Date:Monday 29 - Wednesday 31 October 2018
Venue:EMBL - EMBL- Heidelberg, Meyerhofstraße, 69117, Heidelberg, Germany
Application opens:Monday 07 May 2018
Application deadline:Friday 15 June 2018
Participation:Open application with selection
Dates additional information:Up-front online sessions: 2nd, 9th 16th and 23rd October 2018 (12:00-14:00 CEST) On-site workshop: 29-31st October 2018 Follow-up online sessions (optional): 9th and 16th November 2018
This is a blended learning course on Machine Learning for Image Analysis, consisting of three online sessions with associated hands-on exercises prior to the workshop, a three day face-to-face workshop at EMBL Heidelberg and two optional online sessions with associated hands-on exercises after the workshop. The course is jointly organised by CORBEL, EMBL, German BioImaging and NEUBIAS (COST Action CA15124).
The course will be a great mix of intensive learning, extensive hands-on and community networking in the field of Machine Learning for Image Analysis.
- Up-front online sessions: Participants will review the fundamentals of machine learning in three up-front webinars complemented by online tutorials. The webinars will take place on 2nd, 9th and 16th October 2018, 12:00 - 14:00 but a recorded alternative can be provided.
- On-site workshop: Next, they will apply their knowledge at an on-site workshop (EMBL Heidelberg, October 29-31), in small interactive groups (the workshop has 20 available seats and ~8 trainer/lecturer), to both reference datasets and their own data. Topics to be practiced in these groups include: 2D and 3D segmentation with convolutional neural networks, Content aware image restoration, simulation of ground truth data, labeling strategies and transfer learning. Participants will be asked to pick one group for the practical sessions.
- Follow-up online sessions (optional): After the on-site workshop, two optional advanced training webinar, complemented by a online tutorials, will be given on 9th and 16th November 2018. These will focus simulation of data, transfer learning and boosting.
Practical Sessions (participants will be asked to choose one during registration):
Detection and Segmentation in microscopic images (Thorsten Falk):
- Topics: Segmentation with convolutional neural networks, Pattern in bio-medical image data
- Typical data: Results of multi-labeling high content screens
- Tools: Python, Keras, Tensorflow
Segmentation in 3D microscopy image stacks (Anna Kreshuk & Constantin Pape):
- Topics: 3D segmentation (with convolutional neural networks), 2D, 2.5D and 3D networks and their combinations, Pre- and post-processing tricks, Combination of deep and “shallow” learning
- Typical data: 3D datasets recorded by light or electron microscopy
- Tools: Python, Keras, PyTorch
Transfer learning and how to use synthetic data for supervised deep learning (David Rousseau & Pejman Rasti):
- Topic: Basics of transfer learning
- Typical data: PALM/STORM, 3D cells in spheroid imaged in light sheet fluorescence microscopy and 3D plant roots images in absorption X-ray tomography
- Tools: Python, Keras, Tensorflow
Content-aware image restoration (Martin Weigert):
- Topics: Content aware image restoration, Simulation of images for light microscopy
- Typical data: Pairs of light microscopic images imaged in ideal and suboptimal conditions
- Tools: Python, Keras, Tensorflow
This course is aimed at both core facility staff and research scientists.
Prerequisites for this workshop are programming skills in Python and ideally Tensorflow, Keras or Pytorch as well as basic knowledge of machine learning theory.
Participants should provide an outline of one image analysis task that holds potential to be ideally solved with machine learning. Neural networks have been successfully applied to various medical and biological imaging modalities including PALM/STORM, light sheet fluorescence microscopy, high-throughput microscopy, electron microscopy, X-ray tomography. However, they require observation-outcome-pairs for training. Ideally, you will provide annotated images.
