• About

Collaborate Through Research, Partner for Progress

Bringing researchers around the world for faster delivery of benefits from science and research.

Awards 2017

Dr. Hillary Klonoff-Cohen


Dr. Yoichiro Yamamoto –an eminent scientist and an expert in the field ofmedicineand computer science/mathematics especially including Artificial Intelligence (AI) is presently working as a unit leader of Pathology Informatics Unit at RIKEN Center for Advanced Intelligence Project in Japan. He holds Japanese board certifications in Pathology and Cytology.
Dr. Yamamoto received his M.D. degree from Tohoku University, School of Medicine in 2004. After residency training, he started Ph.D.research under the tutelage of Prof. Manabu Fukumoto – one of the leading scientists in the radiation biology and pathology field. He received his Ph.D. degree from Tohoku University, Graduate School of Medicine in 2009. His Ph.D. thesis focused on the cell dynamics in the carcinogenesis process. The researches resulted in publications in the several fields such as cancer research field and medical physics field.
Since then, Dr. Yamamoto worked as an assistant professor of the Division of Diagnostic Pathology at Nippon medical school hospital. From 2012, he visited USA several timesand researched mathematical modeling of cancer cell dynamics as a visiting scientist at Mayo clinic and as a guest researcher of Program for Evolutionary Dynamics at Harvard University. This mathematical research led to his current Artificial Intelligenceresearch. In 2013, he became Assistant Professor of Department of Pathology at Shinshu University School of Medicine, Japan.In 2014-2015, he got engaged with a research oncomputer science as a visiting scientist at University of Heidelberg in Germany, and became successful in developing the novel method of combining pathology and machine learning.
Dr. Yamamoto became the unit leader of Pathology Informatics Unit at RIKEN in 2017. The mission of his unit is to discover novel diseasemechanisms and therapies through state-of-the-art Artificial Intelligence technologiesand medical research.
Dr. Yamamoto was selected by the World Scientists Forum for “Eminent Scientist of the Year 2017” international award for his research innovations and outstanding contributions in the field of science and medicine especially in combining Advanced Artificial Intelligence and Computer Science with Medical Research and Carcinogenesis.

Artificial intelligence and mathematical analysis in the medical research field
Yoichiro Yamamoto

RIKEN Center for Advanced Intelligence Project, Pathology Informatics Unit, Tokyo, Japan.


Human versus Artificial intelligence (AI) matches have drawn many people's attention. In the 1990s, the IBM chess-playing supercomputer ‘Deep Blue’ had a close match against the human champion Garry Kasparov, and eventually emerged victorious (1). IBM’s ‘Watson’ challenged the popular American television quiz show Jeopardy! in 2011, defeated the human quiz show champion and won a prize of 1 million dollars(2). In 2016, the Google DeepMind AI ‘AlphaGo’, which was equipped with deep learning, won a go match against the professional go player (3).Go has been recognised as the most complex game for machines, and it should have taken several more decades before a computer should have defeated a human in a go match. AlphaGo perceivedeach position of gogame as ‘images’, and deduced the move that a human professional go player will likely make by learning from the enormous records of past human moves. Furthermore, by combining with reinforcement learning method, Monte Carlo method and precise engineering technique, a computer defeated a human in a go matchfinally. In the article published in Nature (3), AlphaGo developers noted that ‘AlphaGo evaluated thousandsof times fewer positions than Deep Blue did in its chess match againstKasparov’and noted that ‘an approach that is perhaps closer to howhumans play’.

AI is an indispensable part of our everyday lives; it has roles in searching the internet, filtering junk mail and trading in the financial market. And now, the competition to apply AI in the medical field has been also intensifying worldwide.

Application of AIin the medical field

The definition of AI is not set in stone, even among experts. Among a lot of definitions, machine learning is the most important element in understanding current AI. Arthur Samuel explains machine learning as the ‘field of study that gives computers the ability to learn without being explicitly programmed’(4).Deep learning is a form of machine learning, which has the stacking of networks for learning.

The recent progress in image analysis using machine learning is remarkable.In medical fields, there are many types of images: X-ray image, CT image, MRI image and Pathology image, etc. Among these, AI Analysis of Pathology images is one of the most challenging tasks because of the huge data size, several GB (or more)per an image, and the complexity.The following is one of the morphometric approaches for analysis of pathological images by using AI. First, to get the computer to recognise images, whole slide images (WSI) were created using WSI scanner. Thereafter, all the target cells’ nuclei on the slides were analysed and classified using machine learning and statistical methods(5, 6, 7).This is easy to imagine when thought of as a computer analysing the ‘cell face’. Figure 1depicts an example of such analysis results, with tall red cells identified by AI as cells that are highly likely to be near malignant lesions and low blue cells identified as cells that are likely to be benign tissue. Using this method, histological types of breast tumors could be classified with 90.9%accuracy only using subtle morphological differences of microenvironmental myoepithelial cell nuclei without any direct information about cancer cells (6). Furthermore, through analysing the typical cells which were selected by computers, we succeeded in developing a new biological mechanism of DCIS progressing to invasive cancer (6).

