Computational Imaging

Available Research Topics

Combining Pthychography and Optical Coherence Tomography

Optical coherence tomography (OCT) is a non-invasive imaging technique, that uses interference and coherence properties of light to create 3D-images of tissue, with micrometer resolution. It is a phase sensitive technique and essentially, the phase information of the light at multiple wavelengths is providing the exact depth information.

(Fourier) Ptychography [1] on the other hand is also a phase sensitive imaging technique. It acquires the phase by several in a certain way overlapping but independent measurements. Some extensions have also suggested more 3D and tomographic reconstructions [2, 3].

In this project we want to explore the discrepencies and similarities between OCT and Fourier [1] or Near-field Ptychography [4] to improve upon both imaging techniques. One direction will be to explore in how far ptychographic techniques could explore a reference-free non-interferometric equivalent to OCT including its advantages and disadvantages.

Another direction would be how an approach related to Ptychography can solve the complex-conjugate ambiguity in OCT, since the phase information will be overdetermined and thereby enable full-range OCT.

This work would be the continuation of a previous Master's project.

Your tasks and milestones

• Design a novel implementation of a full-field OCT system with diffuser-varied illumination.
• Simulate the scattering behavior of this new system
• Build new diffuser full-field OCT system
• Data collection
• Develop effective reconstruction algorithms for collected data
• Evaluate speckle reduction performance.
• Discuss findings at weekly (group) meetings..
• Present thesis results at section meeting..
• Write a final report.

References

G. Zheng, C. Shen, S. Jiang, et al. Concept, implementations and applications of Fourier ptychography. Nat. Rev. Phys. 3, 207–223, 2021, doi.org/10.1038/s42254-021-00280-y

Optical Coherence Tomography Simulation and Deep-Learning Based Reconstruction

Modern imaging techniques increasingly rely on advanced image reconstruction methods to extract meaningful information from raw data. These approaches combine physical models with signal processing techniques and image reconstruction constraints to achieve high-quality results. Examples include digital holography, where diffraction plays a key role, and optical coherence tomography (OCT), which heavily utilizes Fourier transforms grounded in a physical framework.

Recently, deep learning has emerged as a powerful tool for image reconstruction, offering novel ways to enhance imaging quality and efficiency. One such example is GedankenNet [1], which integrates physics-based models with deep learning to improve holographic image reconstruction. Inspired by this, our goal is to develop a similar deep-learning-driven reconstruction method tailored for holographic optical coherence tomography (OCT) [2]. This project will explore how deep learning can enhance OCT imaging quality compared to traditional reconstruction techniques.

Your task will begin with the development of a forward physical model that accurately simulates the raw data generated by a holographic OCT setup. This model will serve as the foundation for training a deep-learning-based inverse model capable of reconstructing and revealing the underlying three-dimensional image data. By leveraging large-scale simulated datasets, you will train a neural network to learn the inverse mapping from raw OCT data to high-quality images. Once trained, the model will be tested on actual OCT raw data and rigorously compared against conventional reconstruction techniques to assess its effectiveness and potential advantages.

This project presents an exciting opportunity to work at the intersection of physics, computational imaging, and artificial intelligence. By combining theoretical insights with cutting-edge deep learning techniques, you will contribute to the advancement of modern imaging methods, pushing the boundaries of what is possible in holographic OCT.

Your tasks and milestones

• Familiarize yourself with GedankenNet and the holographic OCT technique.
• Create a forward model, i.e., simulate the holographic OCT raw data
• Partially reproduce the results of GedankenNet
• Extend the results to holographic OCT
• Discuss findings at weekly (group) meetings
• Present thesis results at section meetings.
• Write a final report.

References

[1] Huang, L., Chen, H., Liu, T. et al. Self-supervised learning of hologram reconstruction using physics consistency. Nat. Mach. Intell. 5, 895–907 (2023). doi.org/10.1038/s42256-023-00704-7

Enhanced Imaging Depth with Dispersion-Encoded Full-Range holographic OCT

Optical coherence tomography (OCT) is a powerful imaging technique used to capture detailed cross-sectional images of transparent and semi-transparent materials, such as biological tissues. It works by measuring how light reflects from different depths within a sample, much like ultrasound but using light waves instead of sound. However, standard OCT systems have a limitation—they only capture part of the depth information, leading to unwanted mirror images (artifacts) in the final reconstruction. This stems from a symmetry in the setup and acquisition that only allows distinguishing path length differences (without its sign) to a reference arm.

