Image and Pattern Analysis for Multidisciplinary Computational Anatomy

26 th November, 2019, Auckland, New Zealand

in conjunction to 5th Asian Conference on Pattern Recognition


Accepted Paper
ID:1, 2, 3, 6, 7

For accepted papers, pleas upload the USB version of pdf on easychar by 1st October.
These pdf will be distributed during ACPR.
The editor will ask the final revised version for post-proceedings.


Scope and Motivation

Multidisciplinary Computational Anatomy (MCA) achieves mathematical foundation to deal with humans both from the static anatomy of organs and dynamic anatomy of living human constructed using multidisciplinary information on the human from cells to body. For the achievement of MCA from the viewpoints of visual computing, we establish a mathematical and theoretical methodologies which allow us to achieve comprehensive understanding of the human body using computer vision, mathematical imaging, pattern recognition and artificial intelligence. The workshop on mathematical image and pattern analysis for (ipaMCA) focuses on mathematical and computational aspects of biomedical imaging and image analysis and on these relations to computer vision and pattern recognition in MCA.


Topics

We call for papers for applications of the methods to anatomy, autopsy, biopsy, physiology and nano-biology. The expected areas for contributions are following. But all aspects of mathematical treatment medical imaging and image analysis are welcome.

In biomedical imaging and image analysis for MCA, the researches aim to design systems that assist medical doctors. In contrast, in computer vision researches focus construct machines that see. The forum aims to derive a bridge on this gap from the viewpoints of mathematical and computational aspects in computer vision and pattern recognition for MCA.


Programme

8:50Registration

Session 1 Chaired by A. Imiya

9:00Opening RemarkKensaku Mori
9:20Invited Talk 1Yoshi Masutani
CA to MCA: Four Aspects toward Comprehensive Understanding of Human Body
10:00Invited Talk 2Avan Suinesiaputra
Computational Anatomy of Cardiac Shape and Motion in Large-Scale Population Studies
10:40Break

Session 2 Chaired by Harvey Ho

11:00Invited Talk 3Shin'ichi Satoh
Nationwide Medical Bigdata, Secure Cloud, and Artificial Intelligence
11:40Invited Talk 4Brendan McCane
Data is key
12:20Lunch

Session 3 (Contributed Papers)Chaired by Yukiko Kenmochi

10min presentation each without discussion

Shun Obikane and Yoshimitsu Aoki
Weakly supervised domain adaptation with point supervision in histopathological image segmentation
Morio Kawabe, Yuri Kokura, Takashi Ohnishi, Kazuya Nakano, Hideyuki Kato, Yoshihiko Ooka, Tomoya Sakai and Hideaki Haneishi
Blood vessel enhancement in liver region from a sequence of angiograms taken under free breathing
Maxime Berg, Changwei Zhang and Harvey Ho
Real-time morphing of the Visible Man liver with intrahepatic vasculatures
Aurélien Bourgais, Sarah Ance and Harvey Ho
Development of 3D Physiological Simulation and Education Software for Pregnant Women
Kento Hosoya, Kouki Nozawa and Atsushi Imiya
Resolution Conversion of Volumetric Array Data for Multimodal Medical Image Analysis
14:50Discussion with All Speakers
15:20Break

Session 4 Chaired by Hideaki Haneishi

16:00Invited Talk 5Mohammad Norouzifard
Glaucoma Detection by an Ensemble of Traditional Classifiers with a Focus on Theoretical and Mathematical Aspects
16:40Invited Talk 6Ken'ichi Morooka
AI-based Cancer Screening System Using Multi-focal Pathological Images
17:20Closing Remarks

Invited Speakers

Professor Yoshitaka Masutani Hiroshima City University Japan

http://rsw.office.hiroshima-cu.ac.jp/Profiles/13/0001203/prof_e.html

CA to MCA: Four aspects toward comprehensive understanding of human body

In the early stage of Computational Anatomy (CA), the discipline focused on shapes and structures in the human anatomy. Aimed at more comprehensive understanding of human body with medical image information, the recent progress of CA developed to Multidisciplinary Computational Anatomy (MCA), which can be featured by four additional aspects; (1) temporal analysis, (2) multi-functional information, (3) multi-scale image analysis, and (4) pathological information. As examples, the following topics related to the mathematical foundations of MCA are presented and discussed in the lecture; (a) diffusion MRI signal models and parameter inference by synthetic learning, (b) stain transformation of pathology images by GAN, (c) multi-scale registration among MRI and pathology images and (d) partially-rigid registration of temporal series of MR images. In addition, open questions in the MCA researches are introduced for researchers in the field of computer vision and pattern recognition.

