Solution and Technology
The heart of GeoNomy
State-of-the-art algorithms for fast and reliable interpretation
Real-time analysis and degree of belief
GeoNomy is using advanced computer imaging techniques and deep learning models which automate, in real time, the extraction, analysis, identification and understanding of useful information from ultrasound images, just the way the operator brain proceeds. The underlying algorithms are trained on thousands of ultrasound images which are collected on real patients in real conditions and qualified by medical professionals thanks to ergonomic tools and automated labeling methods developed by GeoNomy. By incorporating a probabilistic approach to the neural networks, GeoNomy’s solution incorporates an uncertainty factor (degree of belief) accompanying each model’s prediction, and further contributing to the acceptability of AI in the medical field.
GeoNomy, a product
AIPI - Automated Image Pattern Interpretation
GeoNomy offers an autonomous software module (an application) which automatically interprets ultrasound films and detects predefined pathologies if they exist, from any ultrasound machine without modifying the usual examination procedure. It is a Software as a Medical Device (SaMD). This application is called AIPI, it operates in real time when the practitioner performs the examination or can be used after recording a film.
Two ways to use the system
Our product is available in 2 formats:
- an iOS or Android mobile application for tablet or smartphone connected to the ultrasound scanner or ultrasound probe, allowing local processing of information in real time, or
- a web application downloadable on any browser with a secure connection allowing offline image processing in the cloud.
An Innovative Solution
Cutting-edge technology
GeoNomy processes ultrasound films in an innovative way incorporating two superimposed layers of analysis resulting from the inference of 2 distinct AI algorithms:
- a description of the organs, via a semantic segmentation algorithm,
- verification of the normal or pathological nature of the elements described, via an anomaly detection algorithm.
Through the incorporation of Bayesian deep learning techniques, an uncertainty factor is calculated for each prediction and compared to a threshold uncertainty determined during the clinical study. This uncertainty constitutes a degree of belief critical for health care professionals.
An innovative use
Our solution impacts the entire healthcare value chain, enabling earlier diagnoses beyond specialty clinics. It creates new uses by allowing all medical and paramedical players to conduct an imaging examination and for patients to have easier access to ultrasound examinations, particularly in medical deserts, teleconsultation facilities or developing countries.