A promising solution to this issue is dispersion-encoded full-range OCT [1], which uses carefully controlled changes in how different colors of light propagate (dispersion) to separate overlapping signals and provide a complete depth image. In a sense, this breaks the symmetry in the acquired data. This approach has been studied in conventional OCT, but its implementation in holographic OCT remains an open challenge. In addition, we want to explore new algorithmic approaches to obtain these results.

In this Master's project, you will work on integrating dispersion-encoded full-range OCT into a holographic OCT system [2]. Your first task will be to modify an optical setup and introduce dispersion effects, allowing for breaking the symmetry without requiring complex additional hardware. Once this optical system is in place, you will develop computational techniques to process the acquired data, extracting high-quality depth information through signal processing and numerical reconstruction methods.

To further enhance the imaging quality, you will explore deep-learning-based approaches that can improve and accelerate the reconstruction process. The final step will be to test the developed methods on experimental data and compare their performance to traditional OCT techniques.

This project is a unique opportunity to gain experience in optics, computational imaging, and machine learning while contributing to cutting-edge advancements in high-resolution imaging for biomedical, industrial, and scientific applications.

Your tasks and milestones

• Review and implement existing literature on dispersion-encoded full-range OCT
• Simulate the effects of dispersion mismatch on OCT signals
• Implement and test numerical reconstruction algorithms for phase retrieval and artifact suppression
• Develop a dispersion encoding strategy for holographic OCT to achieve full-range imaging
• Compare the developed method to existing full-range OCT techniques
• Validate the technique on both simulated and experimental OCT datasets
• Discuss findings at weekly (group) meetings
• Present thesis results at a section meeting
• Write a final report

References

[1] F. Köttig, P. Cimalla, M. Gärtner, and E. Koch, "An advanced algorithm for dispersion encoded full range frequency domain optical coherence tomography," Opt. Express 20, 24925-24948 (2012). doi.org/10.1364/OE.20.024925

Your tasks and milestones

• Familiarize yourself with GedankenNet and the holographic OCT technique.
• Create a forward model, i.e., simulate the holographic OCT raw data
• Partially reproduce the results of GedankenNet
• Extend the results to holographic OCT
• Discuss findings at weekly (group) meetings
• Present thesis results at section meetings.
• Write a final report.

References

[1] Huang, L., Chen, H., Liu, T. et al. Self-supervised learning of hologram reconstruction using physics consistency. Nat. Mach. Intell. 5, 895–907 (2023). doi.org/10.1038/s42256-023-00704-7

Deep-Learning-based Aberration and Motion Correction for Holographic OCT

Optical coherence tomography (OCT) is a powerful imaging technique used to capture detailed cross-sectional images of transparent and semi-transparent materials, such as biological tissues. It works by measuring how light reflects from different depths within a sample, much like ultrasound but using light waves instead of sound. However, real-world OCT imaging faces several challenges, including motion artifacts [3], dispersion-induced distortions [3], and optical aberrations [1,2] that degrade image quality.

These distortions arise from comparable mathematical problems [1], making it possible to address them using a unified computational approach. Recent advancements in deep learning have shown great potential in solving complex inverse problems, making it a promising candidate for enhancing OCT imaging by correcting dispersion, motion artifacts, and optical aberrations.

In this Master's project, you will develop a deep-learning-based correction framework for holographic OCT. Your first step will be to simulate and analyze the effects of dispersion, motion, and optical aberrations in OCT data. Based on this, you will design and train a neural network capable of identifying and correcting these distortions, ensuring high-fidelity image reconstruction. The network will be trained using synthetic data with known distortions and then validated on experimental holographic OCT data.

Once the deep-learning model is developed, you will compare its performance against traditional correction methods, evaluating improvements in resolution, contrast, and imaging robustness. Additionally, you will explore how the learned corrections generalize across different imaging conditions and datasets.

This project offers an exciting opportunity to work at the intersection of optics, computational imaging, and artificial intelligence. By developing a unified deep-learning approach to correct dispersion, motion, and aberrations, you will contribute to advancing high-resolution imaging techniques with applications in biomedical research, industrial inspection, and scientific imaging.

