Your career path is impressive, combining academic research, medicine, and entrepreneurship. What led you to found Matricis.ai?
After an academic career in physics and biophysics at the École Normale Supérieure in Lyon, followed by a doctorate at the University of Paris-Saclay, I wanted to move into industry. So I completed my training with an MBA for scientists at the Collège des Ingénieurs before taking my first job in reinsurance risk analysis.
Very quickly, I realized that this position did not stimulate me as much as I would have liked. It was then that my friend Élise Mekkaoui, who suffers from endometriosis, shared her idea of developing a diagnostic support tool. From my own experiences and those of the women around me, I was aware of the huge unmet need in gynecology and I was seduced by the technological challenge behind the product. I decided to join her and we founded Matricis.ai.
How does the artificial intelligence developed by Matricis.ai differ from other diagnostic support tools in radiology?
Like most diagnostic support solutions in imaging, our technology is based on computer vision: it teaches a neural network to recognize patterns on images using a very large number of example images where these patterns are annotated. This approach is commonly used in other fields, such as autonomous cars which must identify road signs on RGB images from their cameras.
What differentiates us is that we are the first to apply this technology to a gynecological pathology. Today, more than 500 AI devices in radiology are available on the market, covering various imaging modalities and pathologies, but the female pelvis is a real desert in terms of medical innovation. Endometriosis, in particular, poses a major challenge because its radiological manifestations are complex to identify, even for expert radiologists. The construction of a reliable learning database, and in particular the annotation of patterns, is therefore a real challenge, and this is where we bring a real breakthrough.
How is AI currently transforming the radiology sector and, more broadly, medicine?
Radiology was one of the first medical fields to benefit from advances in artificial intelligence. The first AI software (for the diagnosis of cardiac pathologies) was approved in 2016, and since then, the technology has continued to improve. Today, one in three radiologists in the United States uses AI on a daily basis to assist them in their work. These tools help reduce errors, standardize diagnoses, and speed up image analysis, which is crucial in a context of a shortage of specialists.
With the rise of language models like ChatGPT 3.5, AI is spreading to other medical fields, including automatic note-taking and medical report writing. These advances allow physicians to focus more on their patients by reducing their administrative burden.
What guarantees do you put in place to ensure that your tools respect medical ethics and do not reproduce biases related to gender or other factors?
Our device is a class 2.a medical device, subject to strict European standards that govern its development and deployment. This regulation also requires clinical studies to compare the performance of radiologists with and without the help of AI on the same images.
Regarding biases, we are not exposed to gender bias since we deal exclusively with gynecological pathologies. On the other hand, ethnic biases are a point of attention. In France, ethnic statistics are prohibited and our data is totally anonymized, we minimize this risk by building diversified databases from several medical centers in France and abroad.
You are supported by the Allianz France accelerator, which is focusing on generative AI for its new class. How is this an asset for your development?
Our long-term goal is for our software to be reimbursed by Social Security and insurance companies, which will require a well-structured regulatory and economic framework. The support of the Allianz France accelerator allows us to better anticipate this step and structure our reimbursement strategy.
In addition, Allianz is very committed to women’s health, which opens the door to promising collaborations, particularly around the prevention and early detection of gynecological diseases.
The Allianz France accelerator, a JFD partner and led by Sylvain Theveniaud, has a mission to scout innovation trends for Allianz France and its customers. It helps promote and develop impactful initiatives and encourages and organizes the creation of synergies with the accelerated startups. Nearly 50 projects have been launched with startups since the creation of the accelerator in 2015. The Allianz France accelerator has also been committed for a long time to better representation of women in the digital economy and is part of a desire to support high-potential startups.
Raphaelle Taub, co-founder Matricis.ai, and JFD Club member
Raphaelle Taub is a normalienne, a doctor of physics and a graduate of the MBA of the Collège des Ingénieurs. She co-founded and directs Matricis.ai, a deep-tech start-up from the APHP and INRIA that is developing the first artificial intelligence to help radiologists in the detection of gynecological pathologies.
