André Pedersen

Brattørkaia 17A

7010 Trondheim

Trøndelag, Norway

Hi! I’m a Senior Machine Learning Engineer at Sopra Steria. I have strong experience in using artificial intelligence and image analysis techniques for various medical applications, aiming to develop solutions that can assist clinicians in their daily practice. Currently, my main focus is on using generative AI and large multimodal models to solve various problems in the industry.

I was a MSc student at the Arctic University of Norway (UiT), from August 2014 to June 2019, in applied physics and mathematics, specializing in machine learning and statistics. I completed a summer internship at SINTEF Digital in 2018, which I ended up collaborating with on my Master’s thesis. From January 2019, I worked part-time at SINTEF while finishing my degree.

I started my PhD fellowship October 2019 in collaboration with the same research group as for my Master’s thesis. I was a PhD Candidate until 2023, where I published several papers on computational pathology, while I also held a position as a Research Scientist at the Medical Image Analysis group at SINTEF. I successfully defended my PhD November 2024.

In parallel to my PhD work, I have developed open, standalone software for C++ and Python, mainly using Qt5/PySide6 (e.g., FastPathology and Raidionics). I have published open command line tools (e.g., livermask), developed Python packages (e.g., gradient-accumulator, torchstain), and published articles to high-impact scientific journals related to medical image analysis and deep learning (on topics such as image classification, semantic segmentation, image-to-image registration, high-performance computing, semi-supervised learning, and natural language processing).

I have also written a book chapter and acted as a reviewer for scientific journals, such as Medical Image Analysis, Nature Scientific Reports, Frontiers in Medicine, IJCARS, QIMS, and BMC Medical Imaging. Lastly, I have (co-)supervised five Master’s students working on using deep learning for supervised/semi-supervised segmentation of 3D medical images (CT), multilabel histopathology image classification, and bronchoscopy video navigation.

news

Nov 15, 2024	Officially a Doctor! Successfully defended my PhD in Medical Technology!
Oct 2, 2024	AeroPath paper published in PLOS ONE (see here). Paper, annotated dataset, trained models, demo web app, and source code are all made openly available (see here).
May 12, 2024	Achieved a certification in Machine Learning in Production from DeepLearning.AI. See the certificate here.
Jan 23, 2024	Achieved the Microsoft Azure AI Engineer Associate Certificate. See the certificate here.
Jan 19, 2024	Achieved the Microsoft Azure Data Scientist Associate Certificate. See the certificate here.
Jan 7, 2024	Achieved a certification in Generative AI with Large Language Models from DeepLearning.AI. See the certificate here.

today's publication

Fron.Rad
Meningioma Segmentation in T1-Weighted MRI Leveraging Global Context and Attention Mechanisms

David Bouget, André Pedersen, Sayied Hosainey, Ole Solheim, and Ingerid Reinertsen

Frontiers in Radiology Sep 2021

Abs Bib PDF Code

Purpose: Meningiomas are the most common type of primary brain tumor, accounting for 30% of all brain tumors. A substantial number of these tumors are never surgically removed but rather monitored over time. Automatic and precise meningioma segmentation is, therefore, beneficial to enable reliable growth estimation and patient-specific treatment planning. Methods: In this study, we propose the inclusion of attention mechanisms on top of a U-Net architecture used as backbone: (i) Attention-gated U-Net (AGUNet) and (ii) Dual Attention U-Net (DAUNet), using a three-dimensional (3D) magnetic resonance imaging (MRI) volume as input. Attention has the potential to leverage the global context and identify features’ relationships across the entire volume. To limit spatial resolution degradation and loss of detail inherent to encoder–decoder architectures, we studied the impact of multi-scale input and deep supervision components. The proposed architectures are trainable end-to-end and each concept can be seamlessly disabled for ablation studies. Results: The validation studies were performed using a five-fold cross-validation over 600 T1-weighted MRI volumes from St. Olavs Hospital, Trondheim University Hospital, Norway. Models were evaluated based on segmentation, detection, and speed performances, and results are reported patient-wise after averaging across all folds. For the best-performing architecture, an average Dice score of 81.6% was reached for an F1-score of 95.6%. With an almost perfect precision of 98%, meningiomas smaller than 3 ml were occasionally missed hence reaching an overall recall of 93%. Conclusion: Leveraging global context from a 3D MRI volume provided the best performances, even if the native volume resolution could not be processed directly due to current GPU memory limitations. Overall, near-perfect detection was achieved for meningiomas larger than 3 ml, which is relevant for clinical use. In the future, the use of multi-scale designs and refinement networks should be further investigated. A larger number of cases with meningiomas below 3 ml might also be needed to improve the performance for the smallest tumors.
@article{bouget2021attention, abbr = {Fron.Rad}, bibtex_show = {true}, author = {Bouget, David and Pedersen, André and Hosainey, Sayied and Solheim, Ole and Reinertsen, Ingerid}, year = {2021}, month = sep, pages = {711514}, title = {Meningioma Segmentation in T1-Weighted MRI Leveraging Global Context and Attention Mechanisms}, volume = {1}, journal = {Frontiers in Radiology}, doi = {10.3389/fradi.2021.711514}, pdf = {https://www.frontiersin.org/articles/10.3389/fradi.2021.711514/full}, code = {https://github.com/dbouget/mri_brain_tumor_segmentation}, rgvalue = {6} }