CUSTOM AI DEVELOPMENT · MEDICAL
AI-POWERED MEDICAL SOLUTIONS
Medical image analysis and data processing that support earlier, better-informed decisions.
We build bespoke AI for medical research and healthcare: deep learning for histopathological image analysis and cell segmentation, OCR and NLP that extract patient information from records, and models that support risk assessment and treatment evaluation.
Co-authored by our team: MesoGraph, Cell Reports Medicine (2023)
Deep learning for histopathological images
Small dedicated networks and genetic algorithms
THE PROBLEM
WHERE MEDICAL DATA GOES UNDER-USED
Clinical and research teams hold images and records that could support earlier decisions, if they could be analysed at scale.
LATE DIAGNOSIS
Some conditions, such as mesothelioma, are often diagnosed late.
MANUAL IMAGE REVIEW
Analysing slides and microscopic images by eye is slow and variable.
UNSTRUCTURED RECORDS
Patient information is locked in documents and free text.
TREATMENT UNCERTAINTY
Predicting how patients will respond to treatment is difficult.
WHAT WE BUILD
AI FOR MEDICAL IMAGING AND DATA
Bespoke models for research groups and healthcare organisations.
MESOTHELIOMA CELLULAR PROFILING
Deep learning models that analyse histopathological images to detect malignant mesothelioma and classify subtypes, supporting earlier diagnosis.
GOBLET CELL SEGMENTATION
Convolutional networks that segment mucus goblet cells in microscopic images, providing cell counts and morphology for research into asthma, COPD and IBD.
PATIENT DATA PROCESSING
OCR and NLP that automate the extraction of patient information from medical records, identifying key health indicators.
TREATMENT OUTCOME PREDICTION
Multimodal models that combine medical imaging and patient history to predict treatment success probabilities.
REAL APPLICATIONS
- Early detection of high-risk patients: predicting severe responses to conditions such as COVID-19
- Disease threat categorisation: assessing population health risks and treatment success probabilities
- Drug effectiveness quantification using genetic algorithms
Have a medical imaging or data problem in mind? Tell us about it.
OUR ENGINE
PROPRIETARY TECHNOLOGY, NOT A RESOLD PLATFORM
We do not resell a general-purpose platform. Three in-house technologies do the heavy lifting, each matched to the task.
NEURAL NETWORKS
Small, agile networks for classification, surrogate development, forecasting and pattern recognition.
In medicine: small dedicated networks that detect high-risk patients.
IMAGE PROCESSING
Larger convolutional networks for image classification, image segmentation and background removal.
In medicine: classification and segmentation of histopathological and microscopic images.
GENETIC ALGORITHMS
Hybrid evolutionary optimisation using multiple cross-over, mutation and evolution techniques.
In medicine: evaluating new treatments and drugs.
RESEARCH
PUBLISHED MEDICAL AI RESEARCH
Peer-reviewed publications in medical image analysis, including automatic profiling of mesothelioma subtypes from histological images.
PUBLICATIONS
Eastwood, M., Sailem, H., Marc, S.T., Gao, X., Offman, J., Karteris, E., Fernandez, A.M., Jonigk, D., Cookson, W., Moffatt, M. and Popat, S. (2023). MesoGraph: Automatic profiling of mesothelioma subtypes from histological images. Cell Reports Medicine.
Marc, S.T., Belavkin, R., Windridge, D. and Gao, X. (2023). An Evolutionary Approach to Automated Class-Specific Data Augmentation for Image Classification. International Conference on the Dynamics of Information Systems. Springer Nature Switzerland.
THE TEAM
Marc, from the Prime AI team, co-authored the 2023 papers listed here, including MesoGraph in Cell Reports Medicine.
HOW WE WORK
FROM FIRST CALL TO PRODUCTION
01
FREE CONSULTATION
Explain your challenge, or just explore how AI could streamline and enhance your current business practices. No commitment.
02
DISCOVERY & SCOPING
We review your data, systems and success criteria, then propose an approach, timeline and budget range.
03
PROOF OF CONCEPT
A focused prototype on your own data, so you can judge accuracy and value before scaling.
04
PRODUCTION & MLOps
Deployment to cloud, on-premise or edge, with monitoring, retraining and ongoing support.
FAQ
MEDICAL AI FAQ
Histopathological image analysis, such as mesothelioma profiling and goblet cell segmentation; extraction of patient information from records with OCR and NLP; patient risk assessment; and treatment outcome prediction.
Convolutional neural networks are trained on annotated images to classify and segment cells, tissue and abnormalities, giving consistent quantitative results such as cell counts and morphology.
Yes. OCR and natural language processing automate the extraction of patient information from medical records so it can be analysed at scale.
Yes. We deploy models across cloud, on-premise or edge environments, with CI/CD, version control, monitoring, drift detection and automated retraining to keep them accurate over time.
Every project is scoped individually, based on the problem, the data available and the accuracy and integration required. We usually start with a focused proof of concept on your own data, and you get an approach, timeline and budget range before any commitment.
You explain your challenge, or simply explore how AI could streamline and enhance your current business practices. We’ll tell you honestly whether AI is the right tool and, if it is, what a sensible first step looks like. There’s no commitment.
RELATED
CUSTOM AI IN RELATED INDUSTRIES
TELL US ABOUT YOUR MEDICAL AI CHALLENGE
Book a free consultation. We will tell you honestly whether AI is the right tool and, if it is, what a sensible first step looks like.