Portfolio
National and international projects developed by Vicomtech making use of the design and application capabilities of Plotly and Plotly-Dash. In this portfolio, we show different use cases in the field of health technologies such as epidemiology, cancer, medical imaging and wellness assessment involving Plotly for visual analysis and clinical decision tools.
Our latest projects
Tags:
Our projects
DESIREE
Description
Desiree is a web-based software ecosystem for the personalized designed in an European Project, collaborative and multidisciplinary management of primary breast cancer
Parallel Coordinates View
This view allows to analyze depending on the scenario relationships between variables and the toxicity associate for example the age, the diameter of the tumor and the number of lesions plotted in different classes if have toxicity or not.
Visual Analytics
In this View you will be able to filter by different breast cancer scenarios and treatments distributions like chemotherapy, endocrine therapy, surgery, radiotherapy depending on the age, number of excised nodes, invasive tumor size … For each selection toxicity, non-compliance details and compliance rates are plotted.
Components
- Dropdowns for filtering by scenario selected, X axis and Y axis with numerical variables as age, number of excised nodes, invasive tumor size, lymphatic invasion…
- Interactive Scatter plot:
Numerical variables axes and categorical variable previously selected in X Axis and Y Axis dropdown grouped by treatment chemotherapy, endocrine therapy, surgery and radiotherapy
Click on interaction by treatment bubble
- Pie charts:
Percentage of patients who has been suministrated any of the treatments (chemotherapy, endocrine therapy, surgery and radiotherapy) depending on the scenario selected
Percentage of toxicities (yes/no) depending on the scenario selected
Percentages of toxicities types as bruising, breast pain, loss weight, vomit, harm pain, faint depending on the scenario selected
- Histograms:
Compliance rates: Grouped by yes/no and types as clinical parameters, treatment (clinical)…
Non-compliance details distribution by treatment (patient) as distance constrains or monetary constrains; treatment (clinician) as no surgery available and clinical parameters as age, comorbidities and tumor size.
COVIDSS
General View
It allows to filter by patients characteristics in order to improve the data analysis, find new evidences and relation between features too.
Components
- Dropdowns for filtering by sex (Male/Female/Both), intensive care unit (Yes/No)
- Interactive Scatter plot:
Numerical variables axes and categorical variable for grouping selected by user
Greater than age slider
Click on interaction between scatter and pie chart
Diagnostics View
It shows a Sankey diagram which relates different categorical characteristics depending on the columns selected.
Components
- Dropdown for selecting the columns to relate and sankey diagram.
Laboratory View
It allows to see the distributions of the value results of laboratory determinations and shows diagram with the top ten determinations with worst results.
Components
- Dropdown for selecting a determination and histogram for representing the top ten worst results.
COVID Diabetics Patients View
It allows to compare the different characteristcs beetween patients who sufferend from diabetics with general COVID patients.
CAPTAIN
Description
Older adult evolution analysis, developed to assist caregivers, in the context of the CAPTAIN project, to manage virtual coaching of a population of older adults, while providing personalized assistance, to:
a) assess the effectiveness of personalized coaching plans
b) identify missing aspects that require coaching in the population
c) design specific coaching plans following a fair-comparison approach
d) suggest coaching plans to new participants based on past experiences collected from prior participants.
Components
- Histograms: This visualization allows the user to understand the distribution of the population scores for each assessment and to easily check if there has been an overall change in a particular variable.
- Dendrograms: For grouping the participants based on their results in the first assessment
- Pie chart: provides the resultant pie chart of the dendrogram. It shows what is the predominant cluster and the percentage of population on each cluster.
- Radar chart: Radar chart of the clusters obtained in the dendrogram. For example, it shows the variables associated with the Physical and Cognitive fields for each cluster.
- Parallel graph: It allows to represent the changes in each field per participant, determined by the color of the cluster the participant belongs to.
- Evolution graph: It shows the evolution of selected characteristics as nutritional activity, social activity… and coloring by physical improvement or another feature.
ADELAIDE
Within the framework of aDELAIDE, Vicomtech is bringing their expertise in medical image processing to support the development of automatic analysis and tagging of images to classify them and detect structures or pathologies, based on AI.
Components
- Button to select the dicom directory folder.
- Button to load and visualize the study saved in .dcm format from the selected directory.
- Graph to visualize 2D slices of the study plotted with plotly, together with a vertical slider that allows selecting the 2D slice of the volume.
- Button to render the 3D volume of the loaded study.
- Interactive dash VTK volume representation of the dicom image.
- Button to process study: preprocess and predict organ status (Visible/Not visible/Partially visible) and location using a tensorflow model trained previously.
- Data table showing the organ status results after the processing.
- Four graphs to visualize 2D slices of the organ VOIs extracted from the predicted organ locations, each graph connected to their corresponding vertical slider.
Project Summary
Radiology departments have huge image DBs collected over decades and stored in hospitals’ PACS. Yet, there is a tremendous opportunity to improve this image data mining, sorting and organising it efficiently. Extracting, harvesting and building large-scale annotated datasets from these DBs is very important for patient management and clinical development. Adelaide, a cognitive radiology project financed by Eurostars and developed by Cetir, Medexprim and Vicomtech, aims at defining, developing and deploying a solution that allows querying the Picture Archiving and Communication Systems (PACS) based on tags generated from complex image analysis and semantic radiological report mining to facilitate patient cohort generation for clinical trials or research studies.
EREETA
Project Summary
Exploratory Data Analysis (EDA) and Machine Learning training of models in the cloud.
In this project, a service located in the cloud allows to explore, clean and train different types of datasets. It takes in account two main parts: EDA and ML.
ML View
- Boxplot : Cross Validation train (green) and test (red) depending on the metric selected (AUC, accuracy, precision, recall…)
- Curves : Different ROC curves for each model, precision- recall
- Histogram: Different metrics result representation
- Confusion matrix: Representation of confusion matrix of the model and threshold selected
- Histogram : Feature importance distribution
EDA View
- Histogram: For showing missing values in feature
- Distribution (Curve and Histogram): Parameters distribution for checking if any variable has impact in the outcome
- RadarChart and boxplot : For showing the variables distribution depending on the outcome result.
Contact with us
Visit us
Parque Científico y Tecnológico de Gipuzkoa, Paseo Mikeletegi 57,
20009 Donostia – San Sebastián (España)
Call us
+(34) 943 309 230
Email us
tech.transfer@vicomtech.org