Case Studies / Evangelisches Krankenhaus Alsterdorf
HEALTHCARE · MEDICAL IMAGING & NEUROLOGY
Automated epilepsy lesion detection on MRI with AI at the Hamburg Epilepsy Center
Focal Cortical Dysplasias (FCDs) – a leading cause of epilepsy – are notoriously difficult to spot on MRI scans. Only a handful of specialists worldwide can reliably detect them, leaving many patients undiagnosed. We built a 3D neural network that automates FCD detection with higher sensitivity than conventional visual analysis. The results were published in leading epilepsy research journals.
Client: Hamburg Epilepsy Center, Evangelisches Krankenhaus Alsterdorf – a leading German institution for epilepsy diagnostics and treatment.
KEY RESULTS
SOTA
State-of-the-art results in automated FCD detection
3D CNN
Custom convolutional neural network for MRI analysis
Published
Results peer-reviewed in leading epilepsy journals
Largest
Largest FCD MRI dataset compiled to date for training
INDUSTRY
Healthcare / Neurology
USE CASE
Epilepsy lesion detection on MRI
AI APPROACH
3D CNN + autoencoder regularization
DATA
MRI brain scans (largest FCD dataset)
The challenge
Focal Cortical Dysplasias are malformations in the brain’s cortex and one of the most common causes of drug-resistant epilepsy. When detected and surgically removed, many patients can become seizure-free. The problem: FCDs are extremely difficult to identify on MRI scans.
Unlike tumors, which typically present as clearly visible masses, FCDs vary widely in location, size, and shape. They blend into surrounding brain tissue without clear boundaries. Detecting them requires deep, specialized expertise in epileptology, and the number of specialists worldwide who can reliably identify them is very small. The result: many patients with treatable epilepsy never receive the right diagnosis.
The core problem: a critical diagnostic step – identifying FCDs on MRI – depended on a tiny pool of global specialists. Manual MRI analysis was time-consuming, subjective, and inaccessible to most hospitals. Patients were being missed.
What we built
Working directly with the neurologists and epileptologists at the Hamburg Epilepsy Center, we developed an AI system that automates the detection and segmentation of FCDs in 3D MRI brain scans.
Custom 3D convolutional neural network. We designed a neural network architecture specifically for this task, analyzing MRI images in three dimensions to detect the subtle structural anomalies that characterize FCDs. The architecture included autoencoder regularization to improve generalization and robustness across diverse patient scans.
Largest FCD training dataset. One of the biggest challenges in medical AI is data scarcity. Through extensive collaboration with the Epilepsy Center, we compiled the largest dataset of MRI images with confirmed FCD diagnoses assembled to date. This was essential for training a model that could perform reliably beyond the specific cases it learned from.
Higher sensitivity than visual analysis. The resulting model demonstrated higher sensitivity in detecting FCDs compared to conventional expert visual analysis of MRI scans. The clinical team assessed the model as highly useful for FCD screening in practice.
Continual learning. The system was designed with a continual learning approach, improving over time as new data becomes available while minimizing the risk of catastrophic forgetting (losing previously learned knowledge when training on new cases).
The results
BEFORE
FCD detection depended on a tiny number of global specialists. Manual MRI analysis was slow, subjective, and unavailable at most hospitals. Many patients went undiagnosed
AFTER
Automated screening tool with state-of-the-art detection accuracy. Higher sensitivity than conventional visual analysis. Peer-reviewed and validated for clinical use.
The model achieved state-of-the-art results in FCD detection and was validated in clinical practice at the Hamburg Epilepsy Center. The findings were published in leading epilepsy research journals, contributing to the broader medical community’s understanding of how AI can support neurological diagnostics.
Beyond the immediate clinical impact, this project demonstrated theBlue.ai’s capability in deep learning for medical imaging, building custom architectures for problems where off-the-shelf solutions don’t exist, assembling the data infrastructure to train them, and delivering results that meet the rigorous standards of peer-reviewed clinical research.
Technology used
More Case Studies
See how we’ve helped other companies
AUTOMOTIVE · LEADING LUXURY MANUFACTURER
Intelligent virtual assistant replacing manual planning queries across SAP and cloud systems
Product planners spent hours manually querying SAP BW and multiple data warehouses for every decision. We built a bilingual voice-and-text assistant that retrieves planning data on demand – no system expertise needed.
Hrs → Sec
Data retrieval
DE + EN
Voice & text
SAP BW
Integrated
MANUFACTURING · RADAWAY
Making email-based order processing reliable with LLMs
Staff were manually reading customer emails, identifying products, and entering orders by hand. We turned a promising AI prototype into a production system that handles it end to end, across languages, formats, and attachments.
-90%
Manual intervention
95%+
Match accuracy
LOGISTICS · FR. MEYER’S SOHN
Eliminating manual data extraction from thousands of daily shipping emails
Operations staff were manually reading German and English logistics emails to pull out routing and scheduling data, every single day. We built an AI pipeline that extracts, structures, and delivers the data automatically.
–80%
Manual effort
2 langs
DE & EN
On-prem
Deployed