Sonio
Sonio makes prenatal ultrasound software for sonographers, maternal fetal medicine specialists and the obstetricians who read the scans, built as one platform with AI reading aids and a reporting and image management layer. Images and clips flow from the ultrasound machine through Sonio Edge into Sonio Detect, which recognizes supported fetal views and anatomical structures as the exam is acquired and checks each view against standardized quality criteria, such as whether a structure is present or the zoom is right. Items left unverified are highlighted for the sonographer, who can reassign an image or change its status. The FDA cleared Detect v3 under K252433 in March 2026 after testing on 22,496 fetal ultrasound images kept apart from development data, and Detect is CE marked as a Class IIa device.
Sonio Pro manages worklists, DICOM images, report templates, CPT coding and HL7 exchange with record and billing systems, extracts the best clips from cine loops, and shares images with patients through QR codes and protected links. It works with ultrasound machines from any manufacturer, while Detect v3 is cleared on validated GE, Canon, Philips and Samsung systems. Sonio Suspect is a separate reading aid that flags supported abnormal findings for interpreting physicians, cleared first as a web version and in June 2026 as a scanner integration that runs on the equipment. Pediatrix Medical Group runs the platform at a maternal fetal medicine practice.
Founded in 2020 in Paris, with cofounder Cécile Brosset as chief executive, Sonio raised about $30 million before Samsung Medison bought it for about $92 million, closing in August 2024. It operates independently from Paris with US offices in New York and Cambridge, Massachusetts, and sells through demos and institutional agreements.
Capability Axes
An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read
Sonio Detect uses a learned image model to identify supported views and structures and check quality criteria, and Sonio Suspect uses a separate model to flag supported abnormal findings for a physician. Those features depend on the software interpreting images. Sonio Pro adds worklists, reporting, image review, clip sharing, CPT coding and exchange with hospital systems, and it runs without the AI reading aid.
The models add real time feedback and structured findings inside that workflow, which an ultrasound team uses as it acquires images or when a clinician reviews the exam. Samsung Medison bought the company for its AI, with plans to fold it into its ultrasound roadmap.
Detect matches images to views and highlights a criterion in yellow when the condition was not verified or the view was not recognized. The clinician can reassign an image, change the verified status or enter a characteristic by hand during or after the exam. Suspect presents findings to the interpreting physician as a concurrent reading aid, and a prioritized reader dashboard puts the most urgent cases in front of physicians first.
The user terms keep interpretation and diagnosis with the clinician. The sonographer page's promise to never overlook the implications of patient risk factors is tied to dynamic exam protocols rather than to AI output. No numeric confidence threshold or required escalation to a second reader is set.
Sonio describes deep learning for medical image analysis, with Detect's modules combining image analysis and OCR neural networks. The regulatory summary identifies specific versions, supported views and the change from the earlier model, and Detect v3 adds localization of views and structures with bounding boxes.
The filing describes an evaluation dataset kept apart from development and the device's operating points, and the training set behind Detect was put at more than a million real ultrasound images in 2024. The training data is not characterized by source and cohort, the full architecture is not given, and when a new model reaches a practice after a clearance is not stated.
Detect and Suspect are cleared for specified functions, and their FDA summaries set out which changes separate each device from its prior version. The 2026 Suspect submission describes the same core model running locally on a third party scanner, alongside its earlier configuration in Sonio's managed cloud. Samsung Medison has owned the company since 2024, and the platform keeps working with machines from every manufacturer rather than only Samsung's.
Hosted infrastructure uses only HDS certified cloud providers. The cloud providers in the reporting and image workflows are not named, the parties that receive exam data for hosted analysis are not listed, and whether any third party keeps or can reuse that data is not stated.
The Detect v3 evaluation used 22,496 fetal ultrasound images kept separate from its development set. The FDA summary reports sensitivity of 88.6% for detecting 14 first trimester views and 90.1% for 40 second and third trimester views, which measures view detection rather than diagnosis. Clip extraction from cine loops rests on an internal study of time saved on trackball scrolling.
Pediatrix adopted the reporting and imaging workflow at a maternal fetal medicine practice before this version was cleared. The bench evaluation does not measure outcomes for pregnant patients, and the Pediatrix rollout comes without a before and after comparison of clinical performance or report turnaround.
Exam images and clips move from the ultrasound machine through Sonio Edge into a workflow that can be cloud hosted, on premises or inside a private cloud, with encryption in storage and transit and data held in its region. Patients receive images through QR codes and access protected links. The product terms require users to report suspected performance and safety problems, which gives a concern a route to the company.
No fixed retention window is given for inputs, no method is stated for deidentifying scans used to improve models, and no AI training restriction covers the product. What happens after a reported model problem, from triage to telling other sites, is not described.
The security page says the platform meets HIPAA privacy and security requirements alongside the EU General Data Protection Regulation. The July 2026 product terms govern individual access, refer to an institutional master services agreement and place usage obligations on clinicians who handle health data, while the website privacy notice covers visitor information. No business associate agreement is mentioned, and which configuration or subscription would include one is not stated.
Without the agreement text, the downstream obligations on firms that handle ultrasound images or records for the hosted product cannot be seen, which matters most for US practices sending exams to the cloud.
The company states SOC 2 Type 2 compliance, holds ISO 13485 certification for its quality management system and uses only cloud providers certified under France's HDS health data hosting framework. Controls include encryption at rest and in transit, a web application firewall, intrusion detection and zero trust access to internal infrastructure and data, with a security white paper available on request.
