.svg){kind=logo}
Transpara
AI reading support for screening mammograms from Dutch company ScreenPoint Medical. Transpara marks suspicious regions on 2D and 3D exams, scores each case by risk, assesses breast density and compares priors, all inside the radiologist's existing PACS.
What is Transpara?
Transpara is a suite of AI applications built by ScreenPoint Medical B.V., a company based in Nijmegen in the Netherlands and founded by Professor Nico Karssemeijer and Professor Sir Michael Brady, both specialists in quantitative breast image analysis. It reads screening mammograms, both 2D full-field digital mammography and 3D tomosynthesis, and returns its findings inside the radiologist's existing PACS and hanging protocol rather than on a separate workstation.
The suite has three modules. Transpara Detection marks regions that look suspicious, scores each on a 98-point scale, and places the whole exam in one of three categories, Elevated, Intermediate or Low, based on its highest region score. Low-risk exams are deliberately left unmarked so they can be reviewed quickly; ScreenPoint cites a negative predictive value of 99.97% for that category and says such exams account for 70% or more of a screening list. Transpara Density produces a BI-RADS-like composition category alongside a volumetric density measurement, which is what makes it relevant to density reporting under the US MQSA Final Rule. Temporal Comparison tracks a finding across up to three prior mammograms, from nine months to six and a half years back, to judge whether it has changed or stayed stable.
Transpara is a regulated medical device, and ScreenPoint publishes the paperwork rather than only the claim: FDA 510(k) clearances K241831 and K232096, a CE mark under the MDR through notified body 0344, and a mention of Health Canada licensing, with clearance claimed in 40 countries. The company also quotes its CE intended-use wording verbatim, covering decision support, use as an independent second reader, or workflow optimisation, including as a partial replacement for the second reader where double reading is practised.
The evidence base is unusually deep for medical AI: more than fifty peer-reviewed studies and three randomised controlled trials, MASAI in Sweden, PRISM in the United States and AIMS in Norway, all run by third parties rather than by the vendor. Scale figures, by contrast, are self-declared and vary between pages, from eleven to thirteen million mammograms processed across more than thirty countries.
What it does
- Mark regions suspicious for breast cancer on 2D and 3D screening mammograms
- Score each region on a 98-point scale and sort the exam into Elevated, Intermediate or Low risk
- Leave low-risk exams unmarked so they can be triaged to faster or single reading
- Assess breast density with a BI-RADS-like category and a volumetric measurement
- Compare a finding against up to three prior mammograms, from nine months to six and a half years back
- Deliver all of it inside the existing PACS and hanging protocol, without a separate workstation
- Feed density figures into reporting required by the US MQSA Final Rule
When to use Transpara / When not to
A quick filter to help you decide if Transpara is the right fit.
When to use Transpara
- Breast radiologists reading high volumes of screening mammograms who want a consistent second look on every case
- Organised screening programmes considering a move from double reading to AI-supported single reading
- Imaging departments running mixed equipment, since the algorithm is presented as vendor-agnostic across gantries and PACS
- US practices that must report breast density under the MQSA Final Rule and want it produced during the read
- Research groups and screening authorities that require a published, independently trialled evidence base before adopting AI
When not to use Transpara
- Patients and members of the public: this is a clinical device sold to institutions, with a separate patient-facing site elsewhere
- Anyone hoping to replace the radiologist, since autonomous and rule-out use is explicitly outside the current FDA clearance and CE certificate
- Teams wanting five-year risk prediction today, as that algorithm is labelled investigational use only
- Buyers who need a published price, a trial licence or a self-service sign-up before talking to a salesperson
- Departments looking for AI on other modalities such as breast MRI or ultrasound, which remain research projects rather than products
How to use Transpara
A typical end-to-end flow, from setup to results.
