IMPatienT🗂️️: an integrated web application to digitize, process and explore multimodal patient data.
IMPatienT (Integrated digital Multimodal PATIENt
daTa) is a web application developped in the
MYO-xIA project for patient data digitization and exploration.
It features a standard
vocabulary
creator, optical character recognition (OCR), natural language processing
(NLP), image annotation and segmentation using machine learning, interactive
visualizations and automatic diagnosis prediction.
Demo Usage Instructions
If you are on the demo instance (https://impatient.lbgi.fr), you can login
on top
right with
login: demo ; password: demo
The application is
preloaded with a database of histology reports and a standard vocabulary for
muscle biopsy. Data on this demo instance is reset once every day.
For the PDF automatic analysis, you can find a sample PDF
HERE
working with the preinstalled standard vocabulary. A sample image for image
annotation can be found
HERE.
Partners
IMPatienT is developped and used in collaboration with the Morphological Unit of the Institute of Myology of Paris. A production instance is deployed to help discovering new relevant features for congenital myopathies classification and diagnosis.
IMPatienT🗂️️ Abstract
Background
Medical acts, such as imaging, generally lead to the production of several medical text reports that describe the
relevant findings. Such processes induce multimodality in patient data by linking image data to free-text data and
consequently, multimodal data have become central to drive research and improve diagnosis of patients. However, the
exploitation of patient data is challenging as the ecosystem of available analysis tools is fragmented depending on
the type of data (images, text, genetic sequences), the task to be performed (digitization, processing, exploration)
and the domain of interest (clinical phenotype, histology…). To address the challenges, the analysis tools need to
be integrated in a simple, comprehensive, and flexible platform.
Results
Here, we present IMPatienT (Integrated digital Multimodal PATIENt
daTa), a free and open-source web application to digitize, process and explore multimodal patient
data. IMPatienT has a modular architecture, including four components to: (i) create a standard vocabulary for a
domain, (ii) digitize and process free-text data by mapping it to a set of standard terms, (iii) annotate images and
perform image segmentation, and (iv) generate an automatic visualization dashboard to provide insight on the data
and perform automatic diagnosis suggestions. Finally, we demonstrate the usefulness of IMPatienT on a corpus of 40
simulated muscle biopsy reports of congenital myopathy patients.
Conclusions
IMPatienT is a platform to digitize, process and explore patient data that can handle image and free-text data. As
it relies on a user-designed vocabulary, it can be adapted to fit any domain of research and can be used as a
patient registry for exploratory data analysis (EDA). A demo instance of the application is available at https://impatient.lbgi.fr.
MYO-xIA Project
The MYO-xIA project aims to collect data from patients with
congenital myopathies in order to analyse them using
explainable AI approaches. In the field of congenital
myopathies research, there is an important need for
digitization and standardization of patient data in order
to use modern data analysis methods.
This project has tree mains
objectives:
- Gather and format patient data to be ready for automatic exploitation (patient histology-report form, image annotation, OCR/NLP)
- Help the discovery of new phenotype-genotype-histology association for each congenital myopathy subtype (standard vocabulary creation tool, per gene and per class histology feature statistics)
- Provide tools for automatic diagnosis based on patient data (live prediction of congenital myopathy subtype with histology reports, images, genetic data and phenotype)
Contact
The main maintainer is:
Corentin Meyer - PhD Student @
CSTB Team - iCube - University Of Strasbourg co.meyer@unistra.fr
Citing IMPatienT🗂️
[placeholder]