Initiation to AI in HealthTech

Let's go!

Index

Intro: Why include AI in healthtech and scope of the course

Module 3: Knowledge test

Module 1: Introduction to AI in HealthTech

Module 4: Ethical, Legal, and Social Implications of AI in Healthcare

Module 1: Knowledge test

Module 4: Knowledge test

Module 2:AI Technologies and Tools in Healthcare

Module 5: Case Studies and real world applications

Module 2: Knowledge test

Module 6: How to include AI in your organization and product

Module 3: Data Management and Integration in AI

Module 7: Future of AI in Healthtech

Introduction to the course

Start course!

lightit.io

Unit 1: Introductionto AI in HealthTech

Let's go!

lightit.io

lightit.io

This unit's topics

1.

2.

3.

Definition and core concepts

What functions does AI have in healthcare nowadays?

Evolution of AI in healthcare

lightit.io

Why include AI in healttech and what we wil cover

Start module!

lightit.io

Definition and core concepts

So... What is AI?

It is the use or study of computer systems or machines that have some of the qualities that the human brain has, such as the ability to interpret and produce language in a way that seems human, recognize or create images, solve problems, and learn from data supplied to them. To put it more simply: It is a computer technology that allows something to be done in a way that is similar to the way a human would do it.

lightit.io

What functions does AI have in healthcare nowadays?

Accelerates

Personalize

It helps boost productivity by speeding up processes and providing high-quality foundational elements for tasks.

Makes interactions and information feel more familiar and specifically tailored to an individual.

For example: Document distillation: Condenses large texts into summaries, evidence tables, or visuals (e.g., dashboards, knowledge graphs) to highlight key information quickly. Code classification: Transforms unstructured data into specific alphanumeric codes used in other systems or processes.

For example: Jargon simplification: Converts complex medical terms into clear, simple language suited to the patient’s health literacy. Content tailoring: Adjusts educational materials or care plans based on a patient's specific condition, demographics, and data.

Simulate

Automate

With AI we can create virtual environments or models where you can test out workflows, run experiments, or visualize experiences to learn

It can carry out business and technical repetitive tasks and procedures.

For Example: Record summarization: Produces summaries of patient visits or care interactions for clinicians. Component compilation: Combines data from multiple systems into a complete, review-ready document, often with recommendations or images.

For example: Interaction visualization: Creates 3D digital models (e.g., cells, chemicals) to support discovery, teaching, and diagnosis. Hypothesis validation: Using virtual machines to test and refine ideas before real-world implementation.

lightit.io

Evolution of AI in healthcare

Okay, let's focus on the main milestones in the evolution of AI in healthcare.Here is a timeline highlighting the key shifts:

Shift to AI systems that learn patterns from data. Included the development of neural networks.Applications emerged: Medical Imaging analysis (X-rays, MRIs, CTs). Milestone: IBM's BlueGene (1998) advanced computing for research like drug discovery.

Emergence of Large Language Models (LLMs) for text tasks. Virtual Health Assistants (e.g., Google Health AI, Babylon Health) simulating conversation for patient advice. NHS trialled an AI triaging app.

1951: First AI program developed. 1956: The term "Artificial Intelligence" coined at the Dartmouth Conference. Primarily academic research.

AI research focused on rule-based and expert systems. Systems followed pre-defined rules for decision-making. Key Limitation: Couldn't learn from new data; rigid and lacked flexibility.

AI is used to tailor medical care to individual patients based on unique data (genetics, lifestyle, etc.). Areas of Impact: Genomics analysis, Wearable Technology monitoring (heart rate, etc.), Robotic Surgery assistance.

The Early Beginnings: Rule-Based Systems (1960s - 1980s)

The AI Revolution: Deep Learning & NLP (Around 2010s - Present)

The Rise of Machine Learning (ML) (1980s - 2010s)

Current State: AI-Powered Personalized Medicine

Early Foundations (Pre-1970s)

lightit.io

Evaluation

Evaluation

Take the test and see how much you have learned!

