Rami Eisawy

Rami Eisawy PhD Candidate Image-Based Biomedical Modeling Group (IBBM) deepc AI Platform Office: External, Werk1 Email: name.lastname@tum.de firstname@deepc.ai

Education

• 2019: M.Sc. in Biomedical Computing, TUM
• 2014: B.Sc. in Electrical Engineering, University of Waterloo, Canada

Research Interests

• Anomaly detection
• Unsupervised learning
• Segmentation

Student Projects

MA/IDP: Unsupervised/self-supervised MR Brain Anomaly Detection (Unavailable)

Given the scenario where a lack of labelled data exists, develop an model that is able to detect a range of pathologies (not limited to hyperintensities). A single imaging sequence will be selected (i.e. FLAIR). Current state of the art methods rely on autoencoder reconstruction based pipelines, however clinical success is limited here and this project would focus on supplementing and/or improving this approach.

References

• MOOD-MICCAI 2020 challenge top performer: https://arxiv.org/abs/2011.04197
• incorporating transformers with reconstruction based approaches: https://arxiv.org/abs/2102.11650
• a comparative study: https://arxiv.org/abs/2004.03271

MA/IDP: Federated learning based anomaly detection (Unavailable)

A single algorithm/model has the potential to be deployed to multiple sites either via site-specific calibration and/or domain adaptation. This provides a favourable environment for clients in terms of performance, privacy and data ownership. However, there is the possibility to leverage information across all installations in the form of distributed training. In this setting, we can consider these various site installations as edge devices. The challenge remains in producing a global (server) model that is robust and efficient in terms of its aggregation scheme, tackles the heterogeneous nature of the nodes and whether or not the difficulty in data distribution across clients can be addressed at the server level to lead to a more accurate model. The focus of the project will be the aggregation scheme. Recent projects have tackled this by simply using an average function (FedAvg), personalisation layers (FedPer) and matched averaging (FedMA) to name a few. An internal CT brain model will be provided and ideally, the (Clara Train SDK) is utilised. Previous experience in another framework that supports custom components/models would also be acceptable.

• Create a global model with multiple federated learners (clients) with the aim to improve the task of anomaly detection on CT head scans.

• Determine optimal aggregation scheme for the global model.

MA/IDP: Other projects are available and can be discussed. (Unavailable)

• localisation and/or segmentation tasks on various forms of medical data -- CT/MRI (head)brain data is generally available, however access to other anatomical data is also possible.
• (spatio)temporal divergence detection modeling (i.e. anomaly detection on: Alzheimer's disease ((ADNI)), ((EEG data))).