All sessions + MCQs.pdf

22/10/2023 Emerging Technologies for Medicine and Healthcare 1 Presentation Outline INTRODUCTION TO EMERGING TECHNOLOGIES IN MEDICINE AND HEALTHCARE COURSE OUTLINE COURSE ASSESSMENT 2 1 22/10/2023 • Artificial Intelligence (AI): Artificial intelligence allows machines to simulate human in...

22/10/2023 Emerging Technologies for Medicine and Healthcare 1 Presentation Outline INTRODUCTION TO EMERGING TECHNOLOGIES IN MEDICINE AND HEALTHCARE COURSE OUTLINE COURSE ASSESSMENT 2 1 22/10/2023 • Artificial Intelligence (AI): Artificial intelligence allows machines to simulate human intelligence in their thinking and behavior. • Machine Learning (ML): subset of AI where machines learn from experience through statistics. • Deep Learning: subset of ML where machines learn deeply from data using artificial neural networks. 5 Machine Learning definitions 6 3 22/10/2023 In traditional programming, a programmer writes a list of statements (i.e., a program) that tell the computer how to perform a specific task. traditional eari.in i k if 7 Machine Learning - The computer is given a set of data and a goal to achieve (i.e. expected result), and it learns how to achieve that goal by finding patterns or relationships within the data. - The obtained model can then be used to make predictions about new, unseen data. 8 4 22/10/2023 Deep Learning • Subset of Machine Learning where machines learn deeply from data using artificial neural networks. 9 - Artificial neural networks are systems inspired by the structure and functioning of the human brain. - They consist of interconnected neurons, that work together to process and analyze information, enabling the system to learn and make predictions. 10 5 22/10/2023 Why has machine learning become so popular and widely used in recent years, despite being introduced in 1959? Availability of powerful computers - They can process large amounts of data quickly and efficiently. - They support the creation of more complex machine learning models that handle big datasets and perform advanced tasks. Increased availability of data - Machine learning algorithms require large amounts of data to be effective. - Digital technology has led to an exponential increase in the amount of available data: text, images, video, audio, … 11 is oval as 12 6 22/10/2023 Diagnosis and Disease Identification Approximate human decisions especially in complex and uncertain environments - AI systems can analyze patient data, such as medical images (X-rays, MRIs, …), lab results, … to assist in diagnosing diseases and medical conditions. - AI can help identify patterns and anomalies that might be missed by human physicians. 13 Computer-Aided Diagnosis Al Deep Learning techniques provide powerful solutions for medical image analysis problems 14 7 22/10/2023 Drug Discovery and development Deep Learning techniques can help expand the drug discovery universe by making predic;ons in more novel areas of biology and chemistry Deep Learning Algorithm Prediction of drug activity Feature extraction (activity, fingerprint…) Creation of a database with information on chemical substances 19 Artificial intelligence has numerous applications in healthcare Personalized Treatment Plans - AI can analyze a paCent's geneCc informaCon, medical history, and real-Cme data. - This can lead to more effecCve and personalized therapies. Fraud Detection Research and Clinical Trials Public Health Surveillance AI algorithms can detect fraudulent activities in healthcare insurance claims, helping to prevent healthcare fraud and reduce costs. AI assists researchers in analyzing vast datasets for clinical trials, identifying suitable candidates, and predicting patient outcomes. AI can analyze data from various sources to monitor public health trends, detect disease outbreaks, and respond proactively. 20 10 22/10/2023 These applications highlight how AI is transforming healthcare and medicine by enhancing patient care, optimizing processes, and advancing medical research and development. 21 22 11 22/10/2023 B l o c kc h a i n fo r H e a lt h ca re - Blockchain technology was initially designed as the foundation for Bitcoin. - Today, blockchain is used in various sectors, and healthcare is considered one of the most promising application areas for blockchain. 31 What is Blockchain? Distributed and Decentralized Ledger - Blockchain is a digital way to keep track of transactions. - It's distributed on many computers without a central authority. Collaborative Network - Instead of one computer controlling everything, many computers work together. - They make sure transactions are recorded correctly. Chain of Blocks Blocks are linked together to create a chronological and tamper-resistant chain of data. 32 16 22/10/2023 What is Blockchain? Permanent Transparent and Honest Immutable Once a block is added, it can’t be removed. Everyone can see the transactions, which guarantees fairness. - When a block is added, it becomes very difficult to change. - This ensures a clear and permanent history of all transactions, keeping them safe and reliable. Advantages of blockchain for healthcare 33 Reliable Medical Records: It offers a secure and reliable way to store and manage medical records. Patient Privacy Control: Patients have control over their data, ensuring privacy and consent. Emergency Data Access: In emergencies, it allows quick access to criCcal paCent informaCon. Interoperability: It simplifies data sharing between different healthcare systems Real-Time Collaboration: It improves decision making as blockchain allows several doctors from different location to view the same Medical data in real time. Counterfeit Drug Detection: It helps Identify Fake Medications, ensuring patient safety. 