Question 1/6

Question 1/6

Question 2/6

Question 3/6

Question 4/6

Question 5/6

Question 6/6

Unit 2: AI Technologies in Healthcare

Let's go!

lightit.io

lightit.io

AI Technologies and Tools in Healthcare

Start module!

lightit.io

This unit's topics

1.

2.

Understanding the relationship between AI, Machine Learning (ML), Natural Llanguage Procesing (NPL) and Deep Learning (DL)

Examples of AI, Machine Learning (ML), Natural Llanguage Procesing (NPL) and Deep Learning (DL).

lightit.io

Understanding the relationship between AI, Machine Learning (ML), Natural Llanguage Procesing (NPL) and Deep Learning (DL)

AI

ML

NLP

DL

lightit.io

Examples of AI, Machine Learning (ML), Natural Llanguage Procesing (NPL) and Deep Learning (DL)

ML is being used to: Analyze medical images (X-rays, MRIs, CTs) to detect anomalies. Predicting the probability, onset, and progressionof diseases by analyzing patient data.

To grasp these concepts a bit better lets take a look into some examples.

DL is being used to: Processing aggregated Electronic Health Records (EHRs) for predictive tasks. Analyzing data from wearable devices.

NPL is being used to: Power virtual health assistants and chatbots for patient support and information. Also to assist in patient education .

lightit.io

Evaluation

Evaluation

Take the test and see how much you have learned!

Question 1/5

Question 2/5

Question 3/5

Question 4/5

Question 5/5

Unit 3: Data Management and Integration in AI

Let's go!

lightit.io

lightit.io

Data Management and Integration in AI

Start module!

lightit.io

This unit's topics

1.

2.

Process and good practices for preparing healthcare data for AI models.

Interoperability and Integration of Healthcare Data.

lightit.io

Process and good practices for preparing healthcare data for AI models.

So, to make the most out of AI we need the organization to be ready from a structural and technical point of view. This implies preparing healthcare data for AI models and that involves a multi-step process encompassing collection, preprocessing, and the necessary step of preparing data for training.

This process is crucial because data is the fuel that powers AI; without high-quality data, even the most sophisticated AI models will fail to deliver value. Before starting any AI initiative, a thorough assessment of your data landscape is crucial.

lightit.io

Process and good practices for preparing healthcare data for AI models.

Let's go over a process and good practices for preparing healthcare data for AI models:

Define AI Objectives and High-Impact Use Cases

Assess Data Readiness (Collection & Preprocessing Assessment)

Identify and Plan to Address Data Gaps and Infrastructure Needs

Implement Data Preparation Processes (Collection, Preprocessing, and Preparation for Training)

Ensure Robust Privacy and Security

lightit.io

Process and good practices for preparing healthcare data for AI models.

lightit.io

Interoperability and Integration of Healthcare Data.

Let's focus on one of the most important and talked about aspects of data management in healthcare: Interoperability. We know AI can help... but how?

lightit.io

Evaluation

Evaluation

Take the test and see how much you have learned!

Question 1/5

Question 2/5

Question 3/5

Question 4/5

Question 5/5

Unit 4: Ethical, Legal, and Social Implicationsof AI in Healthcare

Let's go!

lightit.io

lightit.io

Ethical, Legal, and Social Implications of AI in Healthcare

Start module!

lightit.io

This unit's topics

1.

2.

Bias in AI Models and Ethical Considerations

Legal considerations

lightit.io

Bias in AI Models and Ethical Considerations

lightit.io

Bias in AI Models and Ethical Considerations

Data Privacy and Security

Regulatory Compliance

Trust

The Challenge:

The Challenge:

The Challenge:

Healthcare data is highly sensitive, making its privacy and security paramount. AI systems often require vast datasets, increasing the potential risk if not properly managed. Increased cyberattacks are a major cause that can compromise patient data and disrupt critical healthcare operations with the use of AI in the healthcare system.

Building trust is imperative for the relationship between patients and an AI-based healthcare delivery system to succeed. Public perception of AI in healthcare varies, and concerns about accuracy and data security can be barriers to adoption.