- Try the interactive virtual demo, which is open on the ScreenPoint site without filling in a form
- Submit the demo request form on the Learn More page to be routed to the right representative
- Discuss scope with that representative or with a regional distributor, since no price is published
- Check whether an AI platform you already run, such as Ferrum Health, Microsoft's Precision Imaging Network or Aidoc, can activate Transpara directly
- Scope the deployment with your own IT team, which ScreenPoint presents as a joint project
- Integrate the modules into the existing PACS and hanging protocol so results land in the normal reading environment
- Read exams concurrently, with region marks and the exam risk category displayed alongside the images
- Use the Low Risk category to triage cases toward faster or single reading, according to your local protocol
- Read density and prior comparison results in the same pass, where those modules are licensed
- Contact support by email or by the EMEA and US phone lines, and take the routine algorithm updates as they ship
Pros & Cons
Pros
- An evidence base that is rare in medical AI: more than fifty peer-reviewed studies and three randomised controlled trials
- Those trials, MASAI, PRISM and AIMS, are run by independent third parties, not by the vendor
- Regulatory paperwork is actually published: FDA 510(k) numbers K241831 and K232096, CE notified body 0344, MDR marking
- The CE intended-use wording is quoted verbatim, and the company states plainly what is not covered by it
- Vendor-agnostic by design, working across gantries, acquisition types and reading environments
- Results appear in the existing reading workflow, with no extra workstation and no second login
- A named Data Protection Officer, a full postal address and a documented vulnerability disclosure process
Cons
- No public pricing whatsoever: the site has no pricing page, no plans and no order of magnitude
- No terms and conditions online either, so liability, service levels and ownership of outputs are undocumented
- The published privacy statement does not cover the processing of medical images, which is the part a hospital most needs
- Scale figures contradict each other between pages: eleven, twelve and thirteen million mammograms, fifty or fifty-five studies
- The headline performance numbers all come from a single trial, MASAI, on 2D mammography in Sweden
- Nothing is published about data hosting, no DPA is offered, and not one subprocessor is named
- No quality-system certification is published, and ISO/IEC 27001 is claimed only as an alignment rather than a certification
Pricing & Plans
No pricing is published. ScreenPoint Medical operates no public price list, and the site carries no pricing page at all: a figure is obtained by submitting the demo request form and speaking to a representative or a regional distributor. No permanent free plan and no free trial are announced. The freely accessible virtual demo is a demonstration of the product, not a trial of it. The USD 16 million announced in April 2026 is investment raised by the company and says nothing about what the software costs.
Data, GDPR & hosting
A consolidated view of how Transpara handles your data.
GDPR overview
Implementation is concrete, but the acronym is never used: the word GDPR appears nowhere on the site, which speaks instead of European Data Protection Laws and applicable Privacy and Data Protection laws. The substance is present. A Data Protection Officer is designated and reachable at dataprotection@screenpointmed.com or by post at Toernooiveld 300, Nijmegen. The privacy statement, last updated February 2025, sets out the legal bases relied on and enumerates the rights of access, rectification, erasure including at third-party recipients, restriction, objection and withdrawal of consent, then points to the right to complain to a local supervisory authority. Transfers outside the EEA are acknowledged, with standard contractual clauses named as the safeguard. As the publisher is established in the Netherlands, no Article 27 representative is required or designated.
Who owns the data?
ScreenPoint Medical B.V. acts as data controller for the personal data of website visitors, job applicants, vendors and business contacts, and states that internal access is limited to a need-to-know basis. For the medical images themselves it acts only as a processor on behalf of the hospitals and clinics using Transpara, and its published privacy statement explicitly places that processing outside its own scope. Ownership of images, algorithm scores and reports is therefore documented nowhere public: the site carries no terms and conditions and no customer contract. Personal data may also be shared with the US affiliate ScreenPoint Medical Inc. for support functions.
Reuse rights
Because the site publishes no terms and conditions, there is no public statement on whether a customer may reuse the images, the Transpara scores or the resulting reports without asking permission. What the privacy statement does cover is ScreenPoint's own use of contact-level personal data: managing customer and vendor relationships, invoicing, answering enquiries, recruitment, direct marketing, physical and IT security, audience measurement, due diligence during a merger or takeover, and cooperation with authorities. It relies on contract performance, legitimate interest, legal obligation or consent, and states that no automated decision is taken without human intervention. Whether customer data is ever used to train or retrain the models is not addressed anywhere on the site, in either direction. Reuse of the website material itself is restricted: ScreenPoint prohibits reproduction without express permission and says it enforces its trademarks.
