Clinically-grounded development
Together with our in-house specialist physicians, we build AI that is genuinely useful in real clinical practice.
Medical AI Custom Development
A medical AI custom-development service — from problem definition and PoC through model development, validation, and operation. From diagnostic support to prognosis prediction and risk stratification, we build the right solution for your challenge, choosing from a wide range of machine-learning methods including deep learning.
What is medical AI custom development?
Medical AI custom development is a service in which a specialist company designs, builds, validates, and operates medical AI — models and systems for medical-image diagnosis, prognosis prediction, risk stratification, and more — tailored to the needs of hospitals, pharmaceutical companies, and research institutions. Building on the expertise gained developing medical-imaging diagnostic AI (SaMD), Kewel selects the best approach from a wide range of machine-learning methods, including deep learning, according to the problem and the data — and supports you end to end, from planning and development through implementation and operation.
What Kewel takes on
- Medical image analysis — lesion detection and segmentation, disease classification, findings presentation, benign/malignant classification, and more (CT, MRI, X-ray, OCT, fundus, etc.)
- Prognosis prediction and risk stratification using diverse clinical data such as lab values, images, and time-series
- Data design, annotation, and dataset construction, carried out under the supervision of our in-house specialist physicians
- Explainable AI (XAI) implementation and SaMD-aware design (regulatory affairs handled together with a regulatory CRO)
- End-to-end support from PoC design through clinical evaluation, operation, and continuous accuracy improvement
Challenges we address
- Automatically detect or segment specific lesions/findings from medical images
- Predict post-operative prognosis or stratify risk
- Turn research-stage results into AI usable in clinical practice
- Delegate AI model development and evaluation as part of clinical or academic research
- Delegate everything from data design and annotation to dataset construction
- Develop with explainability (XAI) and regulatory / SaMD readiness in mind
- Extract quantitative biomarkers for pharmaceutical R&D
Development Process
- 01
Problem definition & PoC design
We clarify the clinical and business problem and design a proof of concept to test feasibility.
- 02
Data design & annotation
We design the required data, annotate it, and augment it as needed — under the supervision of our in-house specialist physicians in our own fields of expertise.
- 03
Model development & validation
We develop models with state-of-the-art methods and validate them, visualizing the rationale with explainable AI (XAI).
- 04
Clinical evaluation & SaMD readiness
We support clinical evaluation and, where applicable, design toward approval as a medical device (SaMD). Regulatory affairs themselves are handled in collaboration with a regulatory CRO.
- 05
Operation & improvement
We stay involved through deployment, monitoring, and continuous accuracy improvement.
Why Kewel
Method chosen for the challenge
From a wide range of machine-learning methods including deep learning, we pick the best fit for the problem and the data — balancing accuracy and explainability.
Experience across domains
Proven development across ophthalmology, gastroenterology, orthopedics, neurology and more — spanning CT, MRI, X-ray, OCT, fundus cameras, and other modalities.
Explainable AI (XAI) & SaMD
We visualize prediction rationale and design toward medical-device (SaMD) requirements. Regulatory affairs are handled together with a regulatory CRO.
NVIDIA Inception member
Selected for NVIDIA Inception, with access to GPU and technical support.
Proprietary technology & patents
We hold proprietary technology, including novel models and patents in medical-image AI.
Track Record & Partnerships
We develop our own medical AI across multiple domains including ophthalmology, gastroenterology, orthopedics, and neurology. Alongside this, we are building contracts and partnerships with medical institutions across these fields — from clinics and community hospitals to university hospitals. We are also a member of the NVIDIA Inception program.
Custom Development by Clinical Domain
We take on custom development tailored to the modalities and diseases of each domain — ophthalmology, gastroenterology, orthopedics, and neurology. See the details for each field.
Frequently Asked Questions
- What is medical AI custom development?
- It is a service in which a specialist company designs, builds, validates, and operates medical AI — models and systems for image diagnosis, prognosis prediction, risk stratification, and more — tailored to the needs of hospitals, pharmaceutical companies, and research institutions. Kewel selects the best approach from a wide range of machine-learning methods, including deep learning, based on the problem and the data.
- Which clinical domains and modalities do you support?
- We have implementation experience across ophthalmology, gastroenterology, orthopedics, and neurology, and handle modalities such as CT, MRI, X-ray, OCT, and fundus cameras. Beyond images, we also work with numeric and time-series clinical data.
- Can you develop with non-image data too?
- Yes. We work with a wide range of non-image data — lab values, biosignals, text, post-operative course, and more. Depending on the problem, we address classification, regression, anomaly detection, prognosis prediction, risk stratification, and beyond, choosing between deep learning and classical machine learning based on the nature of the data.
- How does the process work?
- Problem definition & PoC design → data design & annotation → model development & validation (with XAI to visualize rationale) → clinical evaluation & SaMD readiness → operation & improvement. We recommend starting with a PoC to test feasibility.
- Can you develop even with limited data?
- We support data design, annotation, and augmentation. Even with limited data, we can proceed by first validating feasibility through a PoC.
- Can you support regulatory / SaMD?
- We handle design and development with explainability and SaMD readiness in mind. For regulatory affairs, we currently work with a regulatory CRO (contract research organization) while also building up our own in-house capability. Projects aimed at medical-device (SaMD) certification are welcome.
- Can we just consult first? What about cost?
- You can consult us from the problem-scoping stage. Cost depends on the target data, goal, and scope, so please get in touch first.
Looking for a medical AI development partner?
We can start from problem scoping and a PoC. Please feel free to get in touch.