The regulatory landscape for AI in healthcare is still evolving, requiring organizations to navigate new guidelines and standards for AI algorithms and their use.

lightit.io

Evaluation

Evaluation

Take the test and see how much you have learned!

Question 1/5

Question 2/5

Question 3/5

Question 4/5

Question 5/5

Unit 5: Case Studies andReal-World Applications

Let's go!

lightit.io

lightit.io

Case Studies and real world applications

Start module!

lightit.io

This unit's topics

1.

5.

4.

2.

3.

Kaiser Permanente: Ambient Listening Technology

EO Care: AI in enhancing patient education

LUMIA: AI mental health support

Northwell Health - AI-powered iNav

Puppeteer: patient-facing AI agents

lightit.io

1. Northwell Health AI-powered iNav

Impact: How has it helped?

How AI is used

The system has significantly improved the process for detecting certain cancers, like pancreatic cancer Northwell Health states that the iNav system has slashed time to treatment by 50% This demonstrates how AI can lead to accelerated diagnoses and speed the delivery of patient care. AI's ability to find subtle patterns in data would be crucial for identifying cancerous masses that might otherwise be missed or detected later.The organization is also exploring licensing iNav to other hospitals, suggesting a potential for broader positive impact in other healthcare systems.

The AI-powered iNav was developed in-house by Northwell Health and it is used to analyze images of patients’ MRIs and CT scans proactively. The system is designed to search for evidence of cancerous masses or lesions, specifically highlighting its use in finding pancreatic cancer, which often has a low survival rate due to late detection. This kind of image analysis aligns with general applications of AI in healthcare, such as Image recognition, which identifies and categorizes aspects of an image and Computer Vision, a role of AI in analyzing medical images, diagnosing conditions from X-rays, MRIs, and CT scans, and assisting radiologists.

"II has totally revolutionized our ability to get these people connected to care.” Dr. Daniel King, a developer of iNav

lightit.io

2. Puppeteer - Making the most of conversational AI

Puppeteer builds conversational AI agents to improve healthcare communication. Using large language models, the company helps create applications that can talk with patients in a natural, human-like way via text and voice. These agents aim to make healthcare more efficient, reduce the burden on providers, and offer patients a more supportive experience. Let's take a look at how it works and how it can help support patients

lightit.io

What did we learn from this case?

Build vs Buy dilemma

HIPPA

When developing from scratch you need to make sure everyone signs a BAA and that can take time. Consider this when planning.

Consider your tech stack before excecuting anything. It can determine the time and cost of your project.

Time is of the essence

EHR integrations

This is the most complicated step if your product is already developed and you want to incorporate AI. There are some (costly) ways around it and nothing is imposible. Just consider friendly integration options if you still can.

Time to market in such a dynamic industry is key. Depending of the project and resources you can estimate from one week to several months to deploy to production.

lightit.io

3. Kaiser Permanente - Ambient Listening Technology

Impact: How has it helped?

How AI is used

This innovative AI solution aims to reduce the time physicians spend on tedious documentation significantly. By automating transcription and summarization, AI takes over a repetitive task. This reduction in administrative burden allows physicians to dedicate more attention to patient needs and concerns. The ultimate impact is to enhance patient care and improve efficiency. Reducing physician stress is also noted as an advantage of AI in healthcare, which could result from decreased documentation burden.

Kaiser Permanente implemented ambient listening technology that utilizes AI. This technology works by automatically transcribing and summarizing conversations between doctors and patients during appointments.This application falls under the domain of Natural Language Processing (NLP), which focuses on teaching machines to understand, interpret, and generate human language. Ambient listening technology leverages Speech Recognition Technology, a key component of Conversational AI, to enable voice interactions and create transcripts. AI's ability to process information quickly is key to this function.