Prof. Yoshi Masutani currently conducts Medical Imaging Lab. in Hiroshima City Univ. Graduate School of Information Sciences, Japan. He obtained Ph.D. degrees for biomedical engineering field in 1997 and for medical science in 2010, both at University of Tokyo, Japan. He has joined several multi-disciplinary research institutes where clinicians, scientists and engineers collaborate, such as Univ. Hospital of Hamburg-Eppendorf, Univ. of Chicago Hospital, and Univ. of Tokyo Hospital. Prof. Masutani’s research specialty covers biomedical image analysis and software development, especially for dMRI.

Professor Shin'ichi Satoh (NII, Japan)

http://www.satoh-lab.nii.ac.jp/

Nationwide Medical Bigdata, Secure Cloud, and Artificial Intelligence

Robustness and reliability of Deep Learning (DL)-based medical image diagnosis require sufficient training data in terms of both qualityand quantity. The robustness demands the quality of the images acquired from many different medical institutions, and the reliability could only be achieved by the large quantity of data a.k.a bigdata.To obtain such multi-institutional, large-scale training data, we should overcome different ethical code adopted by each hospital across Japan. We therefore built a cloud platform for (i) collecting multi-modal large-scale medical images from nationwide hospitals through 6 leading Japanese medical societies and (ii) collaboration research on DL-based Computer Aided Diagnosis (CAD) being carried out by 6 highest standard institutes of informatics in Japan. In association with the medical societies, over 50 informatics researchers have conducted various DL-CAD researches on our platform using more than 60 million collected images. I will discuss the outcome so far including academic results presented at major conferences and published on medical journals, as well as ongoing implementation of the DL-CAD to a clinical environment.

Prof. Shin'ichi Satoh is currently the director of the Research Center for Medical Bigdata, National Institute of Informatics, Japan. His research interests include image processing, video content analysis and multimedia database. Currently he is leading the video processing project at NII, addressing video analysis, indexing, retrieval, and mining for broadcasted video archives. He received his BE degree in Electronics Engineering in 1987, his ME and PhD degrees in Information Engineering in 1989 and 1992 at the University of Tokyo. He joined National Center for Science Information Systems (NACSIS), Tokyo, in 1992. He is a full professor at National Institute of Informatics (NII), Tokyo, since 2004. He was a visiting scientist at the Robotics Institute, Carnegie Mellon University, from 1995 to 1997. He has served on technical program committees for several international conferences, including ACM Multimedia, ICME, ICPR, ICCV, SIGIR, and WWW. He served as program co-chairs for Pacific-Rim Conference on Multimedia in 2004 (PCM2004), Multimedia Modeling Conference (MMM2008), International Conference on Multimedia Retrieval in 2011 (ICMR2011), ACM Multimedia 2012, International Conference on Internet Multimedia Computing and Service in 2012 (ICIMCS2012), International Workshop on Content-Based Multimedia Indexing in 2012 (CBMI2012), Pacific-Rim Symposium on Image and Video Technology in 2011 and 2013 (PSIVT2011, 2013), and Multimedia Modeling Conference (MMM2017). He will serve as a general co-chair of International Conference on Multimedia Retrieval in 2018 (ICMR2018). He served as a track co-chair for International Conference on Pattern Recognition (ICPR2014). He is on the Editorial Board of the following journals: IEEE Transactions of Circuits and Systems for Video Technology (TCSVT), International Journal on Multimedia Tools and Applications (MTAP), International Journal of Multimedia Intelligence and Security (IJMIS), International Journal on Multimedia Information Retrieval, and International Journal on Computer Vision (IJCV).

