Sonio, founded in Paris in 2020, makes AI software that assists prenatal ultrasound examinations. Its EIC project said about half of congenital malformations are missed before birth and that current guidelines require 20 ultrasound views showing 42 anatomical structures, which doctors can forget. It set out to build tools that recognise views and structures, flag abnormal signs and recommend what to check next, according to the final report on CORDIS.
The European Innovation Council granted €2,500,000 towards a €5.72 million project from January 2023 to December 2024, under blended finance, according to CORDIS. The report says Sonio Detect won FDA clearance in June 2023 and a wider V2 in April 2024, going from 13 views and 15 structures to 26 views and 53 structures, while work continued on Sonio Suspect for anomaly detection. On 30 August 2024 Samsung Medison completed its $92.4 million acquisition of all of Sonio's shares after French government approval, Radiology Business reported. Sonio Suspect received FDA 510(k) clearance in February 2025.
Company file
- Legal entity
- Sonio SAS, VAT FR30883053571, Paris
- Founded
- 2020
- Owner
- Samsung Medison (100%, since August 2024)
- Products
- Sonio Detect, Sonio Suspect, reporting and quality tools
- Approvals
- FDA 510(k): Detect V1 (2023), Detect V2 (2024), Suspect (2025)
- EIC project
- d3pm, 1 Jan 2023 to 31 Dec 2024, closed
- EIC funding
- Blended finance
Where Sonio Stands Now
Sonio's website now presents it alongside Samsung Healthcare. Samsung Medison's chief executive said at the deal that "the real-time ultrasound exam guidance and quality assurance developed by Sonio is becoming more relevant" given workforce shortages. Sonio's revenue within Samsung is not reported separately.
"With the workforce shortages affecting the global healthcare industry, the real-time ultrasound exam guidance and quality assurance developed by Sonio is becoming more relevant to ensure high quality care."
Kyu Tae Yoo, CEO, Samsung Medison, September 2024The Technology: A Checklist That Reads the Scan
The report cites evidence that checklists raise exam completeness from 49% to 97%, but says they are rarely used because they are cumbersome. Sonio Detect recognises views and structures automatically, acting as a live checklist. The report also notes a limit: trained mostly on normal images, Detect "could also be prone to hallucinations" on some malformations, which is why the project built Suspect, trained to spot abnormal signs.
What the EIC Project Promised, and What CORDIS Shows
The project pursued three tools: Detect for completeness, Suspect for abnormal signs and Expert for rare disease recommendations. Detect was cleared twice during the project and Suspect shortly after. The report does not describe a cleared version of Expert. CORDIS lists the project as closed.
- Sonio DetectDeliveredFDA V1 and V2.
- Sonio SuspectDeliveredFDA, February 2025.
- Sonio ExpertUnverifiedNo clearance reported.
- Commercial outcomeExitAcquired, $92.4M.
The Money
Samsung Medison paid $92.4 million for Sonio four years after it was founded. The EIC award was blended finance; whether the EIC Fund held shares and received part of the sale price has not been disclosed.
What Cannot Be Checked
Questions for the EIC
Whether the EIC Fund invested in Sonio and what it received on the sale, and how many clinics use Sonio's software, are not public.
Why Europe Should Care
The EIC helped a French start-up build prenatal AI that won US approvals, and the result is now owned by a Korean group. The company remains based in France, but the value it created left European ownership.
What the project promised
AI to make prenatal ultrasound exams complete, flag malformations and support rare disease diagnosis.
What the record shows
Three FDA clearances and a $92.4 million sale to Samsung Medison.
The Verdict
Delivered, then sold
Sonio delivered the products the EIC funded and attracted a global buyer. Europe paid for part of the development; a Korean group now owns the outcome.
Sonio, Samsung Medison and the EIC have not been asked for comment for this article. It is based entirely on public records and published statements, linked throughout.
