ConsultancyAI in oncology
AI for oncology, built for the clinic.
We develop and implement AI for pathology, radiology, radiotherapy and precision oncology, with a focus on clinical validation, regulation and day-to-day practice in the hospital.
Where AI makes a difference in oncology.
Six domains in which we develop, validate and integrate applications. Always starting from the same question: does it help the clinician and the patient?
Pathologists assess large numbers of slides every day. AI analyses whole-slide images at cell level and highlights regions that need attention, so the specialist can assess faster and more consistently.
- Tumour detection and grading
- Mitosis counting
- Quantification of biomarkers such as Ki-67, PD-L1 and HER2
- Analysis of the tumour microenvironment
From screening to follow-up: models detect and segment lesions, measure volumes over time and support response assessment according to RECIST.
- Lesion detection and segmentation
- Volumetry and growth rate
- Radiomics for prognosis
- Automated response assessment
Contouring takes radiation oncologists a lot of time per patient. Automatic segmentation of the tumour and organs at risk speeds up planning and makes adaptive radiotherapy feasible.
- Auto-contouring of organs at risk
- Dose prediction
- Adaptive planning on daily images
- Quality assurance of treatment plans
Molecular profiling produces enormous amounts of data. AI helps interpret variants, discover biomarkers and match patients to suitable therapies and trials.
- Variant interpretation and prioritisation
- Biomarker discovery
- Prediction of treatment response
- Clinical trial matching
Language models summarise records, extract structured data from free text and prepare the tumour board, always with a reference to the source.
- Summaries for the tumour board
- Structured extraction from reports
- Staging and guideline adherence
- Source-grounded answers to questions about the record
Lead times in oncology matter for outcomes. Predictive models help plan capacity for diagnostics, operating theatres and radiotherapy.
- Demand forecasting
- Planning of scanner and radiotherapy capacity
- Early warning for breaches of waiting-time standards
- Triage support
One patient, many data streams.
The biggest gains rarely come from a single model, but from connecting sources. Imaging, pathology, genomics and clinical data each tell part of the story.
We build the data pipelines and models that bring them together into one patient view, while keeping everything traceable to its source.
- Imaging
- Pathology
- Genomics
- Clinical data
From data to everyday care.
A clinical AI model is only finished when it runs safely in the workflow and keeps performing. We support the full journey, or step in wherever you need us.
- 01
Data and governance
Pseudonymisation, consent, data access and FAIR agreements, together with the privacy officer and ethics committee.
- 02
Model development
Annotation together with specialists, and robustness across scanners, protocols and centres.
- 03
Clinical validation
Retrospective, external and prospective, measured on endpoints that matter to the clinic.
- 04
Certification
Technical documentation, risk management and a quality system for the MDR or IVDR and the EU AI Act.
- 05
Implementation
Integration into PACS, LIS and EHR, with a workflow that fits how clinicians work.
- 06
Monitoring
Drift detection, post-market surveillance and controlled retraining when practice changes.
Privacy and regulation are built into the design.
Medical AI is subject to strict frameworks. We know them from the inside and design solutions that comply from day one.
Regulation
Medical devices, in vitro diagnostics and high-risk AI systems.
- MDR (EU) 2017/745
- IVDR (EU) 2017/746
- EU AI Act
- GDPR
Standards and quality
Quality management, software life cycle and risk management for medical software.
- ISO 13485
- IEC 62304
- ISO 14971
- NEN 7510
Privacy techniques
Data stays where it belongs. Models go to the data, not the other way round.
- Federated learning
- Pseudonymisation
- On-premise inference
- Audit trails
Interoperability
Integration into the systems clinicians already use every day.
- DICOM
- HL7 FHIR
- PACS
- LIS and EHR
- Each hospital trains locally on its own data
- Only model updates go to the centre
- The updates are aggregated
- The improved model goes back
Our research: simulating tumour growth at scale.
In our research on malignant pleural mesothelioma, we simulate how a tumour grows in the pleural space. A Cellular Potts Model for the tumour cells is coupled with partial differential equations for oxygen, nutrients and cytokines, in an environment built from segmented CT scans of the lungs.
This links clinical imaging to mechanistic models of tumour growth. Because such simulations are demanding, a dynamic bounding box restricts the computation to the region around the tumour and the work is distributed across multiple cores.
- Title
- Multiscale Parallel Simulation of Malignant Pleural Mesothelioma via Adaptive Domain Partitioning – An Efficiency Analysis Study
- Authors
- Anton Dolganov, Valeria Krzhizhanovskaya, Stefano Trebeschi, Vivek M. Sheraton
- Published in
- Computational Science – ICCS 2025 Workshops, Lecture Notes in Computer Science, vol. 15911, Springer, 2025, pp. 20–32
- DOI
- 10.1007/978-3-031-97570-7_3
Working on AI in oncology?
Whether you are a hospital, a research group or a medtech company: we think along from research question to implementation.
Book a call