Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Computational Psychiatry Research Mapping Project

CPSYMAP
June 14, 2021

Computational Psychiatry Research Mapping Project

This slide explains the concept and the future goal of the new database, Computational Psychiatry Research Map (CPSYMAP): https://ncnp-cpsy-rmap.web.app/
This is a tool for visualizing research papers from computational psychiatry as a two-dimensional "map". Explore the distribution of papers along neuroscientific, psychiatric and computational dimensions using this map!

We had a major update June 2021! The updated contents are described in slide no. 9~14.

・Implementation of a crawler to semi-automatically collect computational psychiatry articles

・Automatically tag articles based on keywords in neuroscience, psychiatry, and mathematical modeling

・Add a tag exclusion function that prevents certain tags from being displayed on the map

・Separate display of papers with semi-automatic registration and manually registered by humans

・Twitter bot to automatically tweet when a paper is manually registered

CPSYMAP

June 14, 2021
Tweet

More Decks by CPSYMAP

Other Decks in Research

Transcript

  1. Computational Psychiatry Research Mapping Project 1 version 2020.06.01 Ayaka Kato,

    Yuichi Yamashita Yoshihiko Kunisato, Tsukasa Okimura, Kentaro Katahira Computational Psychiatry Colloquium
  2. 2 version 2020.06.01 What is Computational Psychiatry? Applying mathematical and

    theoretical methods to psychiatric research to advance knowledge in the field Enhancing our understanding of psychiatric disorders by computational modelling of information processing in the brain Application of machine learning techniques to large scale data on mental disorders to develop a variety of computational models - i.e. discriminative modelling
  3. 3 However, insufficient connections between the perspectives of neuroscience, psychiatry,

    and computational models may hinder the extension of research For the further development of the field, it is necessary to organize individual studies from the following perspectives and make it possible to have a bird's eye view. version 2020.06.01 1. Neuroscience What cognitive functions and units of analysis, including behavioral and neurocircuitry, are targeted? 2. Psychiatry What mental disorders and symptoms are being addressed and what is not being addressed? 3. Computation What methodologies, data, theories, and models are used?
  4. • Tag computational psychiatry papers by using terms “Neuroscience”, “Psychiatry”

    and “Computation” • Organize and visualize the status of research areas along the tags on a two-dimensional map (color shading indicates areas with many papers and areas with few papers) • Create an environment where anyone can search, register, and overview computational psychiatry (CPSY) research as a web application We've developed a database to solve the problem! 4 version 2020.06.01
  5. 6 You can choose the vertical and horizontal axes you

    wish to view takes you to this page version 2020.06.01
  6. 7 View and sort papers in specific cells of the

    map version 2020.06.01 You can use the filter function to narrow down the contents map
  7. 8 You can specify the papers you want to add

    and easily tag them version 2020.06.01
  8. 9 New: June 2021 major update version 2020.06.01 • Implementation

    of a crawler to semi-automatically collect computational psychiatry articles • Automatically tag articles based on keywords in neuroscience, psychiatry, and mathematical modeling • Add a tag exclusion function that prevents certain tags from being displayed on the map • Separate display of papers with semi-automatic registration and manually registered by humans • Twitter bot to automatically tweet when a paper is manually registered
  9. 10 Implementation of a crawler version 2020.06.01 Automatically collect computational

    psychiatry articles from Pubmed using following queries 1. Combination of the DSM-5 categories and name of the model 例)Neurodevelopmental Disorders [MeSH Terms] x "Reinforcement learning” →Following two tags are assigned to the paper 2. Papers searched by ”Computational psychiatry” * MeSH Term: the abbreviation of the Medical Subjest Headings, the organized heading of the medical term used in Pubmed
  10. 11 version 2020.06.01 • Making the list of the keywords

    and assign tags to the paper with similar words with the keywords Example) Abstract …this functional impairment of the inferior frontal gyrus in those at genetic risk of bipolar disorder reflects the dysfunction of broader network dynamics underlying the coordination of emotion perception and cognitive control. Breakspear, Michael, et al. "Network dysfunction of emotional and cognitive processes in those at genetic risk of bipolar disorder." Brain 138.11 (2015): 3427-3439. Corresponding keywords: Gene, genetic Automatically tag articles based on keywords
  11. 12 Add a tag exclusion function version 2020.06.01 As default

    settings for exclusion function, papers with an “unspecified” tag to the used tags for display was selected.
  12. 14 Twitter bot to automatically tweet when a paper is

    manually registered version 2020.06.01
  13. 15 1) A researcher beginning research on computational psychiatry Be

    able to outline the area in which you are about to begin your research, sort out where your niche is, and develop a research plan We welcome all researchers to use this database 2) A researcher who has already published a paper on computational psychiatry Register your or your collaborators‘ research results to make them more accessible to other researchers and potentially increase the number of citations! version 2020.06.01
  14. 16 The broad range of axes that can be visualized

    version 2020.06.01 1. Neuroscience What cognitive functions and units of analysis, including behavioral and neurocircuitry, are targeted? 2. Psychiatry What mental disorders and symptoms are being addressed and what is not being addressed? 3. Computation What methodologies, data, theories, and models are used?
  15. Research Domain Criteria (RDoC) A novel framework for psychiatric research

    based on the findings of behavioral neuroscience, without use of conventional disease categories (proposed by the National Institute of Mental Health: NIMH) https://www.nimh.nih.gov/research-priorities/rdoc/research-domain-criteria-matrix.shtml version 2020.06.01 17 1. Connections to neuroscience
  16. Rows: constructs and domains • Observable basic component functions based

    on behavioral neuroscience are defined as constructs • Psychiatric disorders are considered on a spectrum from normal to abnormal states of constructs • Organize related constructs into a domain RDoC Matrix version 2020.06.01 18
  17. RDoC Matrix version 2020.06.01 19 Column: Unit of analysis Enumerate

    the variables of the unit of analysis from micro to macroscopic levels • Consists of Genes, Molecules, Cells, Neural circuits, Physiology, Behaviors, Self-reports, Paradigms
  18. Categorical strategy vs. dimensional strategy Expectations and concerns about RDoC

