Certificate Course on GenAI and ML in Healthcare

These modules are totally hands-on, and practise based. These are online, live, and totally interactive lab-oriented Modules with the primary objective of disseminating techniques of Healthcare Analytics using Data Visualization, Machine Learning, Deep learning and Gen AI. These technologies enable a practitioner to apply them on data in numerous ways.

1. Learn about different Health Systems across the World
2. Develop insights into healthcare data through visual analytics and optimize treatments
3. Learn techniques to group/segment data and patients
4. Improve patient outcomes through predictive analytics models
5. Building clinical decision support systems using Generative AI
6. Developing Large Language Model products for medical applications

The plan for the four modules is as below. Teaching sequence will also be in this order. Details about each Module are given below under respective Module heads.

About Visual Tools Used

We use the following Visual frameworks:

All these have highly intuitive user interface to perform analytics.

Module—I

Health Systems and Ethics

This foundational module introduces the concept of health systems, their structure, and the ethical considerations that govern healthcare practices. The course covers various health systems across the world, their evolution, and their role in delivering healthcare services. It also delves into the ethical issues prevalent in healthcare, particularly in the context of emerging technologies like AI and machine learning. The content covered in the module includes but not limited to:

1. Introduction to Health Systems:
   • Definition and components of a health system
   • Historical evolution of health systems globally
   • Comparison of health systems in different countries (e.g., U.S., U.K., India, UAE)
   • Role of health systems in public health and healthcare delivery
2. Types of Health Systems:
   • Public vs. Private Health Systems
   • Universal Health Coverage (UHC)
   • Health financing models (e.g., National Health Service, insurance-based systems)
   • Challenges in health systems: Accessibility, affordability, and quality
3. Ethics in Healthcare:
   • Basic principles of medical ethics: Autonomy, Beneficence, Non-Maleficence, and Justice
   • Informed consent and patient autonomy
   • Ethical dilemmas in healthcare: End-of-life care, resource allocation, and confidentiality
   • Case studies on ethical challenges in clinical practice
4. AI and Ethics in Healthcare:
   • Ethical implications of AI and machine learning in healthcare
   • Bias and fairness in AI algorithms
   • Data privacy and security in health informatics
   • The role of ethics in the deployment of AI-driven Clinical Decision Support Systems (CDSS)

Module—II

Machine Learning

We practice those modelling algorithms that consistently garner high performance, are relatively fast and are well known in ML community. Thus, these will be of immense use in many predictive applications.

1. Introduction to Machine Learning Technology
2. Data visualization and discovering structure in data. (Techniques include t-sne, parallel coordinates, mosaic plots) and Feature importance
3. Unsupervised learning techniques
   • K-means clustering
   • Expectation-Maximization algorithm
   • T-SNE & UMAP manifold learning technique
   • Dimensionality reduction & PCA
4. Supervised learning techniques for Classification and Regression
   • Decision trees
   • Ensemble modelling using Random Forest
   • Gradient Boosting Techniques
     • Gradient Boosting Learner
     • XGBoost
     • LightGBM
   • Performance measures: Accuracy, Precision and Recall, F-measure; Area Under the Curve, Cohen’s Kappa, Sensitivity, Specificity
5. Hyper-parameter optimization techniques
6. Interpreting Machine Learning Models

Module—III

Deep Learning

In Module-III we practise Deep-learning techniques. Deep Learning techniques are especially useful for image datasets, for example, chest-X-rays and CAT-Scans. These techniques are also used with sensor data (for example, ECG). If tabular data is sufficiently large, deep learning techniques can also be applied for making predictions:

1. Introduction to Neural Networks (NN)
2. Experiments with MLP networks
3. Regularising NN
   • Dropouts
   • Batch-normalization
   • l1 and l2 regularization
   • Starting weight initialization
4. Deep Learning with Convolution Neural Networks
   • Data Augmentation
5. Using very Deep Convolution Networks
   • Transfer learning with VGG16
   • Transfer learning with ResNet50
   • Transfer learning with InceptionV3
6. Recurrent Neural Networks
   • LSTM, GRUs and Bi-directional LSTM

Module—IV

Generative AI and LLMs

Generative AI and Large Language Models (LLMs) are integral components of the AI landscape, each with distinct roles and capabilities. Generative AI encompasses a range of tools that leverage information from LLMs and other AI models to create new content through machine learning. On the other hand, LLMs are a specific type of AI model that utilizes machine learning with billions of parameters to understand and generate text.

