Research Outline

In response to the recent developments in cell and molecular imaging technology, bio-imaging and informatics research has emerged as a new field of bioinformatics. Although medical image processing is a much-needed technology, under the current situation, statistical mathematics has not fully permeated the medical field. Also, a deeper look at humanities and sociology amidst the ongoing recession has revealed the recent increase in individualization and isolation of the society, and rapid increase in the number of people requiring care (such as seniors living alone and people with serious stress).

In this project, image data obtained from 4 fields (medicine, biology, humanities, and sociology) are to be used to conduct research and development in the field of advanced analytical technology. In addition, the obtained results will be published as a generic image data analysis tool that can be applied to various problems, in order to contribute to academic discovery in each research field.

(Project Director: Tomoko Matsui - The Institute of Statistical Mathematics)

Purpose of the project

We aim to establish the fundamental technology for a multitude of decision support by visually expressing complex phenomena in cutting-edge scientific fields, and by conducting advanced analyses of the image data using a statistical method. The information regarding coding technology will be accumulated for image data, such as bio-imaging data (calcium imaging data, etc.), medical image data, and video data of people and objects, in order to research advanced analytical technologies for image data meant for general use.

Project Promotion System

The basic objectives of this project are as follows:

1. Handling the actual image data associated with medicine, humanities, and sociology.
2. Specific tasks are to be set for each subtheme. Working on these tasks would facilitate the accumulation of knowledge regarding coding technology for each subtheme. The image data analysis technology could then be researched.
3. Generalization of knowledge obtained as described in “2” and development of image data analysis technology

The research for subthemes 1, 2, and 4 is performed based on the following ideas:

Subtheme 1: We aim to build a generic methodology (in each appropriate field) by investigating a wide range of bio-imaging and informatics research topics, such as feature extraction, fractionation, registration, classification, and visualization of image information. The main focus was on the calcium imaging data of Caenorhabiditis elegans (nematode) nervous system.

Subtheme 2: Strong backgrounds resulting from principle problems in the measurement system and artifacts of stationary and nonstationary brain wave are to be superimposed on to the data obtained from the measurement of brain function. The appropriate data must be subjected to a considerable amount of pretreatment in order to adopt fundamental statistical methods, such as cross-correlation analysis and regression analysis. All processes, ranging from pretreatment to subsequent analyses, are rarely completed based on theory alone, as significant trial-and-error and empirical knowledge would be required. Therefore, the research will be coordinated with experimentalists, who would provide cooperation and feedback.

Subtheme 4: The developed and analyzed image data analysis technology is to be published as a generic tool that could be applied in a wide range of fields, such as medicine and human society, which handle image data.

Subtheme introduction

1. Bio-imaging informatics (Medicine and Biology)

We aim to develop a statistical method that can automatically detect the object position, count numbers, track objects, and conduct registration and image matching, using bio-imaging data mainly composed of 3D-videos, such as cell and molecular imaging. In addition, we will attempt to tackle a wide range of research topics in bio-imaging informatics, such as feature extraction, fractionation, registration, classification, and visualization of image information. We also aim to construct a generic methodology in the appropriate field.

(Principal Investigator: Akira Yoshida - The Institute of Statistical Mathematics)

2. Study of the operating principle of the brain, using optical imaging data (Medicine and Biology)

Micro-activities of nerve cells and astrocytes can be measured using calcium imaging data, while the macro-activities of the brain can be measured using near-infrared spectroscopy. Both techniques are classified as optical imaging methods, and have a common spatio-temporal structure. In this study, algorithms used to examine a background drift component exclusion method (an efficient and objective method for the detection of nerve activation) will be developed. In addition, we will attempt to estimate a space propagation mechanism for nerve activation.

(Principal Investigator: Fumikazu Miwakeichi - The Institute of Statistical Mathematics)

3. Study of statistical methods in medical image processing (Medicine and Biology)

The potential of the existing methods that are based on medical image processing, such as calculation anatomy for higher accuracy and increased speed of processing, led by statistical sciences such as sparse modeling, are to be examined for future medical applications.

(Principal Investigator: Shiro Ikeda - The Institute of Statistical Mathematics).

4. Visual information data analysis in humanities and sociology (Humanities and Sociology)

By detecting and predicting events and tracking people and objects from the video data of people and objects in humanities and sociology, we aim to develop a method to discover potentially important (dynamic) factors. In this study, we will focus on the study and development of a method that responds to generally unknown types and numbers beforehand (problems with unknown factors) and dynamic changes (problems with dynamic changes), using a specifically nonparametric and dynamic state space model. Here, we will examine information representation and identification technology for feature values that are most suitable for video analysis. In addition, by introducing a nonparametric Bayesian model into a state space model, we aim to address the problems of unknown factors and dynamic changes. The development of this method as a generic tool shall increase its applicability in the analysis of medicine- and biology-related data.

(Principal Investigator: Shinichi Sato - National Institute of Informatics/Tomoko Matsui, The Institute of Statistical Mathematics).