    Categorical strategy Dimensional strategy (RDoC) • Reliability of diagnosis accumulated by DSM etc Expectations • Use of the accumulated neuroscience knowledge • Less biological plausibility • Less predictive validity Concerns • Not fully capturing the symptoms of mental disorders • Less applicability to clinical practice version 2020.06.01 20
  19. The RDoC has been proposed as a novel framework for

    psychiatric research based on the findings of behavioral neuroscience without the use of conventional disease categories. (proposed by NIMH) However, in many CPSY papers, the correspondence between the contents of the paper and RDoC is not clarified 21 version 2020.06.01 It is necessary to address which construct and what kind of units of analysis, such as cognitive functions or behavior and circuits, are targeted in existing CPSY research through RDoC 1.Connections to neuroscience summary In our database, we have
  20. 22 • RDoC does not use the traditional taxonomy and

    symptomatology of mental disorders • It is also difficult for researchers working on CPSY (especially those who do not specialize in psychiatry) to see what symptoms are important and what should be studied from a psychiatric symptomatology and psychopathology perspective 2. Connections to psychiatric disorder category and symptoms version 2020.06.01
  21. 23 version 2020.06.01 Examples of symptoms of mental disorders that

    are not captured by RDoC Symptoms Main complaint Related mental disorders RDoC Obsessive thought Can't help thinking of irrational ideas, words, etc. against their will Obsessive-compulsive disorders No appropriate construct Delusion Delusional perception, delusion of pursuit, persecutory delusions, etc. Schizophrenia, Bipolar disorders, etc. No appropriate construct
  22. • RDoC cannot cover the traditional taxonomy and symptomatology of

    psychiatric disorders • It is difficult to identify which symptoms are important and what should be studied 24 version 2020.06.01 2. Connections to psychiatric disorder category and symptoms: Summary It is necessary to organize what mental disorders, symptoms, and psychopathological problems are being dealt with and what is not being dealt with in CPSY research In our database, we have these categorize DSM: correspondence with the disease category Symptomatology: Symptoms (anxiety, depression, hallucinations/delusions, etc...)
  23. In CPSY research, various mathematical theories, techniques, and strategies are

    used For example, Theory-driven approach vs. Data-driven approach. Model Fitting vs. Simulated lesion. There are different "cultures" of research strategies, but it is generally difficult to grasp the whole picture. 25 version 2020.06.01 3. Connections with mathematical modelling Kunisato, Y., Katahira, K., Okimura, T., & Yamashita, Y. (2019). Keisanron- teki Seishin Igaku (Computational Psychiatry). Tokyo: Keiso Shobo (in Japanese).
  24. Data-driven approach Discriminative functions/models Input: Neural activity and other biometric

    information Output: class label/probability Methodology: support vector machine, regression, etc. Example: discrimination of disease categories / Prediction of disease states f • Attempt to apply machine learning methods to large-scale data on mental disorders, for data clustering, and the development of discriminatory models version 2020.06.01 26
  25. Perception, Recognition and Decision Making Sensory input Actions/ Behavior A

    "theory-driven approach" using generative models • Mathematical modeling information processing (perception and cognition) in the brain as a “computation” version 2020.06.01 27
  26. 28 version 2020.06.01 3.Connections with Mathematical Modelling: Summary • In

    CPSY research various mathematical theories, techniques, and strategies are used. • There are multiple approaches ("cultures") of research strategy that differ, but it is generally difficult to grasp the whole picture It is necessary to organize what methodologies, data, theories, and models are used In the database, we have Data type: Human data, Simulation etc… Experimental design: Generative model, Model fitting etc… Models: Reinforcement learning, Neural network etc…
  27. • Assist in creating links between CPSY, traditional psychiatry and

    neuroscience. • Provide alternative perspectives for non-specialists about what topics need to be explored in CPSY and which are important in this field • Encourage experts in fields of informatics, physics, and engineering, including experts in psychiatry and neuroscience, to become more involved in CPSY research • Contribute to the integration between different methods, especially the interactive and cyclical development of theory-driven and data-driven approaches • By building a platform for organizing and storing information, we can contribute to the efficiency of research and the vitalization of the field The Future of Databases 29 version 2020.06.01
  28. 30 Ackowledgements • Mr. Yuji Kawase (System Development Engineer) •

    Mr. Chris Salzberg (System Development Advisor) • We thank to Hiroshi Yamakawa (WBAI) for stimulating discussions. • This project was partly supported by JST CREST JPMJCR16E2, JSPS KAKENHI JP18KT0021, and JP19H04998. version 2020.06.01 Organization : Computational Psychiatry Colloquium in Japan A volunteer group for the study of computational psychiatry.