Medical 3D Data Annotation of MRI scans
Find out how we helped Milvue by annotating 3D cerebral vasculature from MRI images so they could train their AI medical models.
Must-read Case Studies
Foodtech App
3D Point Cloud annotation tool
Aerial images
3D Medical annotation
...et bien d'autres qui sont confidentiels!
Understanding 3D scans in medical imaging requires unparalleled expertise and precision.
Milvue, with its AI-based technology, reinvents radiological analysis, speeds up diagnosis and enriches patient follow-up. This innovation places Milvue at the forefront of modern medicine, offering solutions that integrate naturally into the routines of healthcare professionals.
Context
CONTEXT
Thanks to AI-assisted medical imaging, our customer Milvue makes it easier for radiologists to practice.
Milvue’s mission is to be on the side of practitioners and radiologists to advance medical imaging.
Milvue develops algorithms to help radiologists analyze medical data, in particular to detect anomalies in brain scans and prevent strokes.
To train these algorithms, we need 3D medical data in which the cerebral vascular network is precisely annotated.
The annotation task consisted in delineating all major blood vessels appearing distinctly on the MRI image.
To achieve this, Milvue asked People for AI for assistance.
OUR SOLUTION
OUR SOLUTION
For this expert task, we selected and trained 2 of our best annotators.
Training two annotators has allowed us to maintain continuous data production, even when one is absent.
To annotate this data, we used the tool recommended by the customer, 3D Slicer, installed locally on the computers in Madagascar. This enabled us to keep the tool’s latency to a minimum, thus improving the annotator’s comfort and speed.
15 to 25 minutes of annotation per MRI.
Two annotators have become experts in 3D medical annotation.
A proven method with a question file, video captures of the annotation and regular exchange sessions.
Matthieu Warnier
Data Labeling Director @ People for AI
« The task of 3D annotation of Milvue’s cerebral vascular network is one of the most difficult we’ve had to perform. The 3D annotation tool Slicer proved to be very powerful but complex to use. It was a great satisfaction to be able to train experts within our teams who proved capable of carrying out the task with the level of quality and speed required. »
OUR IMPACT
OUR IMPACT
Milvue saved of 250 data scientist hours on the first annotation phase.
Milvue validates the ability to upgrade the skills of people with no initial knowledge on this specific field.
Milvue's first stroke detection models have progressed following this first campaign.
Malo Huard/Rudy Noyelle
CSO @ Milvue Client of People For AI since 2023
« The PFAI teams showed a strong desire to improve their skills in this non-trivial annotation task. We had some doubts about their ability to achieve the expected quality during the first weeks of the experiment. They never gave up, however, and ultimately succeeded beyond our expectations. »
Our ability to implement the 3D Slicer tool and to train annotators on the long run enabled Milvue to rapidly and cost-effectively build up a high-quality medical dataset.
They trust us
Our labeled data will exceed your expectations.
We provide training data. We do not provide your data to anyone else and we will not subscribe you to any newsletter without your consent.
