed through AntibioHelp®. We compared error rates, confidence levels, and response times. RESULTS: We included 64 GPs. When no explicit recommendation existed for a particular situation, AntibioHelp® significantly decreased the error rate (-41%, P value = 6x10-13), and significantly increased GP confidence (+8%, P value = .02). This CDSS was considered to be usable by GPs (SUS score = 64), despite a longer interaction time (+9-22 seconds). By contrast, AntibioHelp® had no significant effect if there was an explicit recommendation. DISCUSSION/CONCLUSION: The visualization of weighted antibiotic properties helps GPs to extrapolate recommendations to patients for whom CPGs provide no explicit recommendations. It also increases GP confidence in their prescriptions for these patients. Further evaluations are required to determine the impact of AntibioHelp® on antibiotic prescriptions in real clinical practice.

Breast cancer is the most common type of cancer in women worldwide, with incidence rate being second highest to other types of cancer. In the current clinical setting, multidisciplinary breast units are introduced to improve the quality of the therapeutic decision based on the best evidence-based practices. DESIREE project aims to provide a web-based software ecosystem for personalized, collaborative and multidisciplinary management of primary breast cancer by specialized breast units, from diagnosis to therapy and follow-ups. In order to provide a multi-model decision support to clinicians in present clinical settings, the project develops and integrates three modalities of decision support, namely guideline-based decision support system (DSS), experience-based DSS and case-based DSS. Visual analytics GUI are developed to properly adapt the results of the DSSs and graphically represent them to the clinician in a user-friendly manner. DESIREE information management system (DESIMS), serves as the interface between the user and DSSs for entering patient data and viewing the results in the visual analytics GUI. In this paper, we present the overall architecture, workflow and integration of the three DSSs in the DESIREE platform.

A symmetric square matrix is a type of dataset frequently found in various domains. An example is character co-occurrence matrices in digital humanities. However, the visualization of these matrices is difficult. In this paper, we propose a method for visualizing symmetric square matrices, by transforming a matrix into overlapping sets, and then by visualizing these overlapping sets using rainbow boxes, a recent set visualization technique. We apply this method to the visualization of character matrices in novels, using two datasets: a small one (21 characters) and a large one (80 characters). We show that this visualization allows the finding of several insights. Finally, we discuss the advantages and drawbacks of this method, and we compare it to other approaches in the literature.

Euler diagrams are commonly used for visualizing small datasets, especially in biology. A well-known example is the diagram of amino acid properties. However, Euler diagrams are not always easy to read for students, and they are complex to produce when the number of sets is above 6. Other approaches exist for set visualization, such as rainbow boxes, but they have not been evaluated properly yet. In this paper, we propose a new diagram for amino acids, using rainbow boxes. We present a crossover user study that compares the new diagram with the usual Euler diagram (often miscalled “Venn diagram”). 78 students in biology were recruited; they responded to questions using the diagrams. We show that rainbow boxes lead to half the error rate (p value = 0.013) and also performed significantly better in terms of response time and user preference. We also explore the possibility of enriching the amino acid diagram with additional information, beyond the 8 properties commonly presented. We discuss the limitations of each technique.

In this paper, we propose a technique for translating visually the reasoning of a perceptron. The artificial neuron, or perceptron, is a simplified model of a biological neuron. It can achieve simple reasoning and solve linearly separable problems. Despite its limited reasoning power, it is enough to deal with several real-life problems. The proposed technique is based on rainbow boxes, a technique for overlapping set visualization, which has been applied to the input vectors of the perceptron. We extended this technique, leading to weighted rainbow boxes. It can visualize several input vectors and output values for a single perceptron. We applied this approach to decision support in antibiotherapy, for the determination of the most appropriate antibiotic in urinary infections, by taking into account the properties of each drug (e.g. efficacy, risk of adverse effects, etc). Finally, a user study with 11 physicians showed that most of them found the visualization interesting and easy to read.

Overlapping set visualization is a well-known problem in information visualization. This problem considers elements and sets containing all or part of the elements, a given element possibly belonging to more than one set. A typical example is the properties of the 20 amino-acids. A more complex application is the visual comparison of the contraindications or the adverse effects of several similar drugs. The knowledge involved is voluminous, each drug has many contraindications and adverse effects, some of them are shared with other drugs. In this paper, we present rainbow boxes, a novel technique for visualizing overlapping sets, and its application to the properties of amino-acids and to the comparison of drug properties. We also describe a user study comparing rainbow boxes to tables and showing that the former allowed physicians to find information significantly faster. We finally discuss the limits and the perspectives of rainbow boxes.