After this course you should be able to:
Explain the fundamentals of machine learning methods suitable for image analysis
Consult users/colleagues in strategies to obtain ground truth
Give advice in training and using a neural-network
Perform simple quality control on the results of one selected ML approach
|Day 1 - Monday 29 October 2018|
|09:00 - 09:15||Registration|
|09:15 - 09:30||Welcome and Course introduction||Tobias Rasse, Anna Kreshuk|
|09:30 - 11:00||Flash talks by participants|
|11:00 - 11:15||Tea and Coffee break|
|11:15 - 12:45||Overview group projects||All instructors|
|12:45 - 13:30||Lunch||
|13:30 - 15:30||Practical (sample data)||in project groups|
|15:30 - 15:45||Tea and Coffee break|
|15:45 - 18:00||Practical (own data)||in project groups|
|18:00 - 18:45||Common pit faults in using CNN||Pannel discussion|
|19:00 - 19:45||Dinner|
|19:45 - 20:45||(Optional) continue with project||in project groups|
|Day 2 - Tuesday 30 October 2018|
|09:00 - 09:30||Blended workflows||Anna Kreshuk|
|09:30 - 10:00||Simulation of date||David Rousseau, Martin Weigert|
|10:00 - 11:00||Check & analyse results||in project groups|
|11:00 - 11:15||Tea and Coffee break|
|11:15 - 12:00||Practical (own data)||in project groups|
|12:00 - 12:30||Progress and issues||All|
|12:30 - 13:15||Lunch|
|13:15 - 15:00||Practical (own data)||in project groups|
|15:00 - 15:15||Tea and Coffee break|
|15:15 - 16:30||Practical (own data) and prepare presentation||Project groups|
|16:30 - 17:30||Presentation of completed projects||All|
|Day 3 - Wednesday 31 October 2018|
|09:00 - 10:00||Learned watershed: a case study on interleaving neural networks with dynamic programming||Fred Hamprecht|
|10:00 - 10:15||Tea and coffee break|
|10:15 - 10:45||Fine tuning and domain transfer||David Rousseau, Pejman Rasti|
|10:45 - 12:45||Practical||Project groups|
|12:45 - 13:30||Lunch|
|13:30 - 15:30||Project presentations||All|
|15:30 - 16:15||Wrap up and feedback|
|16:15||End of course|
Full details for the practical sessions (participants choose to join one group):
Detection and Segmentation in microscopic images (Thorsten Falk, University of Freiburg)
Neural Networks are extremely powerful in learning various image processing tasks. Their biggest strength is their easy adaptability to very different applications by simply training them with application-specific corresponding observation-outcome pairs. Especially in biology where imaging is by no means standardized and the number of applications on various scales is uncountable, neural networks can be the swiss army knife for image analysis. However, the required observation-outcome-pairs for training are also their biggest weakness. The key for good models is a sufficient amount of high quality training data covering the distribution of possible observations.
The focus of this workshop will be instance detection and segmentation in biomedical image data which can be solved with a comparably simple convolutional-deconvolutional feed-forward network. Image segmentation is one of the most frequently required but also one of the most demanding tasks with respect to training data generation. Each pixel of multiple images must be accurately and consistently annotated to obtain training data for a neural network. You will learn how to properly record, select and annotate images to train a neural network for your task. You will employ data augmentation techniques to reduce the image annotation effort to a moderate level. Finally, the concept of domain adaptation to transfer learnt models from one imaging device to another will be explained in a nutshell.
The workshop will require but also teach manual annotation in Fiji/ImageJ and use a new Deep Learning plugin for simple 2D tasks. 3D tasks can be tackled in a similar fashion but will require python and the caffe framework (this may change to keras/tensorflow depending on preparation time). Attendees need to provide 2D or 3D microscopic data for structure detection and/or segmentation with corresponding point markers indicating the positions and types of structures to be detected or (potentially multi-class) segmentation masks for training networks for structure segmentation.
Segmentation in 3D microscopy image stacks (Anna Kreshuk and Constantin Pape, EMBL)
We will focus on segmenting 3D data with convolutional neural networks. We’ll talk about network architectures that were shown to work well in 2.5D and 3D, about pre-processing and data augmentation, as well as necessary post-processing. We will also introduce and compare tools for groundtruth annotation and proof-reading in 3D.
Transfer learning and how to use synthetic data for supervised deep learning (David Rousseau, Pejman Rasti, Université d’Angers)
"A striking fact when gazing at the first layers of deep neural networks is that these layers almost look like Gabor wavelets. While promoting a universal framework, these machines seem to systematically converge toward tools that humans have been studying for decades. This empirical fact is used by computer scientists in the so-called transfer learning where the first layers of an already trained network are re-used to save time and improve results when limited data sets are available. Another current limitation to the application of deep learning is that this requires huge training data sets to avoid overfitting. Such training data sets have to be annotated when supervised learning is targeted. And, at the moment, the available data sets available among the bioimage analysis community are far from covering the large amount problematics in bioimaging. Here also some possible alternatives exists. This includes the generation of realistic automatic annotated data sets with synthetic simulated images.
During this hands on you will learn the basics of transfer learning with a Jupiter notebook and will learn how to use synthetic data for supervised deep learning. This can be applied to any type of bioimaging problem. A large panel of examples will be given for illustration among which 2D microtubule imaged with PALM/STORM, 3D cells in spheroid imaged in light sheet fluorescence microscopy and 3D plant roots images in absorption X-ray tomography."
Content-aware image restoration (Martin Weigert, MPI CPG)
Fluorescence microscopy of living cells or organisms often results in noisy and axially blurred images, owing to the limited laser power and number of focal planes that are compatible with specimen health or temporal resolution. This is in contrast to acquisitions of fixed specimen, where such constraints are often not present, and where the signal-noise-ratio (SNR) and axial sampling can be chosen almost freely.
In this part of the tutorial, we will explain how machine learning can be used to improve the quality of images that were acquired under adversarial imaging conditions, with the help of adequately acquired high quality images. Participants will learn how to construct image restoration pipelines that are tailored to specific imaging situations and organisms, and will be guided through all the necessary steps: From data pre-processing, model training, and application to new images to the final deployment via Fiji plugins that can be shared with collaborators.
Ideally, the participants should already have a specific experimental situation in mind, where normal imaging conditions would preclude sufficient image quality. To take full advantage of this practical session, they are encouraged to bring already acquired corresponding images/volumes of samples acquired at different (low and high quality) conditions.