THE UNITED REPUBLIC OF TANZANIA INFORMATION AND COMMUNICATION TECHNOLOGIES COMMISSION THE 2ND ANNUAL ICT PROFESSIONALS CONFERENCE 2018 (AIPC-18) Machine.

Slides:



Advertisements
Similar presentations
Clinical decision support system (CDSS). Knowledge-based systems Knowledge based systems are artificial intelligent tools working in a narrow domain to.
Advertisements

Data Mining Glen Shih CS157B Section 1 Dr. Sin-Min Lee April 4, 2006.
Personal Digital Assistants: Revolutionizing Medical Care Lauren McKenna and Dallas Warren ORF/PSY 322- H/M Interactions May 5 th, 2005.
Introduction to Machine Learning Anjeli Singh Computer Science and Software Engineering April 28 th 2008.
© Prentice Hall1 DATA MINING TECHNIQUES Introductory and Advanced Topics Eamonn Keogh (some slides adapted from) Margaret Dunham Dr. M.H.Dunham, Data Mining,
McGraw-Hill/Irwin ©2005 The McGraw-Hill Companies, All rights reserved ©2005 The McGraw-Hill Companies, All rights reserved McGraw-Hill/Irwin.
An expert system is a package that holds a body of knowledge and a set of rules on a subject that has been gained from human experts. An expert system.
Data Mining: A Closer Look
Computer Science AND DOCTORS Jolena Co Truong- 6 th period.
Data Mining By Andrie Suherman. Agenda Introduction Major Elements Steps/ Processes Tools used for data mining Advantages and Disadvantages.
LLNL-PRES This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.
1 © Goharian & Grossman 2003 Introduction to Data Mining (CS 422) Fall 2010.
METU Informatics Institute Min720 Pattern Classification with Bio-Medical Applications Lecture Notes by Neşe Yalabık Spring 2011.
Data Mining. 2 Models Created by Data Mining Linear Equations Rules Clusters Graphs Tree Structures Recurrent Patterns.
Data Mining Chun-Hung Chou
Data Mining Chapter 1 Introduction -- Basic Data Mining Tasks -- Related Concepts -- Data Mining Techniques.
Copyright © 2004 Pearson Education, Inc.. Chapter 27 Data Mining Concepts.
Advanced Analytics on Hadoop Spring 2014 WPI, Mohamed Eltabakh 1.
Data Mining BY JEMINI ISLAM. Data Mining Outline: What is data mining? Why use data mining? How does data mining work The process of data mining Tools.
Computational Approaches for Biomarker Discovery SubbaLakshmiswetha Patchamatla.
Data Mining and Decision Support
Chong Ho Yu.  Data mining (DM) is a cluster of techniques, including decision trees, artificial neural networks, and clustering, which has been employed.
CISC 849 : Applications in Fintech Namami Shukla Dept of Computer & Information Sciences University of Delaware iCARE : A Framework for Big Data Based.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
WHAT IS DATA MINING?  The process of automatically extracting useful information from large amounts of data.  Uses traditional data analysis techniques.
WHAT IS DATA MINING?  The process of automatically extracting useful information from large amounts of data.  Uses traditional data analysis techniques.
Big Data Quality Panel Norman Paton University of Manchester.
EuResist? EuResist is an international project designed to improve the treatment of HIV patients by developing a computerized system that can recommend.
Network Management Lecture 13. MACHINE LEARNING TECHNIQUES 2 Dr. Atiq Ahmed Université de Balouchistan.
Sparse Coding: A Deep Learning using Unlabeled Data for High - Level Representation Dr.G.M.Nasira R. Vidya R. P. Jaia Priyankka.
Introduction to Machine Learning, its potential usage in network area,
Brief Intro to Machine Learning CS539
CSE 4705 Artificial Intelligence
P.Demestichas (1), S. Vassaki(2,3), A.Georgakopoulos(2,3)
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Machine Learning for Computer Security
Information Organization: Overview
Applying Deep Neural Network to Enhance EMPI Searching
ANOMALY DETECTION FRAMEWORK FOR BIG DATA
Machine Learning overview Chapter 18, 21
Machine Learning overview Chapter 18, 21
DATA MINING © Prentice Hall.
An Artificial Intelligence Approach to Precision Oncology
Eick: Introduction Machine Learning
School of Computer Science & Engineering
It’s All About Me From Big Data Models to Personalized Experience
Introduction to Clinical Pharmacy
Data Mining 101 with Scikit-Learn
Introduction C.Eng 714 Spring 2010.
Reading: Pedro Domingos: A Few Useful Things to Know about Machine Learning source: /cacm12.pdf reading.
CH. 1: Introduction 1.1 What is Machine Learning Example:
Component 11/Unit 7 Implementing Clinical Decision Support
Fenglong Ma1, Jing Gao1, Qiuling Suo1
Luke Do, Jessica Olmedo, Arely Romero, and Vianca Santana
Overview of Machine Learning
INTRODUCTION.
iSRD Spam Review Detection with Imbalanced Data Distributions
3.1.1 Introduction to Machine Learning
Text Analytics and Machine Learning Workshop
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Data Mining, Machine Learning, Data Analysis, etc. scikit-learn
Scalable and accurate deep learning with electronic health
Christoph F. Eick: A Gentle Introduction to Machine Learning
Big DATA.
Information Organization: Overview
Simplifying Healthcare
Machine Learning for Space Systems: Are We Ready?
Machine Learning overview Chapter 18, 21
Machine Learning.
Getting Started with Microsoft Azure Machine Learning
Presentation transcript:

THE UNITED REPUBLIC OF TANZANIA INFORMATION AND COMMUNICATION TECHNOLOGIES COMMISSION THE 2ND ANNUAL ICT PROFESSIONALS CONFERENCE 2018 (AIPC-18) Machine Learning (ML) for Healthcare: Opportunities By: Ireneus Kagashe, Ph.D. Date: 24th – 26th October 2018 AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques in healthcare. AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques. AIPC-18

What is Machine Learning? Machine learning (ML) is one of the techniques (algorithms) used in data mining. Data mining (DM) – the process of discovering interesting patterns and knowledge from large amounts of data. ML algorithms basically use computer applications to automatically learn and recognize complex patterns from data and make intelligent decisions on other sets of data based on the learned patterns. AIPC-18

What is ML?.... ML should be used to solve problems when: No human experts, so data cannot be labeled or categorized; There are human experts, but it is very hard to program; There are human experts and it can be programmed, but where it is not cost effective to implement. Usually caused by the following : Huge volumes of data to be analyzed Data has lots of features/variables Limited prior knowledge of the data ML can be categorized into: Supervised learning Unsupervised learning Semi-supervised learning AIPC-18

Categorization of ML: Supervised Learning You have input data or variables (x) and their corresponding label or output variable (Y) and you use an algorithm to learn the mapping function (model) from the input to the output. Y = f(x) The model is then used to predict output of new data. Predicting the future based on the past. Training Data learn model past predict future New AIPC-18

Categorization of ML: Supervised Learning … Supervised ML problems: Classification: The output variable (label) is a category, such as “red” or “blue”; or “malaria” and “no malaria”. Regression: The output variable is a real value, such as “Shillings” or “score”. Classification or Regression problems intend to find labels for your data Common supervised ML algorithms: Linear regression for regression problems. Random forest for classification and regression problems. Support vector machines for classification problems. Naive Bayes for classification problems. Artificial Neural Networks and Deep Learning for classification problems AIPC-18

Categorization of ML: Unsupervised Learning There is only input data (X) and no corresponding output variables. Algorithms are given plenty of leeways to learn and present the interesting structure in the data. AIPC-18

Categorization of ML: Unsupervised …. Unsupervised learning problems: Clustering: Identifying similar groups of data in your data set. Or separate groups having similar characteristics and assign them into clusters. E.g. grouping patients by symptoms, grouping customers by purchasing behavior. Association: Discover rules that describe large portions of your data. E.g. in a symptoms dataset, possible associations could be such as people with runny noses also tend to have a fever. Clustering or association problems intend to automatically group similar data together. Some popular examples of unsupervised learning algorithms are: k-means for clustering problems. Apriori algorithm for association rule learning problems. AIPC-18

Categorization of ML: Semi-supervised You have a large amount of input data (X) and only some of the data is labeled (Y). You use supervised learning techniques to train the model and make best guess predictions for the unlabeled input data. Then concatenate the newly labeled data with the initially labeled data and then feed that data back into the supervised learning algorithm as training data to create a new model that will be used to make predictions on new unseen data. Why this method? It can be laborious to label data as it may require domain experts. Whereas unlabeled data is cheap and easy to collect and store. AIPC-18

Performance Evaluation of ML Prediction Models When we apply ML algorithms, we expect them to make accurate predictions. But: How certain are we that the fitted/learned model can generalize well to unseen data? How do we select a good predictor/model? Can there be a different learning algorithm better-suited for prediction? Solution: Performance evaluation of ML prediction models. To evaluate the performance of the prediction models we need to have training data (labeled/annotated) and test data (labeled); Usually in the ratio of 70% training and 30% test. AIPC-18

Performance Evaluation of ML Prediction Models… Common evaluation metrics: Accuracy = TP+TN/(TP+FP+FN+TN) Precision = TP/(TP+FP) Recall = TP/(TP+FN) F1 Score = 2*(Recall * Precision) / (Recall + Precision) AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques in healthcare. AIPC-18