34 17 22/10/2023 More and more IoT wireless sensors are used in healthcare applica5ons 37 Remote medical supervision of patients BANSMOM Motivation Chronic diseases, including cardiovascular diseases, diabetes, heart failure or chronic respiratory diseases are frequent. Remote Patient Monitoring Data Forwarding and Analysis IoT devices collect health metrics remotely (heart rate, glucose rate, blood pressure, temperature, etc.). • Collected patient data is sent to a software application. • Healthcare professionals and/or patients can access and view the data. • Algorithms analyze the data for treatment recommendations and alerts. Alerts for Intervention • IoT devices can generate alerts based on data analysis. • Alerts can trigger healthcare professionals to intervene as needed, such as when detecting low heart rate, high blood pressure, … Advantages • IoT devices enable us to reduce healthcare costs, improve chronic disease management, and maintain system autonomy. 38 19 22/10/2023 Emerging Technologies in Medicine and Healthcare 2 AI for Medicine and Healthcare Blockchain for Medicine and Healthcare IoT for Medicine and Healthcare Quantum Technologies for Medicine and Healthcare 41 What Is Quantum Compu9ng? 2 Digital computers are made up of millions or billions of transistors which exist in two states - 'on' to represent 1 and 'off' to represent 0 Quantum computers use qubits (or quantum bits) that can exist as a 1 or 0 or both at the same time, or somewhere on a spectrum between the two For n qubits, a single operation can operate on 2 n values simultaneously Quantum algorithms can solve problems in seconds that would take a digital computer the entire life of the universe just to attempt to solve. 42 21 22/10/2023 Quantum Drug Discovery 2 Pharmaceutical companies take up to 10 years and often billions of dollars to discover a new drug and bring it to market. Biologists face two main challenges: § they have to test thousands of possible variables in the chemical formula to find the required properties needed to treat a variety of diseases. § they are limited only to molecules up to a certain size that a classical computer can actually compute. Thanks to quantum computers, all these variables can be processed simultaneously, and we can analyze larger-scale molecules which will likely reduces costs and increases productivity and efficiency in the field of drug discovery. 45 Quantum computing for healthcare systems 2 Improved Imaging solutions • The traditional MRIs can identify areas of light and dark, and the radiologist must have to evaluate the issues. • Quantum imaging solutions can differentiate between tissue types, which allows more detailed and precise imaging. Improved therapies The ability of quantum computers to handle complex problems can help physicians to define the best therapy plan for a patient. 46 23 22/10/2023 A program is a sequence of statements that tells the computer how to solve a specific problem Input 3 What’s a program? § What’s a good program? üIt outputs a correct answer for any input üIt outputs an answer in reasonable cost • reasonable time of input length • reasonable space of input length § What’s a bad program? üIt takes a loooong time for some input üIt uses a huuuge memory for some input Note that there exists unsolvable problems by any program 4 2 22/10/2023 11 Print() Function • The Python print() function allows us to display various types of data, such as numbers and text. 12 6 22/10/2023 Print Function The Python print() function takes multiple inputs, turns them into text, separates them with spaces, and adds a 'newline' character ('\n') at the end, all on a single line. print(12, 24, -2) 12 24 -2 Δ print('hi', 'there', -2) hi there -2 We have "1" on the first line, a blank line in the middle, and "2" on the third line. 13 Print option sep= By default, print() separates the items by spaces. The optional sep= parameter sets a different separator text. print(12, 24, -2, sep=':') 12:24:-2 print('but', 'not', 'including', sep='**') but**not**including print('but', 'not', 'including', sep='') butnotincluding 14 7 22/10/2023 Print option end= By default, print() puts a single '\n' after all the items. The optional end= parameter sets a custom string to appear after all the items. Write the statements that display '2+4+6=12' using the sep and end parameters. 15 Simple Assignment Statements • A specific value can be assigned to a variable 16 8 22/10/2023 Simple Assignment Statements • An easy way to understand the impact of an assignment is to consider that when a variable changes, its old value is replaced. x=2 print(x) 2 Before x 2 x = 2.3 After x 2.3 x = 2.3 print(x) 2.3 17 Input() Function The Python function input() allows the user to input something that is stored in a Python variable. 18 9 22/10/2023 eval() Function • Here is another sample interaction with the Python interpreter: The eval function will evaluate this expression and return a value, which is then assigned to the variable “a” 21 Datatype Conversion • Besides, we can convert the string output of the input function into an integer using the built-in int function 22 11 22/10/2023 Simultaneous Assignment • Python allows us also to assign multiple values to multiple variables all at the same time CAN be easily done with simultaneous assignments 25 Expressions • You can produce new data (numeric or text) values in your program using expressions § This expression uses the addition operator § This expression uses the multiplication operator § This expression uses the addition operator but to concatenate strings together 26 13 22/10/2023 Expressions 27 Conditional Python statement Syntax: if condition: do_something else: do_alternative 28 14 22/10/2023 In Python, indentation is crucial for defining the structure of your code • • • • • Python uses indentation to define blocks of code. Indent with either four spaces or a tab. Stick to one method; mixing can cause issues. Indentation enhances code readability. It clarifies the structure and relationships between blocks. 