Feb 24 PM 13:10 – 15:10 (WS01-WS05)
Data Mining Clinical Notes
Christina Chen/Satoru Hashimoto/Naoaki Ichihara
There are valuable insights that can be gained from information hidden in free text clinical notes. In this workshop, participants will be introduced to the process of extracting data from free text. The first half of this session will consist of leveraging regular expressions (RegEx) to analyze formatted clinical notes. The second part will introduce a program called cTakes (visual natural language processing system) and SNOMED CT (standardized collection of medical terms). Participants will have plenty of hands-on practice. No programming skills are required.
Tree-based Approaches for Prediction, with Applications to Healthcare Data
Alistair Johnson/Shingo Tsuji/Keith Boroevich
This interactive workshop will introduce participants to tree-based approaches for classification and regression tasks. The first hour will introduce the approaches: starting from the simplest model, a decision tree, and covering more sophisticated models such as bagging, boosting, random forests, and ending on gradient boosting. In the second hour participants will work through building predictive models on a wide variety of tasks including breast cancer detection and flower classification. Finally, in the last hour, participants will be given a dataset for a real-world clinical problem, and must use their newfound knowledge to improve upon the current state of the art. This workshop has the following requirements: laptop with Anaconda 5.0.1 (Python 3.6 version) installed (instructions will be provided by the Japanese office prior to the conference through mailing list), and basic understanding of Python and Jupyter notebook, which will be covered during the main programme (please attend the session if you are a new user).
Sparking Entrepreneurship in Healthcare & Biotechnology
Leo Celi/Jun Suto/Keiji Yano/Hironobu Matsushita/George Konno
This workshop will take a case-study approach to exploring the often the under-appreciated non-technical barriers to innovation. Examples will be drawn from personal experience in building hubs of entrepreneurship and innovation, with an emphasis on the MIT ecosystem. The target audience of this talk includes prospective and current entrepreneurs, professors and industry professionals interested in increasing innovation or entrepreneurship in their organizations, and government officials with an interest in building innovation hubs. No programming skills are required.
Physiological Signal Processing
Chen Xie/Ryo Uchimido/Artem Lysenko
This workshop explores a subset of the 2015 Physionet/Computing in Cardiology Challenge: "Reducing False Alarms in the ICU". The target arrhythmia type to be investigated is ventricular tachycardia. In the first two hours, methods to preprocess, derive features from, and combine information from multichannel ECG and blood-pressure waveforms will be shown. Example features include spectral components, signal peak shapes and distributions, and signal quality indices. Finally, the combined features will be used to train and test supervised classification methods to identify true and false alarms. In the final hour, participants will apply their own ideas to the challenge. The workshop will be carried out in Python/Jupyter notebook.
AI for medical image analysis
Mengling Feng/Kazuki Nishida
Artificial intelligence (AI) is the buzzword for all things relating to technology these days. In particular, healthcare is seen as an area in which AI may be gainfully deployed to improve medical care, especially with big data, exponential computing power and a burgeoning demand on healthcare systems due to aging populations. In this talk, I will share two use cases that my group has engaged recently.
Our use case focuses on the analysis of medical images, in particular, the mammogram images. My team recently participated the DREAM digital Mammogram challenged funded by the US white house fund and FDA. The aim was to improve the predictive accuracy of digital mammography for the early detection of breast cancer. The challenge was constructed around 640,000 mammogram images for over 80,000 patients. We developed a patch-based deep neural network structure to extract abnormal lesions and patterns from mammograms so to detect breast cancer. After months of training, the proposed AI model managed to achieve detection performance close to human experts. We are now working with the Breast Cancer Screening Center at National University Hospital aiming to develop and deploy an AI-aided tool to assist radiologists to achieve more accurate, consistent and faster breast cancer screening. Our workshop will share the fundamentals on how deep neural networks can be applied on medical image analysis, and I will also share the lessons learned via our mammogram project. Only pen and paper will be required.
