Interpretable and transparent machine learning algorithms for medical image computing at University of Bath on FindAPhD.com

This project is no longer listed on FindAPhD.com and may not be available.

Click here to search FindAPhD.com for PhD studentship opportunities

Dr Mohammad Golbabaee, Dr Matthias Ehrhardt Applications accepted all year round Competition Funded PhD Project (UK Students Only)

Bath United Kingdom Applied Mathematics Artificial Intelligence Biomedical Engineering Computational Mathematics Computer Vision Data Analysis Machine Learning Medical Physics Software Engineering Stochastic Processes

About the Project

The UKRI Centre for Doctoral Training in Accountable, Responsible and Transparent AI (ART-AI) at the University of Bath is inviting applications for a PhD project based in the Department of Computer Science and under the joint supervision of Dr Mohammad Golbabaee (Computer Science dept.) and Dr Matthias Ehrhardt (Mathematics dept.). The position is an opportunity to conduct cutting-edge research at the intersection of computational medical imaging, machine learning and mathematical analysis, with the possibility of collaboration with world leading healthcare industries.

This project is concerned with the development of interpretable and transparent machine/deep learning algorithms for image reconstruction in medical imaging.

Image reconstruction is a computational process that deals with the formation of an image from abstract data obtained from sensory measurements. An example where these techniques find an application is Magnetic Resonance Imaging (MRI) which is omnipresent in modern medicine with applications ranging from neurology to oncology. Challenges such as missing/incomplete data, noise and artefacts are inherent to imaging applications and are dealt with by computational algorithms e.g., for image super-resolution or compressed sensing [1-3]. While AI in the form of deep learning algorithms produces state-of-the-art results, the opaque nature of deep neural networks has made it difficult to understand and interpret the results [4]. Addressing these factors are crucial for medical imaging applications and for enabling the clinical up take of these methodologies.

In this project you will bridge this gap by developing new algorithms and analysis techniques for computational medical imaging towards transparent and reliable solutions to the aforementioned problems. You are expected to further advance current developments within this theme including (but not limited to) reliable generative models to capture prior knowledge about data and their transparent and inspectable integration within model-based (possibly optimisation-inspired) image reconstruction algorithms [5-9]. The project outcomes may involve research papers in which the developed methodologies are backed with theoretical analysis and guarantees for the predicted outcomes, and/or involve software solutions that enable inspection by visualising, analysing and explaining predictions made by the deep learning model.

Applicants should hold, or expect to receive, a First or Upper Second Class Honours degree in Computing, Mathematics, Engineering or a related subject. A master’s level qualification would be advantageous. A strong background in mathematics is essential. Good programming skill and experience of machine learning are highly desirable.

Formal applications should be accompanied by a research proposal and made via the University of Bath’s online application form. Enquiries about the application process should be sent to [Email Address Removed].

Start date: 3 October 2022.

Funding Notes

ART-AI CDT studentships are available on a competition basis and applicants are advised to apply early as offers are made from January onwards. Funding will cover tuition fees and maintenance at the UKRI doctoral stipend rate (£16,062 per annum in 2022/23, increased annually in line with the GDP deflator) for up to 4 years.
We also welcome applications from candidates who can source their own funding.

References

[1] Lustig, Michael, et al. "Compressed sensing MRI." IEEE signal processing magazine 25.2 (2008): 72-82.
[2] Yang, Guang, et al. "DAGAN: Deep de-aliasing generative adversarial networks for fast compressed sensing MRI reconstruction." IEEE transactions on medical imaging 37.6 (2017): 1310-1321.
[3] Chen, Yuhua, et al. "Efficient and accurate MRI super-resolution using a generative adversarial network and 3D multi-level densely connected network." International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2018.
[4] See an interesting blog on explainable AI and references thereof by Oge Marques: https://www.datasciencecentral.com/profiles/blogs/explainable-ai-for-medical-images
[5] Bora, Ashish, et al. "Compressed sensing using generative models." International Conference on Machine Learning. PMLR, 2017.
[6] Monga, Vishal, Yuelong Li, and Yonina C. Eldar. "Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing." IEEE Signal Processing Magazine 38.2 (2021): 18-44.
[7] Aggarwal, Hemant K., Merry P. Mani, and Mathews Jacob. "MoDL: Model-based deep learning architecture for inverse problems." IEEE transactions on medical imaging 38.2 (2018): 394-405.
[8] Ryu, Ernest, et al. "Plug-and-play methods provably converge with properly trained denoisers." International Conference on Machine Learning. PMLR, 2019.
[9] Ahmad, Rizwan, et al. "Plug-and-play methods for magnetic resonance imaging: Using denoisers for image recovery." IEEE signal processing magazine 37.1 (2020): 105-116.