Post graduate course in

Methods and Algorithms for Image Segmentation

(MAIS 2017, April-May 2017)


Organizer and lecturer Guest Professor Heung-Kook Choi

For whom
This course is primarily intended for PhD students in image processing at Department of Information Technology, but open to anyone with a priori knowledge in image processing and an interest in image segmentation.

Background
Image segmentation is one of the most important, but also most difficult steps of any automated, or semi-automated analysis task. More powerful computers combined with algorithm development during the past year has shown an enormous increase in the performance of methods based on deep convolutional networks for a range of scientific tasks; image segmentation being one of them. However, when applied to image data, the output of a deep convolutional network is a probability map, that in turn has to be further processed to provide a final segmentation result. This course will focus on image segmentation in general, and also look into segmentation of probability maps.

Goal
The goal is to gain an understanding of the general concepts of a rage of image segmentation approaches, including traditional and state of the art algorithms. The participants should also know how to implement and evaluate segmentation methods on a general set of images as well as their own data.

Contents
The course consists of three parts. The first part consists of three traditional lectures on general image segmentation. Some written material will be handed out and some tasks should be handed in. The second part is a series of seminars held by the participants. The participants will dig deeper into a specific segmentation method (watersheds, level sets, global/local thresholding, active contouring, texture, morphology, etc.) by reading relevant scientific publications (original method publications as well as more recent improvements/version). Part of the task is to also find an existing implementation, or make their own, test and evaluate it on a common dataset available to all course participants. Finally, the third part is an oral and written presentation of the selected algorithm, including a demonstration of the method's implementation and evaluation on the common dataset as well as the participant's own data. Focus will be on pros and cons of the method, pre- and post-processing options, limitations and possibilities for improvements.The participants will also be asked to read one paper in relation to each presented method, selected by the other participants.

Language English

Pre-course task Discuss suitable datasets and algorithms with your supervisors to select the methods that are most relevant/interesting/promising for your research field.

Venue Room 2414b, Polacksbacken, Uppsala

Literature Handouts and scientific articles.

Examination Participation in the lectures and seminars, a few exercises, and an oral and a written report on the selected segmentation algorithm, together with an implementation of the image segmentation approach.

Credits 5 hp

Maximum number of participants 10

Deadline for applying to the course: 03 March 2017.

Syllabus of the MAIS 2017

Part 1:
Class 1: 07 April (Friday) 10:15-12:00
      # Course introduction
      # What’s the image segmentation role for the image processing and image analysis?
      # Creating an algorithm structure of simple pipeline or combining with any methods
      # Pixelwise, shape, mass or clustering segmentation
      # Image segmentation using textures, shapes, colors, structures, morphologies
Class 2: 10 April (Monday) 10:15-12:00
      # An algorithm development depends on the image types: Microscopy, Confocal microscopy, Electronic microscopy, MRI, CT, PET, SPECT, Ultrasound, Satellite, Camera, etc.
      # Do you access the segmentation using of some image features?
      # Classical image segmentation of the statistic classification
Class 3: 11 April (Tuesday) 08:15-10:00
      # Overview all kinds of segmentation methods: Thresholding, Filters, Region growing, Active contouring,
         Convolution neural networks, Watershed, Level sets, Texture, Morphology, Genetic method, Fuzzy theory,
Statistic method, etc.
      # Mathematical method based on a creating modeling

Part 2:
Class 4: 21 April (Friday) 08:15-10:00
      # Paper presentation, Suggestion to additional problem solutions, Q&A and Comments
Class 5: 21 April (Friday) 10:15-12:00
      # Paper presentation, Suggestion to additional problem solutions, Q&A and Comments
Class 6: 24 April (Monday) 10:15-12:00
      # Paper presentation, Suggestion to additional problem solutions, Q&A and Comments
Class 7: 25 April (Tuesday) 08:15-10:00
      # Paper presentation, Suggestion to additional problem solutions, Q&A and Comments
Class 8: 26 April (Wednesday) 08:15-10:00
      # Paper presentation, Suggestion to additional problem solutions, Q&A and Comments

Part 3:
Class 9: 04 May (Thursday) 08:15-10:00
      # Demonstration for the segmented image result, Q&A and Comments
Class 10: 04 May (Thursday) 10:15-12:00
      # Demonstration for the segmented image result, Q&A and Comments
Class 11: 05 May (Friday) 08:15-10:00
      # Demonstration for the segmented image result, Q&A and Comments
Class 12: 05 May (Friday) 13:15-15:00
      # Demonstration for the segmented image result, Q&A and Comments
Class 13: 05 May (Friday) 15:15-17:00
      # Demonstration for the segmented image result, Q&A and Comments
Class 14: 09 May (Tuesday) 08:15-10:00
      # Image segmentation applications in the world
      # Comprehensive summary and criticism
      # Course closing


Recommendation Journals for the paper selection

Journals                                                                                                    ISSN               IF
1. IEEE Transactions on Image Processing                                              1057-7149       3.735
2. IEEE Transactions on Medical Imaging                                               0278-0062       3.756
3. IEEE Transactions on Pattern Analysis and Machine Intelligence       0162-8828       6.077
4. Medical Image Analysis                                                                        1361-8415       4.565
5. Methods                                                                                                 1046-2023       3.503
6. Pattern Recognition                                                                               0031-3203       3.399
7. Transactions on Information Technology in Biomedicine                    1089-7771       2.493

8. Cancer Cytopathology                                                                           1934-6638       2.471
9. Cell and Tissue Research                                                                       0302-766X     2.948
10. Cellular Oncology                                                                               2211-3428       3.562
11. Cytometry: Part A                                                                               1552-4930       3.181
12. Cytometry: Part B                                                                               1552-4957       2.822
13. Journal of Histochemistry & Cytochemistry                                       0022-1554       2.538

14. Nature Medicine                                                                                  1078-8956       30.357
15. Nature Methods                                                                                   1548-7091       25.328
16. Science                                                                                                0036-8075       34.661
17. Science Translational Medicine                                                          1946-6234       15.843

Recommendation a papper selection of conferences

1. International Association for Pattern Recognition (IAPR)
2. International Conference on Image Analysis and Processing (ICCAP)
3. IEEEE Conference on Computer Vision and Pattern Recognition (CVPR)
4. IEEE International Conference Image Processing (ICIP)
5. Scandinavian Conference on Image Analysis (SCIA)
6. International Conference on Machine Learning (ICML)
7. International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI)


Reference images

1. 10 microscopic images of breast carcinomas: The immunohistochemically(IHC) stained cell nuclei tissue sections have taken breast carcinoma patients. and he histological tissue sections were stained a developed color image of Diaminobenzidine(DAB) and counterstained with Hematoxylin.
2. 15 MR images: 5 Normal Control images (NC), 5 Mild Cognitive Impairment images (MC) and 5 Alzheimer’s Disease images (AD).