Generative AI and Large Language Models (LLMs) are increasingly being utilized in healthcare for various applications. A survey conducted in 2024 revealed insights into the adoption and impact of these technologies in the healthcare industry. The survey highlighted the importance of healthcare-specific, task-oriented language models, particularly in improving patient interactions, streamlining clinical processes, and advancing research. Respondents expressed a preference for LLMs in applications like transcribing doctor-patient conversations, medical chatbots, and answering patient questions, indicating their potential transformative impact on patient-facing tasks.

Furthermore, the integration of generative AI and LLMs in healthcare is seen as a significant advancement that can revolutionize the industry. These technologies can assist in tasks such as clinical note documentation and transcription, enhancing patient engagement through chatbots and virtual assistants, synthesizing biomedical literature, and optimizing clinical trials. Despite challenges related to data privacy and regulatory compliance, the benefits of generative AI in healthcare are evident, with applications that can streamline operations, improve patient care, and drive innovation in the field.

We may mention that we avoid using commercial LLMs in preference to highly reputed models available on HuggingFace and Ollama library. Benefits of these open-source LLMs are that these are fully customizable, can be fine-tuned on one’s data and totally take care of privacy and secrecy concerns—so important in medical field.

1. General Architecture of Transformers
2. Zero-shot classification and few-shot learning
3. Ollama and anythingLLM installation
4. Embedding, vector databases and search
5. Prompt Engineering
6. Developing knowledge products in healthcare using web-UIs for LLM
7. Using Flowise to develop RAG applications for biomedical applications

Modules Pedagogy

We strongly believe that a course in Healthcare Analytics can only be practice-based rather than pure theory based. We also believe that a practice-based course requires constant interaction with the teacher during lecture hours in real time. Our teaching pedagogy is like this: First, the algorithm (or theory part) is conceptually explained without getting into mathematics and then a project is undertaken to implement the techniques. Healthcare datasets for implementation are made available in advance. During the lecture, we go through one of the visual frameworks such as KNIME workflow and explain the steps. At his end, the student goes through the same steps on his laptop. Consequently, results are available at our end as also with the students immediately. In short, both the teacher and students are working on their respective laptops simultaneously; students solve their problems and ask any questions to clarify. The whole experience is just as if everyone is sitting in a laboratory and working together. Students are required to have a laptop with a minimum of 16GB of RAM, 32GB would be preferable—RAM is cheap and existing RAM can easily be upgraded.

The healthcare industry generates lots of data and this data is analyzed by professionals specializing in numerous fields. The program would be very useful for Medical Practitioners, Bio-technologists, bio-informatics or in general, students of Life–Sciences Biology, Biotechnology, Biochemistry, Bioinformatics, Cell biology (cytology), Ecology, Molecular biology, Microbiology, Marine Sciences —will find the program very useful.

Officers Managing Hospitals, Data Scientists or programmers or Engineers, Doctors or those in academics or Healthcare workers will find program extremely beneficial.

Total Duration for all Modules is 56 hours spread over 8-weeks. Program will be delivered on Saturdays and Sundays. There are 8 hours of teaching per week. Students are expected to perform exercises. This methodology of “learning concept->performing class projects->Do self-exercises” leads to better and stress-free absorption.

Participating students should be having a laptop or desktop with minimum 16gb of RAM. More RAM is advisable. Preferably the processor should be not lower than i5.