- HTX | Private AI for European Enterprises — ORCA · MANTA · KOI · TIDE/
- AI Use Cases for European SMEs | Private AI Solutions | HTX/
- AI for Healthcare | Private Clinical Decision Support GDPR | HTX/
AI for Healthcare | Private Clinical Decision Support GDPR | HTX
Private AI
for healthcare.
Clinical decision support, medical documentation, and patient data analysis. Clinical data never leaves your hospital. GDPR, MDR, and EU AI Act compliant.
classification error
lack data protection
on-premise
Healthcare needs AI. But not at any cost.
Artificial intelligence is transforming medicine: from diagnostic imaging to surgical planning, from genomic analysis to clinical decision support. The benefits are clear: reduced diagnostic variability, faster documentation, support in managing complex protocols.
But healthcare is also Europe's most regulated sector. Clinical data qualifies as special category data under GDPR (Article 9), subject to enhanced protection. The Medical Devices Regulation (MDR 2017/745) imposes strict requirements for software as a medical device. And the EU AI Act classifies healthcare AI as high-risk (Annex III, point 5), requiring transparency, human oversight, risk management, and data governance.
The result: healthcare organisations face a dilemma. On one hand, AI could significantly improve care quality and operational efficiency. On the other, most AI solutions available on the market — ChatGPT, Copilot, Gemini — send data to US servers, making regulatory compliance impossible for European health data.
38.4% of published studies on LLM use in healthcare do not implement adequate patient data protections. This is not acceptable in a real clinical setting.
Clinical AI that runs inside your hospital.
HTX has developed two products specifically for healthcare, both built on PRISMA — our private AI infrastructure deployable entirely on-premise within your hospital.
KOI — Clinical Decision Support
KOI is our AI system for ASA-PS classification in anesthesiology. It analyses the patient's clinical record and proposes a physical status classification with detailed clinical reasoning, reducing classification error by 89% compared to standard human assessment.
KOI's advanced reasoning models achieve 97.5% accuracy on standardised clinical benchmarks. The anesthesiologist always maintains full control of the final decision — KOI provides a second opinion, not a replacement for clinical judgement.
-89% classification errorORCA — Medical Documentation AI
ORCA applied to healthcare becomes a tool for querying clinical protocols, guidelines, standard operating procedures, and medical documentation in natural language. Clinicians ask questions and receive answers traced to the original source.
Consider how often a clinician needs to look up a specific protocol, verify a dosage, or consult a guideline. ORCA makes this information accessible in seconds, with precise citations and zero hallucinations.
Source-cited answersChemoMaker and Biovalley: AI for clinical protocols.
The partnership between HTX, Biovalley Investments, and the ChemoMaker project represents a concrete example of AI applied to clinical protocols in oncology.
ChemoMaker is a robot for automated preparation of chemotherapy drugs — a process requiring maximum precision and traceability. HTX integrated MANTA to enable natural language querying of pharmaceutical production databases: preparation times, dosages, batches, anomalies.
Healthcare staff can ask "Which preparations exceeded standard times last week?" and receive precise answers in seconds, without SQL skills or IT department involvement.
This is a concrete case of how private AI — deployed on-premise, with data never leaving the facility's perimeter — can improve clinical safety and operational efficiency in a high-criticality setting.
GDPR, MDR, AI Act: compliance by design.
In healthcare, regulatory compliance is not a nice-to-have — it is a non-negotiable prerequisite. Here is how PRISMA, the infrastructure powering KOI and ORCA, ensures compliance with all relevant regulations.
Personal data protection
Clinical data is special category data (Art. 9 GDPR) requiring the highest level of protection. With PRISMA, all data stays on-premise within your hospital infrastructure. No data is transferred to external servers. No data is used to train third-party models. End-to-end encryption across the entire pipeline.
Art. 9 GDPRMedical devices
The Medical Devices Regulation (MDR 2017/745) imposes specific requirements for Software as a Medical Device (SaMD). KOI is currently classified as Research Use Only (RUO) and is pursuing ISO 13485 and IEC 62304 certification for medical device marking planned for 2027. Every development phase follows quality by design principles.
ISO 13485 in progressEU AI regulation
The AI Act classifies healthcare AI as high-risk (Annex III). This requires: algorithmic transparency, human oversight, documented risk management, data governance, and post-market monitoring. KOI is designed with native human-in-the-loop: the physician sees the AI's reasoning and always decides independently.
High-risk — Annex IIIWhat changes in clinical practice.
Adopting private AI in healthcare produces measurable benefits across three dimensions: care quality, operational efficiency, and patient safety.
Reduced diagnostic variability
Inter-observer variability is a well-known challenge in medicine. In ASA-PS classification, physicians agree on the correct classification only 70% of the time. KOI reduces this variability with advanced reasoning models achieving 97.5% accuracy, offering a consistent, evidence-based second opinion.
Faster documentation
Clinicians spend a significant portion of their time searching for information in protocols and guidelines. ORCA makes this information accessible in seconds with natural language answers traced to the source. Less time searching, more time with patients.
Clinical safety
More accurate ASA-PS classification means better anesthesiological planning, more precise pre-operative risk management, and ultimately greater patient safety. Every AI decision is explainable and verifiable by the clinician.
Compliance without compromise
With HTX solutions, you do not have to choose between innovation and compliance. The AI runs entirely within your infrastructure, compliant with GDPR, MDR, and the AI Act. No compromise on protecting your patients' data.
Grounded in scientific research.
KOI is not a product born from a marketing idea. It is the result of a research project funded by the Friuli Venezia Giulia Region (LR 22/2022, art. 7), developed in collaboration with Ospedale del Quadrante (Ramsay Sante) for clinical validation.
The HTX study evaluated 11 language models on 20 standardised clinical vignettes from the most cited ASA-PS benchmarks in the literature. Results demonstrate that advanced reasoning models significantly outperform both first-generation models and average human performance.
This scientific approach — rigorous validation, standardised benchmarks, peer-reviewed publication — is what distinguishes a serious clinical AI product from a technology demo.
FAQ — AI for Healthcare
Are HTX healthcare AI solutions GDPR and AI Act compliant?
Yes. All HTX solutions run on PRISMA, our private infrastructure with end-to-end encryption. Clinical data never leaves the hospital premises. No data is sent to external servers or used to train third-party models. GDPR, MDR, and AI Act compliant by design.
Can KOI replace the clinical judgement of an anesthesiologist?
No, and it is not designed to. KOI is a clinical decision support system that reduces inter-observer variability in ASA-PS classification. The final decision always rests with the physician. KOI provides an AI-based second opinion to support, not replace, clinical judgement.
How long does it take to integrate HTX AI solutions in my hospital?
Our method delivers a working pilot in 2-4 weeks on a specific department or use case. The pilot lets you measure real value before proceeding with a broader integration. No months of planning required.
Can HTX solutions integrate with our hospital information system?
Yes. ORCA and KOI are designed with modular APIs that integrate with major hospital information systems. Integration is part of the initial pilot and is configured by our technical team.
What investment does AI adoption in a hospital require?
HTX designs solutions to be accessible even for mid-sized facilities. Costs depend on the configuration (on-premise or HTX EU cloud), number of users, and chosen modules. The initial pilot has a contained cost and allows you to evaluate ROI before any long-term commitment.