Learning and optimization

Parameter optimization is one of the most essential factors in machine learning. The underlying way of thinking in data science, including machine learning, is that the future can be predicted from past data by making a model of data and optimising parameters. Extracting and analysing past data can be achieved through standard operations; however, to predict the future, scientific operations of the establishment and verification of hypothesis is necessary. Based on limited data, it is impossible to predict the observed value subsequently obtained with 100 % accuracy. By making the best prediction based on evidence, it is a method of thinking that can be useful in some circumstances.Analysis is essentially performed in accordance with the following steps:(i) Configure a model including parameters. (ii) Configure standards to evaluate parameters. (iii) Find the most suitable parameter. This can be thought of as the most basic neuron model.The human brain is composed of many neural networks. When focusing on a single neuron, the model is based on the supposition that electrical signals received from other neurons via synapses are transmitted through dendrites and that when the threshold is exceeded, it passes through the axons and is ultimately transmitted to a different neuron via synapses. This model determines whether there is output to adjacent neurons in response to abundant input; the model can be referred to as a binary classification model.A diagram of the model is shown in Figure 2. In the example, there are four inputs, each of which has weight as a parameter. When the total of these inputs reaches a certain level, it exceeds the threshold and is calculated as the output. When the activation function is expressed as f, we obtain the mathematical formula:

If the optimal value of these parameters can be found using past data, we can predict and classify the data using this simplest model.

Mathematical modelling in the field of cancer research

In mathematical modelling, parameter optimization is also an essential factor. Mathematical models of cancer have evolved through accumulated knowledge in molecular biology. The multistage model by Armitage and Doll explained that the cancer age-incidence curve on a double logarithmic scale increases in an approximately linear fashion (8). Frank explained age-specific acceleration of the age-incidence curve by multiple rounds of clonal expansion (9). The Moran process is used to explain the dynamics of cancer initiation as an evolutionary theory, where random selection of cell death and survival occurs in a limited population (10).

The emergence of immune checkpoint inhibitors has drawn considerable attention to the influence of immune cells in cancer treatment. Through the analysis of cultured cells and animal experiments, underlying mechanisms have been successively elucidated; however, the detailed analysis of such effects in human body must be analysed on a histological level, and mathematical analysis of pathology slides serves as an important tool. In case of a temporary elevation in prostate-specific antigen (PSA) levels, PSA bounce, arising following brachytherapy for prostate cancer, we succeeded in mathematically and pathologically demonstrating the interaction between cancer cells and immune cells in human body (11). Several reports have indicated that a very interesting point of PSA bounce is that it even occurs without signs of prostatitis and urinary tract infections and that the prognosis of patients with PSA bounce is often good. We created a mathematical model of the number of cancer cells, number of immune cells and levels of serum PSA.




Equation [1] represents cancer cell dynamics. Equation [2] represents immune cell dynamics, such as CD8+ lymphocytes andEquation [3] represents PSA dynamics (for more detailed information, please see reference 11).Then,we performed parameter fitting toindividuals who underwent brachytherapy alone.We found that PSA bounce following brachytherapy can be mathematically explained based upon the interaction between cancer cells and immune cells. On comparing the parameters of each individual, the simulation revealed a difference between patients with PSA bounce and those without PSA bounce based on the number intratumoural immune cells at the start of treatment. Furthermore, measuring the number of intratumoural lymphocytes in biopsy slides, we found that individuals with PSA bounce had significantly more intratumoural CD8+ lymphocytes (p < 0.05) (11). At present, based on these findings, we are proceeding with a project to perform analysis using AI including deep learning method on a large cohort of prostate cancer patients.



The purpose of medicine is to helppatients. If prognosis and therapeutic effects could be predicted with greater accuracy using AI and mathematical analysis than the current medical systems, then patients could get the benefits such as reduced side effects and curbed medical costs. Furthermore, the combination of AI technologiesand medical big data is useful to discover novel diseasemechanisms and therapies (Fig.3).Comprehensive analysis of clinical data enablesdata aggregation and thus opportunities for exploratory analysis whichlead to improved knowledge about patient outcome and survival.