29 Example: Avoiding division by zero 30 15 22/10/2023 While loops • The while loop tells the computer to do something as long as a specific condition is true. • It essentially says: “while this is true, do this.” 31 Example A while loop to print the numbers 0, 1, 2, ... 9 i i < 10 0 True 0 1 True 1 9 True 9 10 False All done … 32 16 22/10/2023 For loops • The for loop, allows to perform a specific action repeatedly for each element in a collection. for variable in a_collection # Code to be executed in each iteration 33 Example 34 17 21/11/2023 Machine Learning definitions 3 In traditional programming, A program is a sequence of statements that tells the computer how to solve a specific problem 4 2 21/11/2023 Tom Mitchell (1997) Task-specific Learning is specific to a particular task or set of tasks (T). Performance is measured The success of a machine learning program is measured by a performance metric P, such as accuracy, speed, efficiency, … Experience is necessary Experience (E) can mainly be acquired from training data. Improvement is the goal Machine learning aims to enhance performance on task (T), as evaluated by metric (P). 7 EXAMPLE Suppose your program is designed to classify X-ray images as either "normal" or "abnormal" to detect lung diseases like pneumonia. The model is trained on a dataset of radiologistverified X-ray images. What is the experience (E), task (T), and performance metric (P) in this setting? T: Diagnosing lung diseases from X-ray images. E: A dataset of X-ray images verified by radiologists, used for model training. P: Accuracy, which measures the percentage of X-ray images correctly classified as "normal" or "abnormal". 8 4 21/11/2023 Introduction Supervised Learning 9 10 5 21/11/2023 Price ($) in 1000’s The first application helps predict Housing price in the city of Portland 400 300 200 100 0 0 500 1000 1500 2000 2500 Size in feet2 We use a dataset containing house sizes in square feet and their prices in thousands of dollars 11 Price ($) in 1000’s Housing price prediction 400 300 200 100 0 0 500 1000 1500 2000 2500 Size in feet2 Given this data, let’s use a learning algorithm to estimate the selling price for a 750-square-foot house. 12 6 21/11/2023 The second application helps predict health scores based on patient age We will analyze the relationship between patients over 40 and their health scores 15 Predicting Health Scores Based on Patient Age Let’s now analyze the relationship between patients of all ages and their health scores Using a straight-line model is not suitable 16 8 21/11/2023 To create a more reliable health prediction model, we can consider other factors like: Body Mass Index Indicates the potential for weight-related health issues Chronic Conditions Existing medical conditions can have a big impact on health Diet What people eat is vital for health 19 The third application predicts whether breast cancer is malignant or benign, based on the size of the tumors We have a dataset containing information about tumor sizes and their classification as either malignant (cancerous) or benign (non-cancerous) 20 10 21/11/2023 Supervised Learning: Dataset has corrected answers. Regression: The prediction results can be any number within a specified range Example: Housing price Classification: The predictions represent predefined classes. Example: Breast cancer (malignant, benign) 25 26 13 21/11/2023 What is linear regression? Linear regression is used to predict the value of a variable by exploiting its relationship with another variable. 27 Consider the pairs represented in the scatterplot 28 14 21/11/2023 • Let's draw a line and see how well it fits the data. • For every data point, we draw a vertical line segment to the line, and the length of this segment shows the error. This line doesn't fit the data well 29 This line doesn't fit well either; the resulting error would be very large, making the model very poor 30 15 21/11/2023 This line looks much better. This is clearly a better model than the earlier attempts. 31 This is the line based on calculations. This is similar to the purple one 32 16 12/12/2023 Deep Learning for the Diagnosis of COVID-19 using X-ray Images and for Cardiac Abnormali@es using ECG Records 1 Deep Learning Subset of Machine Learning where machines learn deeply from data using artificial neural networks 2 1 12/12/2023 Predicting whether breast cancer is malignant or benign based on the patient's age and tumor size Age Tumor Size A straight line on the graph separates malignant from benign tumors 3 - We can use an artificial neuron to predict whether breast cancer is malignant or benign based on the patient's age and tumor size. - Creating a linear boundary on the graph to distinguish between malignant and benign tumors. 4 4 2 12/12/2023 Human Brain Neuron Ar.ficial Neuron - A neuron is a function that takes n real numbers (x1 to xn) as input - produces a single real number as output: F(x1, ..., xn). 5 - A neuron is a function with two parts: • The summation of weighted inputs, including a bias term. • The application of an activation function. 6 3 12/12/2023 Deep Learning for the Diagnosis of COVID-19 using X-ray Images 13 Download and extract the contents of the dataset 14 7 12/12/2023 The second applica.on uses Deep Learning for the diagnosis of cardiac abnormali.es • The dataset includes Electrocardiogram (ECG) records of both normal and abnormal heartbeats. • The obtained model predicts potential cardiac abnormalities for new unseen ECG data. 31 The dataset includes ECG records of both normal and abnormal heartbeats 32 16 12/12/2023 Predicting potential cardiac abnormalities in new, unseen ECG data 33 17