PhD in deep learning for biomedical images of the pelvic floor

The research team of the Pelvic Floor Clinic at the University Hospitals Leuven, in collaboration with the School of Biomedical Engineering and Imaging Sciences, at King’s College London, develops and applies innovative machine-learning techniques to aid transperineal pelvic floor ultrasound examination (TPUS) and, to a lesser extent, pelvic floor Magnetic Resonance Imaging analysis. PFUS is used to assess pelvic floor anatomy and function in women with pelvic floor dysfunction. We aim to develop and validate algorithms that automate aspects of the clinical workflow to save clinicians’ time for patient counselling and to be less operator- and expertise-dependent. This computational research aims to have a clinical impact and improve the current clinical workflow by lowering the threshold for using ultrasound to less experienced users.

Pelvic Floor Ultrasound is a tool routinely used at the Pelvic Floor Clinic of the University Hospitals Leuven. The group has research protocols in place and patients are consented for use of their images. Therefore, a large-scale annotated data set is available for this type of research. Also, three clinician-researchers in the team are engaged into this project. The group collaborates with GE Healthcare, as they are currently the market leader in TPUS, and we aim to have our achievements eventually implemented in clinical practice.

The University of Leuven is one of Europe’s leading research universities, with English as the working language for research. It ranks with position 42 among the top 50 universities worldwide in the Times Higher Education World University Rankings 2022. Leuven is a pleasant middle-sized city, with plenty of activities for its more than 60,000 students. It lies close to Brussels, at the centre of Europe.

Responsibilities

We are looking for one or more talented individual(s) to strengthen our team by developing novel deep learning-based tools that can benefit and automate the current clinical workflow.

We have already a functional automatic pipeline that segments the area of the levator hiatus (which may be considered as a bio-marker) from a transperineal ultrasound volume, and extracts the imaging sequence used to analyse the anal sphincter. There is also an interactive segmentation algorithm that gives the clinician control of segmentation boundaries. Also, we have a large, annotated data set of images and volumes in healthy controls and symptomatic patients, and further collection is permanently ongoing.

 The research goals of this position will be to develop an efficient, intuitive, user-friendly, interactive algorithm for clinical users to adjust further elements of the current pipeline, in order to edit the extracted biomarker as desired. In addition, an intelligent pipeline will be developed to aid the clinicians in interpretation of trauma to the levator hiatus (by quantify ballooning and avulsion) as well as to the anal sphincter (i.e., computer aided diagnosis – CAD). The pipeline must by construction provide insights into why a particular decision regarding anatomical status was made, to ensure the output is understandable and interpretable to clinicians. In the future we aim to have a pipeline in place in the clinic for instantaneous analysis of anal sphincter injuries immediately after delivery.

You will have the opportunity to collaborate with machine learning researchers from academia and industry, and you may see your research applied in clinic with a large industrial partner. The research is in collaboration with clinicians and clinician researchers at the UZ Leuven hospital, who specialise in PFUS.

Responsibilities

• Develop and implement state-of-the-art machine-learning techniques to a real-world problem with clinical impact
• Publishing in top machine learning journals and presenting at medical conferences
• Regular collaboration with researchers from external partners from academia and industry

Opportunities

• Learn about and contribute to the design of cutting-edge applications of machine learning and AI (Artificial Intelligence) in the medical imaging analysis field
• Develop skills in machine learning programming such as PyTorch, scikit-learn and JAX.
• Work in a creative, diverse, and inspiring academic environment with direct clinical impact

Essential

• Master’s Degree in computer science or related (STEM) scientific or engineering degree (i.e., physics, chemistry, mathematics, chemical engineering, mechanical engineering, electrical engineering)
• Strong analytical thinking, strong scientific writing and presentation skills in English and a strong mathematical capability
• Experience with a programming language (i.e., python, C, C+, C#)

Desirable

Strong interest in machine and deep learning, medical imaging analysis and applications of AI within healthcare.

Interested in our achievements?

Recording of mini state-of-the-art lecture “Artificial Intelligence in Image Analysis” at the joint meeting of the American Urogynaecological Society and International Urogynecologic Association, June 18^th, 2022.

Relevant publications (by researchers of the group)

• Bonmati E., Hu Y., Sindhwani N., Dietz H., D’hooge J., Barratt D., Deprest J., Vercauteren T., (2018) Automatic segmentation method of pelvic floor levator hiatus in ultrasound using a self-normalizing neural network. J. Med. Imag. 5(2) 021206. https://doi.org/10.1117/1.JMI.5.2.021206
•   Cattani L., Van Schoubroeck D., Housmans S. et al. (2020) Exo-anal imaging of the anal sphincter: a comparison between introital and transperineal image acquisition. Int Urogynecol J 31, 1107–1113. https://doi.org/10.1007/s00192-019-04122-5
• Williams H., Cattani L., Li W., Tabassian M., Vercauteren T., Deprest J., D’hooge J. (2019) 3D Convolutional Neural Network for Segmentation of the Urethra in Volumetric Ultrasound of the Pelvic Floor. 2019 IEEE International Ultrasonics Symposium (IUS), Glasgow, United Kingdom, pp. 1473-1476. https://doi.org/10.1109/ULTSYM.2019.8925792
• Williams H., Cattani L., Vercauteren T., Deprest J., D’hooge J. (2021) Automatic Tomographic Ultrasound Imaging Sequence Extraction of the Anal Sphincter. Simplifying Medical Ultrasound. ASMUS 2021. Lecture Notes in Computer Science, vol 12967. Springer, Cham. https://doi.org/10.1007/978-3-030-87583-1_4​
• Williams H., Pedrosa J., Cattani L., Housmans S., Vercauteren T., Deprest J., D’hooge J. (2021) Interactive Segmentation via Deep Learning and B-Spline Explicit Active Surfaces. In: de Bruijne M. et al. (eds) Medical Image Computing and Computer Assisted Intervention – MICCAI 2021. Lecture Notes in Computer Science, vol 12901. Springer, Cham. https://doi.org/10.1007/978-3-030-87193-2_30
•  Williams H., Cattani L., Van Schoubroeck D., Yaqub M., Sudre C., Vercauteren T., D'Hooge J., Deprest J. (2021) Automatic Extraction of Hiatal Dimensions in 3-D Transperineal Pelvic Ultrasound Recordings, Ultrasound in Medicine & Biology, ISSN 0301-5629, https://doi.org/10.1016/j.ultrasmedbio.2021.08.009.

Mail your application to:

Prof. Jan Deprest, email: [Email Address Removed]

Prof. Tom Vercauteren email: [Email Address Removed]

Prof Jan D’hooge email: [Email Address Removed]