Your tasks and milestones

• Simulate the effects of dispersion, motion artifacts, and optical aberrations in OCT imaging
• Develop a deep-learning-based correction algorithm to address these distortions
• Train the deep-learning model using synthetic and experimental OCT data
• Validate the correction framework by applying it to real holographic OCT datasets
• Analyze and compare the results against conventional correction techniques
• Optimize the model for computational efficiency and real-time application
• Discuss findings at weekly (group) meetings
• Present thesis results at a section meeting
• Write a final report

References

[1] D. Hillmann, H. Spahr, C. Hain, et al., "Aberration-free volumetric high-speed imaging of in vivo retina," Sci. Rep. 6, 35209 (2016). doi.org/10.1038/srep35209

[2] D. Hillmann, C. Pfäffle, H. Spahr, S. Burhan, L. Kutzner, F. Hilge & G. Hüttmann, "Computational adaptive optics for optical coherence tomography using multiple randomized subaperture correlations," Opt. Lett. 44, 3905-3908 (2019). doi.org/10.1364/OL.44.003905

[3] D. Hillmann, T. Bonin, C. Lührs, G. Franke, M. Hagen-Eggert, P. Koch & G. Hüttmann, "Common approach for compensation of axial motion artifacts in swept-source OCT and dispersion in Fourier-domain OCT," Opt. Express 20, 6761-6776 (2012). doi.org/10.1364/OE.20.006761

Determining the Refractive Index from Oblique OCT Data

Optical coherence tomography (OCT) is a widely used imaging technique that provides high-resolution cross-sectional images of biological tissues and other transparent materials. Unlike ultrasound, which uses sound waves, OCT employs light waves to probe structures beneath the surface. One of the key challenges in OCT is determining the refractive index of a specimen, which is essential for accurate depth measurements and material characterization.

Inspired by the CUTE (Computed Ultrasound Tomography in Echo mode) technique in ultrasound imaging [1,2], this project aims to develop a comparable method for OCT. Our unique OCT setup [3] allows for illumination from multiple directions, enabling the extraction of additional information about the specimen’s optical properties. By leveraging this multi-directional illumination, we aim to reconstruct both the structure and the refractive index distribution of the sample.

In this Master's project, you will work on developing an imaging and reconstruction framework for refractive index mapping in holographic OCT. Your first task will be to simulate how light propagates through samples with varying refractive indices under different illumination angles. Using this information, you will design and implement an inverse reconstruction algorithm to extract the refractive index distribution from the recorded holographic OCT data.

Once developed, the method will be validated on both simulated and experimental OCT datasets. You will compare its performance with conventional OCT techniques and assess its potential for improving quantitative imaging in biomedical and material science applications.

This project offers a unique opportunity to explore cutting-edge developments in computational imaging, optics, and artificial intelligence. By extending holographic OCT to include refractive index mapping, you will contribute to advancing imaging methodologies with broad applications in biomedical diagnostics, materials research, and optical metrology.

Your tasks and milestones

Develop a simulation model for light propagation through samples with varying refractive indices and different illumination angles
Design and implement an inverse reconstruction algorithm for refractive index mapping
Validate the method using simulated and experimental OCT datasets
Analyze and evaluate the performance compared to conventional OCT techniques
Discuss findings at weekly (group) meetings
Present thesis results at a section meeting
Write a final report

References

[1] M. Jaeger, G. Held, S. Peeters, S. Preisser, M. Grünig, M. Frenz, "Computed Ultrasound Tomography in Echo Mode for Imaging Speed of Sound Using Pulse-Echo Sonography: Proof of Principle," Ultrasound Med. Biol. 44(1) (2015). doi.org/10.1016/j.ultrasmedbio.2014.05.019

[2] P. Stähli, C. Becchetti, N. Korta Martiartu, et al., "First-in-human diagnostic study of hepatic steatosis with computed ultrasound tomography in echo mode," Commun. Med. 3, 176 (2023). doi.org/10.1038/s43856-023-00409-3

[3] D. Hillmann, H. Spahr, C. Hain, et al., "Aberration-free volumetric high-speed imaging of in vivo retina," Sci. Rep. 6, 35209 (2016). doi.org/10.1038/srep35209