ML opportunities for healthcare In Tanzania, at present, most health centers (hospitals) use hospital management system (HMS) software. HMS are integrated healthcare solutions that include Outpatient and Inpatient Management, Pharmacy, Laboratory, Radiology, Ward Management, Mobile Application, Online Appointments Scheduling, Doctor Portal, Medical Electronic Billing, Accounting etc. These advancements in electronic health records (EHRs) are remarkable, however, the information they provide is not much better than the old paper-based records they replaced. On the other hand, it is also evident that people discuss health-related conditions and statuses (theirs or others) on online platforms (social media, online health communities[OHCs] etc.) AIPC-18

ML opportunities for healthcare… These platforms generate huge amounts of data that can provide useful population health information. Challenge: Vast amounts of data require automation for analysis. ML can turn complex health-related data (e.g., patient health records, genomic data, data from wearable health monitors/sensors, online health communities, social media, reviews of physicians, medical imagery, etc.) into actionable knowledge that improves healthcare. AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques in healthcare. AIPC-18

ML opportunities for researchers Researchers in this field are expected to: Use new/improved ML methods to provide solutions to new healthcare problems Use new ML methods to provide solutions to old healthcare problems Use existing ML methods to provide solutions to new healthcare problems Use existing ML methods to provide solutions to old healthcare problems AIPC-18

ML opportunities for researchers Researchers can conduct their research in the following areas (not limited though): Preventive analytics and Preventive measures Disease identification and diagnosis of ailments – using data on chief complaint, past hospital visits, lab and radiology results, and other factors to enhance accuracy in disease predictions. For example, which features or algorithms to use to accomplish that? Personalized treatment – how can we use patient medical information and history to optimize the selection of treatment options? AIPC-18

ML opportunities for researchers OHCs (Social Media) Analytics Note: OHCs share their symptoms, concerns, experiences with treatment and healing stories about thousands of conditions. Opportunities: Adverse drug reaction (ADR) detection Adverse drug-drug interaction effect Identification of public opinion or hot discussions on health policies/regulations/disease/drugs etc – sentiment and topic analysis Disease outbreak or trend prediction (e.g. seasonal influenza prediction) AIPC-18

ML opportunities for researchers… Examples of Research Questions: What data analytic methods may help generate reliable early warnings from the diverse sources of noisy data while limiting false alarms? How could diverse sources of data be combined with patient history in electronic health records to improve the quality of early warnings for each patient? What computational challenges may be encountered in real-time processing and convergence of various data sources (e.g., sensors, EHR, genetics, OHC)? What data analytic methods or features can we use/apply to accurately model individuals and tailor behavioral mechanisms or predict diseases? What methods can reduce the amount of data needed to reliably identify the risk factors or predict diseases? How can we effectively de-identify electronic health records to maintain data privacy of individuals whose data is used in research? AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques in healthcare. AIPC-18

ML opportunities for developers Hospital Management System Software There is a need to incorporate disease prediction functionality based on the history of present illness (HPI) or chief complaint and other relevant data. This can reduce “wasteful” procedures  minimize healthcare cost and diagnosis time. AIPC-18

ML opportunities for developers … Online Health Communities (OHCs) Applications Development of web apps for online health communities (OHCs). OHCs share their symptoms, concerns, experiences with treatment and healing stories about thousands of conditions. Examples: DailyStrengths (www.dailystrength.org) PatientsLikeMe (www.patientslikeme.com) JF Doctor (https://www.jamiiforums.com/forums/jf-doctor.61/) ML can be used to recommend posts or threads or answers similar to your shared experiences, questions, and concerns. No need for human moderators for that AIPC-18

ML opportunities for developers… AIPC-18

Outline Introduction to Machine Learning ML opportunities for healthcare researchers information systems developers Challenges of employing machine learning techniques in healthcare. AIPC-18

Challenges Staff unwilling to change their practices – There may be physicians who fear that machine learning is the beginning of a process that could render them obsolete. Solution: ML should augment healthcare providers’ efforts, educate them. Real-world health data are messy, incomplete, and come in a variety of formats. E.g. “&&&&hey ………….. #nuff rizpect cholera!”. Data preprocessing is not a walk in the park. Patient privacy: who has access to the data? who will own the data (and findings)? How will the data be protected? Who should report the findings? Language (English vs Swahili) – some natural language processing tools only can accommodate the English language. This is, on the other hand, an opportunity for NLP researchers to accommodate Swahili. Not all problems will be efficiently solved using ML: Some healthcare problems and research questions don’t actually require ML, but can the just be solved using e.g. simple codes or structured query language (SQL). E.g. causal inferences AIPC-18

Discussion AIPC-18