These cover different scopes, since ISO 13485 governs device quality rather than information security. No SOC 2 auditor's report or reporting period is available to review, so the scope of the attestation for cloud image processing and for local equipment integrations is not clear.
The FDA cleared Sonio Detect v3 on March 16, 2026 under K252433 as a concurrent reading aid that detects fetal views and structures and checks quality criteria during acquisition and interpretation. Sonio Suspect has a separate indication for helping interpreting physicians with supported abnormal findings, with a web version under K243614 and a scanner integration under K261519.
Sonio Pro is a US Class I medical image communications device and is not a medical device in the European Union. Detect also carries Class IIa CE marking under the EU Medical Device Regulation, with BSI as notified body. The Detect clearance limits use to compatible systems and specified views, and each designation applies only to its stated product.
The Detect v3 summary describes risk analysis, design reviews and validation as part of device development under an ISO 13485 quality system, and the company has a head of regulatory affairs and quality who also serves as privacy compliance officer. The summary lists analyses by scanner manufacturer, image quality, gestational age, maternal age and body mass index, and by race and ethnicity where appropriate.
Results are reported for the combined test population, without a full set of subgroup figures. A scientific advisory board of clinicians and a stated commitment to testing the robustness of the image models complete the picture. Together these describe a controlled development process, but how closely performance matches across the measured groups is not shown.
The July 2026 user terms ask professionals to report suspected errors or malfunctions promptly to both Sonio and their institution, and the support workflow takes issues about a specific exam, about images and AI output, and about patient data that needs repair. Detect lets the clinician change assigned views and verification statuses during review, so a wrong label can be fixed before a report goes out. The terms present the product without a warranty and leave diagnosis to the user. Recourse therefore runs through reporting an error to both parties and correcting it in the exam. No correction deadline and no financial remedy for an incorrect finding are offered.
Sonio Pro acts as a worklist for the ultrasound machine, takes DICOM images, connects to PACS and VNA archives and exchanges HL7 messages, including ADT, SIU, ORU and ORM, with FHIR and custom interfaces also supported. Patient and appointment data comes in, while reports, dating details, CPT and ICD 10 codes and DFT billing messages go out, and reports reach outside referring providers by fax.
The product release summary names Epic and athenahealth among the integrations, and Pediatrix is a named clinical deployment. The record system behind the Pediatrix rollout is not named, and which flows run in both directions there is not shown. Detect compatibility follows its cleared scope.
The platform runs in the cloud, on premises or in a virtual private cloud. Hosted deployments use isolated regions for the United States and Europe, with data kept inside its region behind encryption at rest and in transit, a web application firewall, intrusion detection and zero trust controls on internal access.
The cleared scanner integration of Suspect runs the model locally on compatible equipment, and the FDA submission separates that from the earlier version that processed data in Sonio's cloud. These are distinct paths a practice can choose between. The hosting providers are not named, and tenant controls beyond regional isolation, encryption and restricted access are not described.
Every purchase path starts with a demo, of the whole platform or of a single product, and the user terms describe access through an agreement between the company and the customer's institution. Customers can choose reporting or image analysis functions for their workflow, and cloud software updates are included, with the homepage promising zero maintenance costs.
No prices are listed, and neither is a billing unit, a contract minimum or a separate charge for building an interface with a hospital record or billing system. The hosted, on premises and scanner configurations each come through the sales team, and how an institution's fees are calculated is not explained.
The products serve prenatal ultrasound in obstetric and maternal fetal medicine workflows, with versions shaped for sonographers, readers and clinical managers across management services organizations, private practices, academic centers and community centers. Detect v3 was evaluated on fetal scan images and is limited to validated GE, Canon, Philips and Samsung systems, with the FDA filing listing supported views and structures by trimester.
Suspect is a reading aid for specified findings within its own gestational age range. Pediatrix runs the reporting platform at a maternal fetal medicine practice. Coverage is a defined specialty with evaluation for specified views and equipment, and the clearance does not extend to use outside the indicated ultrasound workflow.
Alternatives to Sonio
Other products that do the same kind of work as Sonio, and what each one does.
Gleamer SAS is a radiology AI company based in Saint-Mandé, on the edge of Paris, and since 2 March 2026 part of DeepHealth, the digital health subsidiary of the US imaging services operator RadNet, Inc.
Butterfly Network makes handheld ultrasound probes that image with a semiconductor chip in place of piezoelectric crystals, so a single probe connected to a phone or tablet covers the whole body where conventional systems use several transducers.
Genius AI Detection is the breast imaging AI software line of Hologic, Inc. of Marlborough, Massachusetts, licensed to breast imaging buyers separately from Hologic's wider diagnostics, surgical and skeletal businesses.
Philips Lumify is a handheld ultrasound transducer that plugs into a smart device the customer already owns, over USB on Android or Lightning on Apple hardware, with the app downloaded from the ordinary app stores.
Copenhagen company, founded in 2018, whose Apollo software reads brain MRI while the scan is still running.
DentalMonitoring analyzes smartphone images of teeth and orthodontic appliances between office visits, for orthodontists and dental practices in 50 countries.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | BAA Tier | Implementation | Source |
|---|---|---|---|---|
|
Contact for pricing
|
Institutional agreement through sales | No business associate agreement tier is stated. HIPAA compliance is claimed for the platform | No interface or implementation fee is listed. | Vendor Published |
Sonio offers demos of its platform and of individual products. Prices, billing units and implementation charges for record and imaging interfaces are not listed.