Data retention & training
Hosting summary
No hosting information is published. The site names no hosting country, no region, no provider and no deployment model for Transpara itself, and gives no indication of whether the software runs on-premises, in the customer's own cloud, or on infrastructure operated by ScreenPoint. The privacy statement does acknowledge that personal data is sometimes transferred outside the European Economic Area because of the way the business is structured, and names standard contractual clauses as the safeguard. It lists categories of IT supplier, hosting, email and collaboration, web analytics, help desk and CRM, without naming a single one of them. Data may also flow to the US affiliate ScreenPoint Medical Inc. for support functions. One point deserves care: the marketing site sits behind Cloudflare and is built on HubSpot, and a DNS lookup resolves to a United States anycast node. That describes where the brochure is served, not where patient images are processed or stored. A hospital evaluating the product has to put the hosting question to the vendor directly.
Things to keep in mind
Risks and trade-offs to weigh before adopting Transpara.
- Autonomous and rule-out use is not FDA-cleared and not covered by the current CE certificate, even though the company publicises research on it
- The five-year risk prediction algorithm, benchmarked against iCAD and Mirai, is labelled investigational use only
- Automation bias is the classic risk with a second reader that is always available: an unmarked Low Risk exam is a probability, not a guarantee
- The same site quotes eleven, twelve and thirteen million mammograms, and fifty or fifty-five studies, so treat any single figure with care
- Headline performance comes from one trial, on 2D mammography, in Sweden: transferability to other populations and to 3D is not established by the site
- ISO/IEC 27001 is claimed as an alignment, not a certification, and no quality-system certification such as ISO 13485 is published
- With no terms and conditions online, questions of liability for a missed cancer and ownership of the outputs are left to the private contract
Setup & Integrations
Technical difficulty
Not a self-service product. There is no download, no sign-up and no public API: deployment is a project run with the customer's IT team, integrating Transpara into the existing PACS and hanging protocol so results appear in the radiologist's normal reading environment rather than on a separate workstation. Where a compatible AI platform is already in place, such as Ferrum Health, Microsoft's Precision Imaging Network or Aidoc, activation can go through it instead. ScreenPoint claims broad PACS coverage, non-disruptive integration and routine algorithm updates. No timelines, hardware prerequisites or technical documentation are published.
Integrations
Behind Transpara
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Transpara.
Frequently asked questions
Which types of mammogram does Transpara read?
What is its regulatory status?
Can Transpara read a mammogram without a radiologist?
How much does it cost?
How does it fit into an existing workflow?
Where do the 29% detection and 44% workload figures come from?
What does the Density module actually do?
Is customer data used to train the models?
Who do I contact about data protection?
Should you pick Transpara?
Transpara is one of the better-documented pieces of clinical AI on the market, and that is worth saying plainly. ScreenPoint Medical publishes its FDA 510(k) numbers, its CE notified body, the verbatim wording of its intended use, and, unusually, a clear statement of what its clearance does not cover. The evidence behind it is not vendor marketing: three randomised controlled trials run by independent groups in Sweden, Norway and the United States, alongside more than fifty peer-reviewed studies. For a radiology department weighing up breast AI, that combination is a genuinely strong starting position.
The weaknesses sit on the commercial and documentary side. There is no published price, no order of magnitude and no billing model, and no terms and conditions at all, which leaves liability, service levels and ownership of the algorithm's outputs entirely undocumented. Nothing is said about where data is hosted, no data processing agreement is offered, and not one subprocessor is named. The privacy statement that is published covers website visitors and business contacts, and explicitly excludes the processing of medical images, which is precisely the part a hospital procurement team needs to read.
The site also contradicts itself on scale, quoting eleven, twelve and thirteen million mammograms processed on different pages, and fifty or fifty-five peer-reviewed studies depending on where you look. The headline figures, a 29% increase in detection and a 44% reduction in workload, come from one trial on 2D mammography in Sweden, not from a consensus across the literature.
The verdict, then, is split. Clinically, Transpara is about as well evidenced as breast AI currently gets. Commercially, you cannot evaluate it without a sales conversation, and several questions a buyer would want answered before that conversation are simply not answered anywhere public.
- Choosing a selection results in a full page refresh.
- Opens in a new window.