Feb 24 PM 15:30 – 17:30 (WS06-WS11)
Applied Statistical Learning in Python
Calvin Chiew/Naoaki Ichihara/Keith Boroevich
This workshop aims to introduce clinicians to popular statistical methods used in machine learning, without delving into the underlying mathematical theory. We will focus on the random forest and support vector machine for classification, as well as general concepts of model fit and cross-validation. Other supervised and unsupervised learning techniques will be briefly mentioned. In the hands-on exercise, you will be asked to implement and evaluate these models on a clinical prediction problem. This workshop has the following requirements: laptop with Anaconda 5.0.1 (Python 3.6 version) installed (instructions will be provided by the Japanese office prior to the conference through mailing list), and basic understanding of Python and Jupyter notebook, which will be covered during the main programme (please attend the session if you are a new user).
Public Health for the Data Scientist
Raymond Francis Sarmiento//Kenji Wakabayashi
Tapping on large databases opens up unprecedented opportunities for improving the health of populations. Apart from improving the clinical diagnosis and treatment of individuals, new techniques in data science also have the potential to make huge impact in preventive medicine, systems improvement and global health. During this workshop, we will discuss several case studies which will highlight key concepts in public health with a side focus on epidemiological study design to robustly evaluate the impact of an intervention on population health. The target audience for this workshop is anyone who is interested in the intersection between public health and data science. The purpose of this workshop is to inspire a public health perspective to the work that participants are already engaged with. No programming skills are required.
Introduction to Clinical Medicine for the Data Scientist
Patrick Tyler/Hidenobu Shigemitsu
In her 2012 paper, “Machine Learning that Matters,” Kiri Wagstaff of the Jet Propulsion Laboratory (California Institute of Technology) writes that “much current ML research suffers from a growing detachment from… real problems,” instead perfecting computing techniques on idealized datasets that have limited real world relevance. As medicine advances into the 21st century, care providers and patients need solutions that go beyond deep analysis of the data; they need insights that can translate into better clinical decisions, increased efficiency, and improved outcomes for patients. The goal of this workshop is to give data scientists a view into the clinical world, with a focus on the clinical environment that produces the Medical Information Mart in Intensive Care (MIMIC) database at Beth Israel Deaconess Medical Center, Boston, MA (BIDMC). This database is built from over one decade of intensive care unit (ICU) visits at BIDMC, including medical, cardiac, surgical, and neurological ICUs. The workshop will include an overview of what conditions are commonly treated in the ICU, how patients come to receive care in the ICU, and what happens to patents after they leave the ICU. We will also review the ICU personnel, who enters data, how the user interface appears and works, and how those results map onto the data visible in the MIMIC database. At the conclusion of the workshop, you will better understand the clinical domain represented by the MIMIC dataset. And using MIMIC as a model, we hope to make the interactions between data scientists and front line clinicians more fluid. There will be many interactive activities as part of the session to improve learning, retention, and usefulness to the attendees. Only pen and paper will be required.
Translating a Clinical Question into a Big Data Study Design
Ryo Uchimido/Chen Xie/Artem Lysenko
Big data research is a multi-disciplinary collaborative task. In this workshop, participants will form teams consisting of data scientists and clinicians, and work together to answer the clinical question, "How can we predict whether a hypotensive episode will resolve with fluid administration?" using the MIMIC database. Requirements for this workshop is R/Rstudio installed.
Data analysis using JIPAD
Shawn Sturland/David Pilcher/Satoru Hashimoto/Atsushi Shiraishi
Clinician researchers from Japan, Australia and NZ will demonstrate an approach to analysis of critical care registry data for research, with live interactive analysis of de-identified JIPAD data done in real time and in collaboration with participants. For statistical analysis, STATA will be applied but basically no programming skills will be required.
Data for Improvement Workshop
Mataroria Lyndon/ Hideo Takahashi
Data is a critical element of improvement. This workshop will provide an introduction to improvement science in healthcare. By using case studies and team-based activities, participants will learn practical skills and techniques for quality improvement - systems thinking, measurement, analysis, and implementing change into practice. While this session is focused on healthcare, participants will gain the ability to formulate and create changes that can have a lasting impact in other fields and disciplines. No programming skills are required.