Timezone: »

Medical Imaging meets NIPS
Ender Konukoglu · Ben Glocker · Hervé Lombaert · Marleen de Bruijne

Sat Dec 08 05:00 AM -- 03:30 PM (PST) @ Room 513 ABC
Event URL: https://sites.google.com/view/med-nips-2018 »

Medical imaging and radiology are facing a major crisis with an ever-increasing complexity and volume of data and immense economic pressure. With the current advances in imaging technologies and their widespread use, interpretation of medical images pushes human abilities to the limit with the risk of missing critical patterns of disease. Machine learning has emerged as a key technology for developing novel tools in computer aided diagnosis, therapy and intervention. Still, progress is slow compared to other fields of visual recognition, which is mainly due to the domain complexity and constraints in clinical applications, i.e. robustness, high accuracy and reliability.

“Medical Imaging meets NIPS” aims to bring researchers together from the medical imaging and machine learning communities to discuss the major challenges in the field and opportunities for research and novel applications. The proposed event will be the continuation of a successful workshop organized in NIPS 2017 (https://sites.google.com/view/med-nips-2017). It will feature a series of invited speakers from academia, medical sciences and industry to give an overview of recent technological advances and remaining major challenges.

Different from last year and based on feedback from participants, we propose to implement two novelties.
1. The workshop will accept paper submissions and have oral presentations with a format that aims to foster in depth discussions of a few selected articles. We plan to implement a Program Committee who will be responsible for reviewing articles and initiating discussions. The abstract track organized last year has brought a significant number of submission and has clearly demonstrated an appetite for more.
2. Along the workshop, we will host a challenge on outlier detection in brain Magnetic Resonance Imaging (MRI), which is one of the main applications of advanced unsupervised learning algorithms and generative models in medical imaging. The challenge will highlight a problem where the machine learning community can have a huge impact. To facilitate the challenge and potential further research, we provide necessary pre-processed datasets to simplify the use of medical imaging data and lower data-related entry barrier. Data collection for this challenge is finalized and ethical approval for data sharing is in place. We plan to open the challenge as soon as acceptance of the workshop is confirmed.

Sat 5:45 a.m. - 6:00 a.m.
Welcome (Talk)
Ender Konukoglu, Ben Glocker, Hervé Lombaert, Marleen de Bruijne
Sat 6:00 a.m. - 6:45 a.m.
Making the Case for using more Inductive Bias in Deep Learning (Talk)
Max Welling
Sat 6:45 a.m. - 7:30 a.m.

U-net based architectures have demonstrated very high performance in a wide range of medical image segmentation tasks, but a powerful segmentation architecture alone is only one part of building clinically applicable tools. In my talk I'll present three projects from the DeepMind Health Research team that address these challenges. The first project, a collaboration with University College London Hospital, deals with the challenging task of the precise segmentation of radiosensitive head and neck anatomy in CT scans, an essential input for radiotherapy planning [1]. With a 3D U-net we reach a performance similar to human experts on the majority of anatomical classes. Beside some minor architectural adaptations, e.g. to tackle the large imbalance of foreground to background voxels, a substantial focus of the project was in generating a high-quality test set [2] where each scan was manually segmented by two independent experts. Furthermore we introduced a new surface based performance metric, the surface DSC [3], designed to be a better proxy for the expected performance in a real-world radiotherapy setting than existing metrics. The second project, together with Moorfields Eye Hospital, developed a system that analyses 3D OCT (optical coherence tomography) eye scans to provide referral decisions for patients [4]. The performance was on par with world experts with over 20 years experience. We use two network ensembles to decouple the variations induced by the imaging system from the patient-to-patient variations. The first ensemble of 3D U-nets creates clinically interpretable device-independent tissue map hypotheses; the second (3D dense-net based) ensemble maps the tissue map hypotheses to the diagnoses and referral recommendation. Adaptation to a new scanning device type only needed sparse manual segmentations on 152 scans, while the diagnosis model (trained with 14,884 OCT scans) could be reused without changes. The third project deals with the segmentation of ambiguous images [5]. This is of particular relevance in medical imaging where ambiguities can often not be resolved from the image context alone. We propose a combination of a U-net with a conditional variational autoencoder that is capable of efficiently producing an unlimited number of plausible segmentation map hypotheses for a given ambiguous image. We show that each hypothesis provides an overall consistent segmentation, and that the probabilities of these hypotheses are well calibrated.