Creating space for the patient and the physician connection is what inspired us to implement this technology. And we hope that those connections and improved efficiencies will help with the sustainability of the practice of medicine for many doctors.”Ramin Davidoff, MD, executive medical director and chair of the board with the Southern California Permanente Medical Group

lightit.io

4. EO Care - Personalized treatment with AI agents

eo Care Inc.’s mission is to end the epidemic of unwise cannabis use and bring wise use to all who might benefit. In pursuit of this mission, they provide affordable, data-driven, and clinician-led care as well as innovative, purpose-designed CBD and THC products. Meet PEP, a ChatGPT-powered personal education platform for cannabis. It uses rules and prompts devised by eo’s clinical team in an effort to provide cannabis information that’s as personally relevant and clinically responsible as possible. The AI Agent is trained with eo Care’s own sources of information: an ever-growing library of respected cannabis research provided by trusted academics, government, and industry experts.

lightit.io

What did we learn from this case?

Build to scale

Legal does not trump innovation

Bear in mind how your project can evolve, not only as a product but also considering AI updates. AI advancing at the speed it is advancing begs for a constant QA process.

Think and consider the legal implications before implementing it. This may help to design a better workflow and avoid future mishaps

Dont skip leg day

UX/UI matters

Always consider how the user will interact and how they might feel. Specially if they are patients. Consider that people have a hard time trusting AI and a bad design might make it worse.

Deciding the information and data the model will be trained on is the most important part. Make sure you have checked the data process and tailored it to your organization

lightit.io

LUMIA: AI mental health support

Meet Lumia! An AI-powered meditation app that allows users to customize their meditation journey. Developed by Light-it's Innovation Lab, this app demonstrates how AI can benefit and complement mental health treatments.

What did we learn from this case?

Consider technical limitations and costs from the start

Performance first, concept second

A concept might seem great in theory, but it’s essential to test the user experience in real conditions to assess technical viability and performance constraints. Validate performance early to decide if technical improvements are needed or if the concept requires adjustment.

Initial estimates often underestimate real-world computational demands and scaling requirements. Research actual costs and technical constraints before committing to a product direction that might prove financially unviable.

Plan for unexpected blockers in your timeline.

AI Technology moves fast

Staying updated with the latest AI model releases and improvements can unlock massive product value. Stay informed about advances in your tech stack to avoid missing competitive advantages.

Complex bugs and integration issues are difficult to predict but inevitable in technical projects. Building buffer time into your deadlines can make your roadmap more resilient and realistic.

lightit.io

Evaluation

Evaluation

Take the test and see how much you have learned!

Question 1/2

Question 2/2

Unit 6: How to include AI in your organization and product

Let's go!

lightit.io

lightit.io

This unit's topics

1.

2.

Process and good practices for preparing your organization or product

How do I start? Checklist for a successful implementation

lightit.io

How to include AI in your organization and product

How to include AI in your organization and product

So, what are the steps to include Artificial Intelligence (AI) in your organization and product? We have broken down this into a strategic process that moves beyond simply adopting a technology to fundamentally aligning AI with your business goals and integrating it into your workflows. It's not a silver bullet, but rather a set of sophisticated tools requiring careful planning and execution. The success of AI adoption requires solving concrete business problems that deliver measurable results and drive real value.

lightit.io

How to include AI in your organization and product

The process for successfully including AI begins long before implementing any technology. It starts with a clear understanding of why you are adopting AI.

Impact:How transformative is the solution? Will it automate an existing process or job entirely, or supplement a human agent?

Extensibility: Will this solution have broader potential to be deployed into other areas of my business or other customer groups?

Function: What strengths in datasets and integrations does our business have to create a competitive moat?

Time to market: Based on the complexity of the solution and implementation level of effort, how soon can I begin to realize value?

Permission space: Where do I intend to deploy this solution, and what will those stakeholders' receptivity be?

Measureability: How will we understand and measure the outputs and performance of this model?

lightit.io

How to include AI in your organization and product

Let's go over a process and good practices process for preparing your organization or product for AI:

You must identify concrete goals using frameworks like SMART goals to ensure AI initiatives drive measurable value. This prevents being distracted by trendy but irrelevant technologies.