In the medical filed, AI, in conjunction with medical professionals, has started to help people. Presently, it cannot be said that machine learning technology, including deep learning, is at a state where it can be fully applied in the medical fields. In this day and age, it is important to increasing partnerships between medical professionals and AI researchers for finding ways that would benefit patients.




Fig. 1:Cell classification using AI

Cells were classified by AI into lowblue cells (cells highly likely to be benign tissues) and tall red cells (highly likely to be near malignant lesions). (excerpt from reference 5).


Fig. 2: A mathematical model of a neuron



Fig. 3:AI and medicine

The combination of AI technologies and medical big data is useful to discover novel disease mechanisms/therapies, and useful to select theoptimal treatment for each patient.(excerpt from reference 5).



1) Deep Blue. IBM Research. Retrieved 2007-04-20.

2) http://www.ibm.com/watson/

3) Silver D, Huang A, Maddison CJ, et al. Mastering the game of Go with deep neural networks and tree search. Nature. 529(7587):484-9. (2016)

4) Simon P. Too Big to Ignore: The Business Case for Big Data: Wiley; (2013).

5) Research Spotlight (Yamamoto Y), Archives of Scientific Support Programs for Cancer Research, 6,51-62, (2016).


6) Yamamoto Y, Saito A, Tateishi A, Shimojo H, Kanno H, Tsuchiya S, Ito KI, Cosatto E, Graf HP, Moraleda RR, Eils R, and Grabe N.:Quantitative diagnosis of breast tumors by morphometric classification of microenvironmental myoepithelial cells using a machine learning approach. Sci Rep. 7, 46732 (2017).

7) Saito A, Numata Y, Hamada T, Horisawa T, Cosatto E, Graf HP, Kuroda M, and Yamamoto Y.:A novel method for morphological pleomorphism and heterogeneity quantitative measurement: Named cell feature level co-occurrence matrix. J Pathol Inform. 7, 36 (2016).

8) Armitage P, Doll R.: The age distribution of cancer and a multi-stage theory of carcinogenesis. Br J Cancer 8: 1-12 (1954).

9) Frank SA.: Age-specific acceleration of cancer. Curr Biol 14: 242-246(2004).

10) Nowak MA.: Evolutionary Dynamics: Exploring the Equations of Life. Cambridge, MA: Harvard University Press: 93–105 and 209–248(2006).

11) Yamamoto Y, Offord CP, Kimura G, et al. Tumour and immune cell dynamics explain the PSA bounce after prostate cancer brachytherapy. Br J Cancer. 115(2):195-202 (2016).

Chapter Highlights

  • Australian Chapter:- Dr Pam McGrath and her team, International Program of Psycho-Social Health Research (IPP-SHR), have worked closely with IRPC on a range of research projects.
  • Australian Chapter:- Hamish Holewa, IPP-SHR has contributed to the development of IRPC's journal, the Austral-Asian Journal of Cancer, a HERDC recognised, peer-reviewed, multi-disciplinary cancer journal.

Latest News

  • Guide for students living with chronic health conditions
    Managing a chronic illness is challenging under the best of circumstances, but add the stresses of college, and things can quickly spin out of control. This guide touches on a number of pertinent points, including how to balance studies with disease management. Read more
  • Quadrant - Project Management Software for Researchers
    Quadrant is proud to be a supporter and collaborator of the Austral-Asian Journal of Cancer and International Research Promotion Council.
    Quadrant is a cloud-based project management and data collection software tool that allows participant-based researchers to work collaboratively and efficiently from a self-managed centralised site. Quadrant addresses the need for secure and ethical online research project management and data storage.
    Read More
  • New Room provides an interactive forum for all the scientists and researchers to convey important information and news.


Dr. Hillary Klonoff-Cohen


Dr. Hillary Klonoff-Cohen presently serves as the Saul J. Morse & Anne B. Morgan Professor in Applied Health Sciences, Associate Director of Community Health, and the Director of the MPH and PhD Programs in Community Health at the University of Illinois at Urbana-Champaign.

Prof. Hiroaki Honda


Dr. Yoichiro Yamamoto –an eminent scientist and an expert in the field ofmedicineand computer science/mathematics especially including Artificial Intelligence (AI) is presently working as a unit leader of Pathology Informatics Unit at RIKEN Center for Advanced Intelligence Project in Japan.

Dr Guillaume Moulis, MD, PhD


Dr Guillaume Moulis, MD, PhD is currently Assistant Professor in Internal Medicine at Toulouse University Hospital, France. He is also researcher in Pharmacoepidemiology at UMR 1027 Inserm-University of Toulouse and at the 1436 clinical investigation center, Toulouse, France. He is born on 23rd September 1983 in Toulouse, France.