Automated Detection of Focal Cortical Dysplasias (FCD) on MRI Images with Hamburg Epilepsy Center, Protestant Hospital Alsterdorf
Client Story – Recognition of epileptic changes in the brain with Alsterdorf Evangelical Hospital in Hamburg
About the Project
Hamburg Epilepsy Center is a leading institution in Germany and Europe for comprehensive epilepsy diagnostics, specializing in the detection and treatment of epilepsy. However, the accurate detection of Focal Cortical Dysplasias (FCD) requires very in-depth knowledge about epileptology and the number of specialists worldwide is very limited, which causes many patients never get the right diagnoses.
The task remains a significant challenge due to the varied location, size, and shape of FCDs. They often blend into surrounding tissues without clear, definable boundaries, making their detection a complex task even for experienced medical professionals.
In collaboration with the experts from the Epilepsy Center Hamburg, we leveraged the power of artificial intelligence to automate the detection and recognition of FCDs on medical images. We developed a 3D convolutional neural network with autoencoder regularization specifically tailored for FCD detection and segmentation. This innovative model demonstrated higher sensitivity in detecting FCDs compared to conventional visual analyses. The experts found the developed model to be highly useful for FCD screening in clinical practice.
Detecting FCDs presents unique challenges compared to other applications of AI in radiology, such as brain tissue segmentation or tumor detection. The inherent complexity arises from the significant variations in the location, size, and shape of FCDs, as well as their tendency to blend into surrounding tissues. The manual MRI analysis performed by experts is time-consuming, requires in-depth expertise, and can be subjective, making it challenging to gather a reliable dataset for training AI models. However, through extensive technological research and the collaborative efforts of doctors at the Epilepsy Center, we successfully compiled the largest dataset of MRI images with FCDs to date and achieved state-of-the-art results in FCD detection.
- Enhanced Patient Outcomes: Early and accurate detection of FCDs enables timely interventions, potentially leading to better patient outcomes and a higher likelihood of freedom from epilepsy seizures.
- Enhanced FCD Detection: The automated FCD detection system improves the accuracy and sensitivity of identifying FCDs on MRI images, ensuring a higher detection rate and reducing the risk of undetected FCDs.
- Time Savings: By automating the detection process, medical professionals can save significant time previously spent on manual analysis, allowing them to focus on other critical aspects of patient care.
- Improved Diagnostic Precision: The AI model provides consistent and objective analysis, reducing subjectivity in FCD diagnosis and contributing to more precise treatment planning.
Solution’s Unique Features:
- 3D Convolutional Neural Network: Our customized neural network architecture, designed specifically for FCD detection and segmentation, utilizes advanced deep learning techniques to analyze MRI images in three dimensions, improving detection accuracy.
- Autoencoder Regularization: The incorporation of autoencoder regularization further enhances the robustness and generalization capabilities of the model, ensuring reliable performance across diverse datasets.
- State-of-the-Art Results: Through extensive research and collaboration, we have achieved state-of-the-art results in FCD detection, providing medical professionals with cutting-edge technology for accurate diagnosis and treatment.
- Continual Learning: Our solution gets better in time, thanks to the innovative approach of continual learning, which uses the new data to continuously improve the model and simultaneously minimize the risk of catastrophic forgetting (forgetting the already learned knowledge).
Our partnership with the Hamburg Epilepsy Center and the development of an AI-powered FCD detection system have revolutionized the way FCDs are identified on MRI images. By addressing the challenges associated with FCD recognition, we have enabled more accurate and efficient diagnoses, leading to improved patient outcomes and a higher quality of care in the field of epilepsy treatment.
The results of our work have been published in the most renowned journals on Epilepsy, including the paper below, and new publications are on the way.
How can we help you? The fields marked as required below will help us to process your request and forward it accordingly. You can expect a response within one to two business days.