Professor Ken'ichi Morooka (Kyushu University, Japan)

http://fortune.ait.kyushu-u.ac.jp/morooka_lab/index-e.html

AI-based Cancer Screening System Using Multi-focal Pathological Images

Cervical cancer screening is useful for early detection of cancers with less invasive natures. In the screening, cytotechnologists observe a tissue sample taken out from human body, and _nd pre-cancerous and cancer cells from the sample. Generally, one sample includes tens of thousands of cells. Among them, the number of cancer cells is much smaller than that of normal cells. Moreover, in the case of cervical cancer screening in Japan, only 120 of every 10,000 people may carry cancer cells, and 7 of them will be diagnosed as suffering from cancer. Owing to these, the detection of cancer cells is a hard and time-consuming task. Now, our research group has been developing an AI-based automatic system of cervical cancer screening using multi-focus pathological images. The fundamental techniques for this system include the recovery of 3D cell shapes from multi-focus pathological images, and the classification of the cells using their 3D shapes. I will present the fundamental techniques of our automatic cancer screening system.

Ken'ichi Morooka received his M.S. and Ph.D. degrees from Kyushu University, in1997 and 2000, respectively. Currently, he is an associate professor in Graduate School of Information Science and Electrical Engineering, Kyushu University. His research interests cover computer-aided support system for therapy and surgery by image information processing and machine learning. Moreover, he has been studying medical applications using 3D shapes of human organs and cells.

Professor Brendan McCane (University of Otago)

https://www.cs.otago.ac.nz/staffpriv/mccane/index.php

Data is key

In this talk I will outline a general problem with computational methods in medical imaging and argue that many problems disappear when there is access to sufficient data. I will focus on one particular problem: estimating shape and pose of vertebrae from orthogonal x-rays, but will also touch on other problems.

Prof. McCane graduated with a PhD from James Cook University in 1996 and joined the Department of Computer Science at the University of Otago in 1997. He only intended staying for about four years, but he is still here because New Zealand is such a great place to live. His research interests span the areas of computer vision, machine learning and biomedical imaging, and he has published extensively in all of these areas. His interests range from theoretical to applied and he has worked with people from a wide range of disciplines including physiotherapy, anatomy, surgery, dentistry, paleontology, and agriculture.

Mohammad Norouzifard (Auckland University of Technology, New Zealand)

https://www.linkedin.com/in/mohammad-norouzifard/

Glaucoma Detection by an Ensemble of Traditional Classifiers with a Focus on Theoretical and Mathematical Aspects

Glaucomatous optic neuropathy is listed as the fourth major cause of eye disease by the WHO. In 2015, an estimated 3 million people were blind due to this disease. Besides deep learning strategies recently, structural and functional methods are still in common use to detect and monitor glaucomatous damage via using ensemble or hybrid classifiers such as AdaBoost, XGBoost, or random forests. The talk presents a fused pattern recognition model defined by supervised classifiers and an ensemble learning method. It also discusses Z-scores and PCA statistical analysis. The proposed model achieved an F1 score of 0.82 and an accuracy of 82% using 5-fold cross validation on a small data-set of 107 RNFL data from normal eyes and 68 RNFL data from eyes with glaucoma; 25% of data have been selected randomly for testing. The model may assist general ophthalmologists in their daily screening to confirm their diagnosis, thereby increasing the accuracy of diagnosis, and is also of support for glaucoma screening in research. Experimental tests illustrate that having only a small dataset still ensures highly accurate results by applying the proposed supervised model.

Mohammad Norouzifard is close to the end of his PhD project at Auckland University of Technology (AUT), New Zealand, also working as a teaching and research assistant at AUT and at University of Auckland. His research interests include computer vision, machine learning, deep learning and medical image analysis. In his PhD project, he applied various deep learning and traditional concepts for detecting glaucoma. He received in 2009 his master's degree in AI & Computer Engineering for a project to predict diabetic retinopathy and AMD of patients at early stages. 2007 to 2017, he has been a lecturer and ICT trainer at the University of Applied Science and Technology at Tehran, Iran; during this time, he already started with his research in medical imaging. He published twelve journal and conference articles during his PhD project. Since 2018 he is a member of the Association for Research in Vision and Ophthalmology (ARVO).

