Automated Retinal Vessel Landmark Detection in OCT Time Series Data using Deep Learning

The objective of this master thesis is to explore and implement deep learning models to automatically detect landmarks in dynamic OCT data. The student will focus on developing a model that accurately identifies and tracks anatomical landmarks such as vessel centres. The task can be seen as a landmark detection of segmentation task by leveraging recent advances in deep learning and time series data analysis.

  • Background

    Optical Coherence Tomography (OCT) is a widely used imaging modality in ophthalmology to capture detailed depth-resolved images of the retina. A new methodology allows to capture these OCT images dynamically, allowing for the analysis of flow velocities inside vessels. Currently, the vessels are annotated by eye doctors. This time-consuming process holds a vast, clinically relevant potential for automation.

  • Scope of Work

    1. Literature Review: Review recent advancements in deep learning, with a particular focus on image segmentation, landmark detection and time series data analysis, and their applications in medical imaging.

    2. Data Preparation: Work with an existing dataset of dynamic OCT sequences. Preprocess the data to ensure compatibility with modern deep learning models, including temporal alignment and normalization.

    3. Model Development: Implement and fine-tune state-of-the-art deep learning models (e.g. Convolutional Neural Networks, Vision Transformers, …) to detect and track landmarks in the dynamic OCT data. Explore architectures that effectively combine temporal and spatial information.

    4. Model Training and Evaluation: Train the models on the existing dataset and evaluate their performance using metrics such similarity metrics such as the DICE score, or distance metrics such as the Hausdorff distance.

    5. Optimization: Experiment with different model architectures, loss functions, and training strategies to optimize the performance.

    6. Validation: Validate the model’s performance on unseen data and obtain feedback from medical professionals to assess the clinical usability of the detected landmarks.

    7. Documentation and Presentation: Document the entire process, including challenges, solutions, and results. Prepare a final thesis report and thesis presentation.

  • Expected Outcome

    The project aims to develop an accurate and reliable AI model for the automatic detection of vessel centres in dynamic OCT data. The student would ideally demonstrate improvements over previous models, particularly in the temporal accuracy and clinical relevance of the detected landmarks.

  • Further Information

    • 40% theory, 60% implementation
    • Master’s thesis, 1 student
    • Prior knowledge recommended in
      • image analysis
      • deep learning
      • and/or time series data analysis

Haben wir dein Interesse geweckt?

Ich bin an der Studienarbeit Automated Retinal Vessel Landmark Detection in OCT Time Series Data using Deep Learning interessiert und möchte mehr erfahren.