AI for Healthcare with Keras and Tensorflow 2.0: Design, Develop, and Deploy Machine Learning Models Using Healthcare Data

ISBN-13: 9781484270851 / Angielski / Miękka / 2021 / 381 str.

Anshik Bansal

Opis

Intermediate-Advanced user level

Spis treści

Chapter 1: Healthcare Market: A Primer

Chapter Goal: Know how sub-markets like pharmaceutical, medical

technology, and hospital come together to form the healthcare ecosystem. Learn on how digital and mobile are shaping and reforming traditional health. With technology available and permissible to large masses via internet things like telehealth have become a norm. Also, what kind of

problems are being solved at industry level and at various startups.

Sub Topics:

Healthcare Marketplace Overview

● Map of how different stakeholder comes together to form the system

● Medicare Overview

● Paying Doctors

● Healthcare Costs

Emerging Trends

● Changing role of consumer in healthcare

● Future of Healthcare Payments

● Quality of Healthcare Delivery

Industry 4.0 and Healthcare

Chapter 2: Multi Task Deep Learning To Predict Hospital

Re-admissions

Chapter Goal: A real world case study showing how re-admissions which

costs billions of dollars to the US healthcare system can be addressed. We will be using EHR data to cluster patients on their baseline characteristics and clinical factors and correlate with their readmission rates.

Sub Topics:

● Introduction to EHR data.

● Exploring MIMIC III datasets

● Establishing a baseline model to assess re-admission rates using

ensemble of classification models with handling class imbalance.

● Using auto-encoder to create a distributed representation of features.

● Clustering patients

● Analyzing readmission rate based on clusters.

● Comparative analysis between baseline and deep learning based

model.

Chapter 3: Predict Medical Billing Codes from Clinical Notes

Chapter Goal: Clinical notes contain information on prescribed procedures

and diagnosis from doctors and are used for accurate billings in the current medical system, but these are not readily available. One has to extract them manually for the process to be carried out seamlessly. We are attempting to solve this problem using a classification model using the MIMIC III datasets introduced above.

Sub Topics:

● Introduction to case study data.

● Learn about transfer learning in NLP by fine-tuning the BERT model

for your task.

● Using various attention based sequence modelling architectures like

LSTM and transformers to predict medical billing codes.

Chapter 4: Extracting Structured Data from Receipt Images

Chapter Goal: Just like any other sales job, the sales rep of a Pharma firm is

always on the field. While being on the field lots of receipts get generated for reimbursement on food and travel. It becomes difficult to keep track of bills which don’t follow company guidelines. In this case study we will explore how to extract information from receipt images and structure various information from it.

Sub Topics:

● Introduction to information extraction through Images.

● Exploring receipt data

● Using graph CNN to extract information

○ What is a graph convolutional architecture

○ How is it different from traditional convolutional layers

○ Applications

○ Hands on example to demonstrate training of a graph CNN

● Exploring recent trends in extracting information from template

documents.

Chapter 5: Handle Availability of Low-Training Data in Healthcare

Chapter Goal: Availability of training data has limited the use of advanced

models and general interest for problems in the healthcare

domain. Get introduced to weak supervision techniques that can

be used to handle low training data. Also learn about upcoming

libraries (like Snorkel and Astron) and research in this field.

Sub Topics:

● Explore weak supervision learning using Snorkel and Astron

● Learn to create label functions

● Hands on experimentation with a simple classification problem on

application of concepts from weak supervised learning

Chapter 6: Federated Learning and Healthcare

Chapter Goal: Federated learning enables distributed machine learning in

which machine learning models train on decentralized data.

This is deemed as the future of ML models as sharing patient

level data becomes more difficult for organizations due to

privacy and security concerns.

Sub Topics:

● Introduction to federated learning and what it means for healthcare

● Hands on example on how to use the concepts of federated learning

in one of your project

○ Load and prepare an example decentralized datasets

○ Design a federated learning architecture to predict diagnosis

of inflammation in bladder.