[1] Nikolov et al. (2018) "Deep learning to achieve clinically applicable segmentation of head and neck anatomy for radiotherapy" (soon available on ArXiv) [2] Dataset will be soon available at https://github.com/deepmind/tcia-ct-scan-dataset [3] Implementation available at https://github.com/deepmind/surface-distance [4] De Fauw, et al. (2018) "Clinically applicable deep learning for diagnosis and referral in retinal disease" Nature Medicine (in press). https://doi.org/10.1038/s41591-018-0107-6 (fulltext available from https://deepmind.com/blog/moorfields-major-milestone/ ) [5] Kohl, et al. (2018) "A Probabilistic U-Net for Segmentation of Ambiguous Images". NIPS 2018 (accepted). Preprint available at https://arxiv.org/abs/1806.05034

Olaf Ronneberger
Sat 7:30 a.m. - 8:00 a.m.
Coffee Break (Break)
Sat 8:00 a.m. - 9:00 a.m.
Oral session I (Presentation)
Jonas Adler, Ajil Jalal, Joseph Cheng
Sat 9:00 a.m. - 10:00 a.m.
Lunch (Break)
Sat 10:00 a.m. - 10:45 a.m.


Holger Roth
Sat 10:45 a.m. - 11:45 a.m.
Oral session II (Presentation)
Sil C. van de Leemput, Adrian Dalca, Karthik Gopinath
Sat 11:45 a.m. - 1:15 p.m.
Poster session (Poster presentations)
David Zeng, Marzieh S. Tahaei, Shuai Chen, Felix Meister, Meet Shah, Anant Gupta, Ajil Jalal, Eirini Arvaniti, David Zimmerer, Konstantinos Kamnitsas, Pedro Ballester, Nathaniel Braman, Udaya Kumar, Sil C. van de Leemput, Junaid Qadir, Hoel Kervadec, Mohamed Akrout, Adrian Tousignant, Matthew Ng, Raghav Mehta, Miguel Monteiro, Sumana Basu, Jonas Adler, Adrian Dalca, Jizong Peng, Sungyeob Han, Xiaoxiao Li, Karthik Gopinath, Joseph Cheng, Bogdan Georgescu, Kha Gia Quach, Karthik Sarma, Dave Van Veen
Sat 1:15 p.m. - 2:00 p.m.
Is your machine learning method solving a real clinical problem? (Talk)
Tal Arbel
Sat 2:00 p.m. - 3:00 p.m.
Oral session III (Presentation)
Nathaniel Braman, Adrian Tousignant, Matthew Ng
Sat 3:00 p.m. - 3:15 p.m.
Closing remarks (Talk)
Ender Konukoglu, Ben Glocker, Hervé Lombaert, Marleen de Bruijne

Author Information

Ender Konukoglu (ETH Zurich)
Ben Glocker (Imperial College London)
Hervé Lombaert (Ecole de Technologie Superieure (ETS Montreal))

Hervé is Associate Professor at ETS Montreal, Canada and Affiliated Research Scientist at Inria, France - His research interests are in Statistics on Shapes, Data & Medical Images. He had the chance to work in multiple centers, including Microsoft Research (Cambridge, UK), Siemens Corporate Research (Princeton, NJ), Inria Sophia-Antipolis (France), McGill University (Canada), and Polytechnique Montreal (Canada). He is also a recipient of the François Erbsmann Prize, a top prize in Medical Image Analysis, earned a Best Thesis Award at Polytechnique Montreal, as well as several other prizes and fellowships - Hervé co-organized 4 workshops and special sessions in major international conferences, including the ICML Workshop on Machine Learning Meets Medical Imaging in 2015.

Marleen de Bruijne (Erasmus MC)

More from the Same Authors