Features and roadmap

Principles and commitments of SPE3DLab

The origin of SPE3DLab is the creation of a tool allowing practitioners to implement “precision medicine”.

Datasources are multiplying, everywhere and on multile forms ; people refer to this as “Big Data” or “Smart Data”. These data are related to human activity as individual, as groups or institutions but also tell us about the natural and human environment at different scales. How to use this information? How to question it in the most relevant manner? Our societies find themselves at the crossroad of science and data engineering : how can we extract scientific knowledge from this data? How can we use this data to act and interact?

In the healthcare industry, initiatives aiming at federating or agregate data from various sources and different formats are multiplying, via public or private initiatives. Data repositories are appearing in hospitals ; federative platforms are being built at an international scale, public institutes like the French INDS are emerging. Most of these initiatives are static in the sense the data is stored in a unique place and that their processing only comes as later scientific or industrial projects. There is no integration nor direct retro-action between the production of data, its processing, the resulting decision making and the production of new data. The interactions between produced data, its dynamic processing and the decision of agents has not happened yet and current efforts are not going in that direction.


We propose the solution SPE3DLab that is a question oriented tool:

“I have a question and I would like to analyze my data  without bothering with the handling of the data and their processing”

“I work in a cardiology unit, I have access to my patients electronic records for scientific or medical purposes.”

SPE3DLab connects to the data and allows to work with it without having to take it out from their secured storage or having to compromise the anonymity of the patients.

“Can I predict the risk of re-hospitalization of my patients ?”

SPE3DLab will use the available data to extract the best predictive model.

“I’m happy with the model I have built, how can I use it in my daily practice?”

SPE3DLab makes the predictive model available in an autogenerated web page. The model can also be easily integrated to any Patient Relationship Management software or any dedicated mobile application.


SPE3DLab uses different modules for processing the data depending on the question asked and the type of data being analysed. Eventually, SPE3DLab will be able to interact with the user using natural language (in English, in French, …).

The development and the functioning of SPE3DLab relies on the following principles :

– opensource and documented project

– use of documented and opensource libraries

– use of well documented techniques recognized by the scientific community

– design choices concerning the processing of the data is documented and justified

– the ergonomics, accessibility and confidence in the tool must has a higher priority than the search of the best performance of the models

– the chain of treatment on the data (from the loading of the data to the restitution of the result) must be transparent, traced, and easy to export as executable code if the user asks for it

Presentation of modules

Descriptive analysis

SPE3DLab performs a complete descriptive analysis of your data (min, max, range, SD, mean, median, comparison of missing values, inclusion flowchart). The output of this analysis as tables and charts is ready to be used in your articles.

Hypothesis testing

SPE3DLab allows you to test different hypothesis : differences between two groups (with or without normality normality assumption), distribution tests, …

Causality analysis

SPE3DLab detangles the complex relationships between your variables.

Regression

SPE3DLab allows to find the best explanatory model for your variable of interest.

Clustering

SPE3DLab looks for clusters in your data.

Predictive analysis

SPE3DLab allows you to predict the value of the group of a new record.

Semantic analysis

SPE3DLab allows you to extract the data of text documents (Word, text, pdf) as structured data ready to be analysed.

Integration in your existing application and workflows

The models built with SPE3DLab are ready to be used by your team. Every saved model is made available as an auto-generated web-interface and a JSON API that your applications can invoke.