Dr. Avan Suinesiaputra (University of Auckland, New Zealand)

https://unidirectory.auckland.ac.nz/profile/a-suinesiaputra

Computational Anatomy of Cardiac Shape and Motion in Large-Scale Population Studies

One of the most challenging tasks in cardiovascular research is to understand how cardiac remodels due to external forces. In the onset of cardiac injury or disease, the heart initially experiences adaptive remodelling to compensate the loss of function before progressively deteriorating until symptoms become clinically evident. It is therefore important to understand variations of heart size, shape and function in the asymptomatic population who have risks developing a heart disease. In this talk, I want to share our experience in the Cardiac Atlas Project to analyse large-scale population studies by using statistical shape analysis. I will show how a mathematical model of the heart anatomy has helped us to consistently extract features from different subjects to quantify shape and function abnormalities in patients with myocardial infarction and congenital heart disease. I will also show how machine learning algorithms have leveraged the power of large-scale imaging data to automate cardiac shape analysis. I will conclude the talk by highlighting several challenges in computational anatomy of the heart.

Avan Suinesiaputra obtained his MSc degree with Cum Laude in computational sciences from the University of Amsterdam, the Netherlands, and PhD degree in medical image analysis from the University of Leiden, the Netherlands. He previously obtained his BE in computer science from Institut Teknologi Bandung, Indonesia. He is currently a Senior Research Fellow at the Department of Anatomy and Medical Imaging, University of Auckland, New Zealand. Previously, he was a postdoc researcher at the Department of Radiology, Leiden University of Medical Center, the Netherlands. He has co-authored 28 journal publications, 3 book chapters, and 26 papers in conference proceedings. He was awarded best reviewer by the IEEE Journal of Biomedical and Health Informatics (2014) and golden reviewer by the Society of Cardiovascular Magnetic Resonance (2018). He has been co-organizing annual workshops of the Statistical Atlases and Computational Modelling of the Heart or STACOM (2011-2019). He is currently managing the Cardiac Atlas Project database and website (www.cardiacatlas.org). His main research interests are in the development of computational algorithms for medical images, particularly with machine learning methods.


Organizing committee

Organisers

Program co-chairs

Program Committee Members

  • Angelica Aviles-Rivero (UK)
  • Weidong Cai (Australia)
  • Bernhard Egger (USA)
  • Hideaki Haneishi (Japan)
  • Jan Hering (Czech)
  • Byung Woo Hong (Korea)
  • Hayato Itoh(Japan)
  • Xiaoyi Jiang (Germany)
  • Jacques-Olivier Lachaud (France)
  • Lukas Lang (UK)
  • Yoshitaka Masutani (Japan)
  • Vannary Meas-Yedid (France)
  • Yoshito Otake (Japan)
  • Nicolas Passat (France)
  • Isabelle Sivignon (France)
  • Robin Strand (Seweden)
  • Joao Manuel R. S. Tavares (Portugal)
  • Antoine Vcavant (France)
  • Martin Welk (Austria)
  • Bertrand Kerautret (France)

Important Dates

Submission deadline 5 th September 2019 (No extension)
(In the same week of ACPR notification)
ACPR organisers provide acceptance letters for visa application on day of author notification
Author notification 25 th September 2019
Final manuscript 15 th October 2019

Paper Submission

https://easychair.org/conferences/?conf=ipamca2019

Submissions need to follow the single-blind policy and be formatted in LNCS style, with a maximum of 14 pages (including references).

Workshop proceedings will be published after the conference in the CCIS series of Springer through the ACPR organizers (publication chair). At the conference, workshop papers will be distributed on USB, together with the ACPR main conference proceedings. Publication in the CCIS volume requires that the paper has been presented at the ACPR workshop by one of the co-authors.

Revised long papers presented at the workshop are also invited to special issue of Springer Nature Computer Science. See special issues in the main conference.

Venue

The workshop will be in rooms in the 8th floor of Building WG of AUT, which is just a short 5 minutes walk from Aotea Centre; just follow the ACPR signs.

Sponsors