● Learn about TensorFlow federated

Chapter 7: Medical Imaging

Chapter Goal: Complete end to end analysis of how to develop a deep -

learning based medical diagnosis system using images. Learn about different kinds of image scans available like (cellular images, X-Ray scans etc.) . Also learn about the challenges such as accessibility of data, difference in image quality and how to address it, explainability etc. in disease detection via images.

Sub Topics:

● What is medical imaging

● Different kinds of image analysis

● Deep learning based methods for image analysis

● Understanding how to deal with 2-D and 3-D images

● Solve image classification and segmentation problem

● Understand challenges like accessibility of data, image quality issues,

explainability etc.

Chapter 8: Machine has all the Answers, Except What’s the Purpose of Life.

Chapter Goal: Introduction to concepts of a Question & Answering system.

Comparative analysis of different Question and Answering architectures. Hands-on-Example of building your own Q&A system to ask and query questions over published medical papers on pubmed.

Sub Topics:

● Review and understand various Question & Answering Techniques.

● Comparative analysis of different Question and Answering

architectures

● What is BERT architecture ?

● Using Bio-Bert architecture to train your own Q&A System

Chapter 9: You Need an Audience Now

Chapter Goal: Learned something from the book, excited to show it to the

world. In this chapter we are going to do exactly that, we are going to learn how to bring your models live and let the world interact with it. We will be building a Django app taking the Question Answering case study in point and also learning the basics of using docker for deployment.

Sub Topics:

● Understand technologies like Streamlit, Flask and Django that can help

you deploy your model depending upon the use case.

● What is docker and why should we dockerize our solutions.

● Building a production grade docker application.

● Django basics

● Using services like Heroku or Github SPAs to deploy your Django

App and bring it live.

O autorze

Anshik has a deep passion for building and shipping data science solutions that create great business value. He is currently working as a senior data scientist at ZS Associates and is a key member on the team developing core unstructured data science capabilities and products. He has worked across industries such as pharma, finance, and retail, with a focus on advanced analytics. Besides his day-to-day activities, which involve researching and developing AI solutions for client impact, he works with startups as a data science strategy consultant. Anshik holds a bachelor’s degree from Birla Institute of Technology & Science, Pilani. He is a regular speaker at AI and machine learning conferences. He enjoys trekking and cycling.

Blurb

Learn how AI impacts the healthcare ecosystem through real-life case studies with TensorFlow 2.0 and other machine learning (ML) libraries.

This book begins by explaining the dynamics of the healthcare market, including the role of stakeholders such as healthcare professionals, patients, and payers. Then it moves into the case studies. The case studies start with EHR data and how you can account for sub-populations using a multi-task setup when you are working on any downstream task. You also will try to predict ICD-9 codes using the same data. You will study transformer models. And you will be exposed to the challenges of applying modern ML techniques to highly sensitive data in healthcare using federated learning. You will look at semi-supervised approaches that are used in a low training data setting, a case very often observed in specialized domains such as healthcare. You will be introduced to applications of advanced topics such as the graph convolutional network and how you can develop and optimize image analysis pipelines when using 2D and 3D medical images. The concluding section shows you how to build and design a closed-domain Q&A system with paraphrasing, re-ranking, and strong QnA setup. And, lastly, after discussing how web and server technologies have come to make scaling and deploying easy, an ML app is deployed for the world to see with Docker using Flask.

By the end of this book, you will have a clear understanding of how the healthcare system works and how to apply ML and deep learning tools and techniques to the healthcare industry.

You will:

• Get complete, clear, and comprehensive coverage of algorithms and techniques related to case studies 
• Look at different problem areas within the healthcare industry and solve them in a code-first approach
• Explore and understand advanced topics such as multi-task learning, transformers, and graph convolutional networks
• Understand the industry and learn ML