Hauptseminar/Master Seminar: Deep Learning for Medical Applications

Prof. Nassir Navab, Dr. Shadi Albarqouni; Magda Paschali

Type: Advanced Seminar Course Module IN8901
Type: Master Seminar Module IN2107
SWS: 2+0
ECTS: 5 Credits
Location: CAMP Seminar Room, 03.13.010
Time: Mondays, 10-12
Course Language: English

Announcements

• 24-08-2018: Paper list is almost ready!
• 29-06-2018: Registration, through TUM Matching Platform, should be done between 29.06. to 04.07.18
• 16-06-2018: An introductory meeting: Wednesday, 27.06.2018 (16:00-16:30) in CAMP Seminar Room, 03.13.010.
• 16-06-2018: Website is up!

Introduction

• Deep Learning is growing tremendously in Computer Vision and Medical Imaging as well. Highly impacted journals in medical imaging community, i.e. IEEE Transaction on Medical Imaging, published recently their special edition on Deep Learning [1]. The Seminar will propose a list of recent scientific articles related to the main current research topics in deep learning for Medical Applications together with some interesting papers from other communities.

Registration

• Interested students should attend the introductory meeting to enlist in the course.
• Students can only register through TUM Matching Platform themselves if the maximum number of participants hasn't been reached (please pay attention to the Deadlines).
• A maximum number of participants: 20.

Requirements

In this Master Seminar (formerly Hauptseminar), each student is asked to send three preferences from the list, then he will be assigned one paper. In order to successfully complete the seminar, participants have to fulfill these requirements:

• Presentation: The selected paper is presented to the other participants (20 minutes presentation 10 minutes questions). Use the CAMP templates for PowerPoint camp-tum-jhu-slides.zip, or Latex: CAMP-latex-template.
• Written Report: A document of maximum 2 pages should be submitted before the deadline.
• Attendance: Participants have to participate actively in all seminar sessions.

The students are required to attend each seminar presentation which will be held during this course. Each presentation is followed by a discussion and everyone is encouraged to actively participate. The report must include all references used and must be written completely in your own words. Copy and paste will not be tolerated. Both report and presentation have to be done in English.

You need to upload your presentation together with your report here. A detailed description is written in the Readme file inside the repository. Please create your account on LRZgit and request an access to the repository.

Submission Deadline : You have to submit both the presentation and the written report two weeks right after your presentation session. Your access to the repository will not be any longer granted after the deadline.

Schedule

Date	Session: Topic	Slides	Students
27.06.2018	Preliminary Meeting	Slides
15.10.2018	Paper Assignment --> Updated: There is No Class on the first day!
22.10.2018	Paper Assignment --> Updated: There is No Class. Papers have been assigned.
12.11.2018	Presentation Session 1: Supervised Learning (Detection)
19.11.2018	Presentation Session 2: Supervised Learning (Segmentation)
26.11.2018	Presentation Session 3: Unsupervised Learning
03.12.2018	Presentation Session 4: Semi/Weakly/Webly-Supervised Learning
10.12.2018	Presentation Session 5: Domain Adaptation
17.12.2018	Presentation Session 6: Few/Zero-Shot Learning
07.01.2019	Tentative
14.01.2018	Presentation Session 7: Uncertainty
21.01.2019	Conclusion

List of Topics and Material

The list of papers:

Topic	No	Title	Conference/Journal	Tutor	Student (Last name)	Link
Supervised Learning (Detection)	1	Learning to Segment Every Thing	CVPR 2018	Magda	Soham	PDF
	2	Articulated Multi-Instrument 2-D Pose Estimation Using Fully Convolutional Networks	TMI 2018	Magda	Muhammed Hasan	PDF
	3	Fully automatic detection and segmentation of abdominal aortic thrombus in post-operative CTA images using Deep Convolutional Neural Networks	MedIa 2018	Magda	Granit	PDF
Supervised Learning (Segmentation)	4	Efficient Annotation of Segmentation Datasets with Polygon-RNN++	CVPR 2018	Ari	Rajat	PDF
	5	Pelvic Organ Segmentation Using Distinctive Curve Guided Fully Convolutional Networks	MedIa 2018	Ari	Veronika	PDF
	6	Ω-Net (Omega-Net): Fully automatic, multi-view cardiac MR detection, orientation, and segmentation with deep neural networks	MedIa 2018	Ari	Aleksandr	PDF
Unsupervised Learning	7	Unsupervised Learning by Predicting Noise	ICML 2017	Fernando	Ankita	PDF
	8	Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery	IPMI 2017	Beatrice	Robin	PDF
	9	Unsupervised Representation Learning by Predicting Image Rotations	ICLR 2018	Gerome	Parika	PDF
Semi/Weakly/Webly-Supervised Learning	10	WESPE: Weakly Supervised Photo Enhancer for Digital Cameras	CVPR 2018	Neda	Arian	PDF
	11	Weakly Supervised Learning for Whole Slide Lung Cancer Image Classification	MIDL 2018	Beatrice	Claudio	PDF
	12	Semi-supervised multi-organ segmentation via multi-planar co-training	BMVC 2018	Beatrice	Berkay	PDF
Domain Adaptation	13	Cross-Domain Weakly-Supervised Object Detection through Progressive Domain Adaptation	CVPR 2018	Fernando	Julius	PDF
	14	Unsupervised Reverse Domain Adaptation for Synthetic Medical Images via Adversarial Training	TMI 2018	Ignacio	Yordanka	PDF (Older Version)
	15	Domain Adaptation for MRI Organ Segmentation using Reverse Classification Accuracy	MIDL 2018	Ignacio	Hanwool	PDF
Few/Zero-Shot Learning	16	Meta-Learning for Semi-Supervised Few-Shot Classification	ICLR 2018	Ashkan	Wasiq	PDF
	17	Prototypical networks for few-shot learning	NIPS 2017	Shadi		PDF
	18	Low-Shot Learning from Imaginary Data	CVPR 2018	Gerome	Felix	PDF
Uncertainty	19	What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?	NIPS 2017	Shadi	Mariia	PDF
	20	Uncertainty in multitask learning: joint representations for probabilistic MR-only radiotherapy planning On the Effect of Inter-observer Variability for a Reliable Estimation of Uncertainty of Medical Image Segmentation	MICCAI 2018 MICCAI 2018	Shadi		PDF PDF
	21	A Probabilistic U-Net for Segmentation of Ambiguous Images	arXiv 2018	Shadi	Alessandro Mauro	PDF
	22	Exploring Uncertainty Measures in Deep Networks for Multiple Sclerosis Lesion Detection and Segmentation Uncertainty quantification using Bayesian neural networks in classification: Application to ischemic stroke lesion segmentation	MICCAI 2018 MIDL 2018	Christoph	Navdar	PDF PDF

If you are assigned a paper from MICCAI you are requested to present briefly one more work related to the topic that will be determined by your tutor, since MICCAI papers are significantly shorter than the other given options.

MICCAI: Medical Image Computing and Computer Assisted Intervention
CVPR: Conference on Computer Vision and Pattern Recognition
ICLR: International Conference on Learning Representations
TMI: IEEE Transaction on Medical Imaging
JBHI: IEEE Journal of Biomedical and Health Informatics
MedIA: Medical Image Analysis (Elsevier)
TPAMI: IEEE Transactions on Pattern Analysis and Machine Intelligence
BMVC: British Machine Vision Conference
MIDL: Medical Imaging with Deep Learning
NIPS: Neural Information Processing Systems

Literature and Helpful Links

A lot of scientific publications can be found online.

The following list may help you to find some further information on your particular topic:

• Microsoft Academic Search
• Google Scholar
• CiteSeer
• CiteULike
• Collection of Computer Science Bibliographies

Some publishers:

• ScienceDirect (Elsevier Journals)
• IEEE Journals
• ACM Digital Library

Libraries (online and offline):

• http://rzblx1.uni-regensburg.de/ezeit/ (Elektronische Zeitschriften Bibliothek)
• Verbundkatalog des Bibliotheksverbundes Bayern (BVB)
• Computer ORG
• http://www.ub.tum.de/ (TUM Library)
  • To get access onto the electronic library, see http://www.ub.tum.de/medien/ejournals/readme.html
  • "proxy.biblio.tu-muenchen.de" mit Port 8080 (nur fuer http). Damit klappen zumindest portal.acm.org und computer.org meistens
• Various proceedings of conferences in our AR-Lab, 03.13.036 (These proceedings are not for lending!)

Some further hints for working with references:

• JabRef is a Java program for comfortable working with Bibtex literature databases. Handy feature: if you know the PubMed ID for an article, JabRef can import data from there (via "Web Search/Medline").
• Mendeley is a cross-platform program for organising your references.

If you find useful resources that are not already listed here, please tell us, so we can add them for others. Thanks.