Define Clear Business Objectives

Conduct an internal audit to find bottlenecks, inefficiencies, or opportunities where AI can help, distinguishing between challenges that have clear technical solutions and those requiring behavioral or cultural changes.

Identify Organizational Challenges

We have gone over this before. Check the detailed checklist to make sure this step runs smoothly

Assess Data Readiness

Create a detailed project plan that outlines implementation phases, resource requirements (talent, infrastructure), and risk mitigation. Assess internal skills and infrastructure needs, and consider potential external partnerships.

Plan Resources and Capabilities

Decide whether to develop solutions in-house, partner with external vendors, or use a hybrid approach, balancing control, speed, expertise, and resource investment.

Evaluate Implementation Strategy

lightit.io

How to include AI in your organization and product

Start with pilot projects to demonstrate value. The transition from prototype to a production-ready solution requires planning for scalability, robust execution, cross-functional collaboration, and automation (like MLOps).

Implement and Bridge to Production

Once pilots prove value, expand across the organization, identifying high-impact areas first and using structured feedback loops from early deployments to refine the solution. Monitor performance using business-aligned metrics

Scale Successfully

Implementing AI requires thoughtful change management, addressing cultural transformation alongside technical deployment. This involves creating champions for AI adoption, providing clear communication and training for teams, and proactively addressing potential resistance or fears about job displacement. Highlight how AI augments human capabilities.

Manage Organizational Change

Given the sensitive nature of healthcare data, ensuring robust data privacy and security is paramount, including adherence to regulations like HIPAA and GDPR. Cybersecurity strategies are crucial to protect against attacks. It's also important to address potential bias in data and algorithms and ensure transparency and accountability. Human expertise and oversight are essential to ensure responsible application and avoid issues like AI "hallucination" or fabrication of false information.

Address Ethical, Legal, and Security Considerations

lightit.io

How do I start? Checklist for a successful implementation

lightit.io

Unit 7:Future of AI in HealthTech

Let's go!

lightit.io

lightit.io

Future of AI in HealthTech

Start module!

lightit.io

This unit's topics

Future of AI in HealthTech

lightit.io

Emerging AI Trends Shaping the Future of HealthTech

lightit.io

Final Video

lightit.io

Congrats!

lightit.io

Certificate

YOU COMPLETED THE COURSE:

Initiation to AI in HealthTech

Congratulations!

Certificate of achievement

Empowered by Light-it, you can now illuminate the path to a healthier future with your healthtech expertise.

Dael Stewart

Managing Director at Light-it

February 2025

Course completed!

Visit Light-it

Define AI Objectives and High-Impact Use Cases:

The process begins not with data itself, but by defining clear, measurable AI objectives using frameworks like SMART goals. You must also map potential high-impact use cases. This critical first step determines the data needed to support your AI initiative.

Regulatory Compliance

Action Steps you could take:

• Stay informed about the evolving regulatory landscape for AI in healthcare.
• Collaborate with healthcare organizations, AI researchers, and regulatory bodies to establish guidelines and standards for AI algorithms and their clinical use.
• Adopt agile governance and take a portfolio approach to strategic bets, accounting for risk, regulation, and trust.
• Proactively deploy generative AI plans that will thrive in the emerging regulatory landscape.
• Establish governance inclusive of risk, compliance, and ethics from early stages.

Professional organizations have created frameworks to address the specific challenges involved in developing, reporting, and validating AI in medicine. These frameworks often emphasize educating the people who build the technology and promoting transparency in its use. Meanwhile, regulatory oversight is still in the early stages. Agencies such as the U.S. FDA and the European EMA are beginning to issue guidelines and have made regulating AI in healthcare a strategic priority to help shape a tech-driven future. Based on recent announcements and executive orders, more comprehensive regulations are expected soon.

Implement Data Preparation Processes (Collection, Preprocessing, and Preparation for Training):

This is where the plans are put into action:

• Data Collection: Collect the required data from relevant sources, ensuring it is inclusive and representative to reduce algorithmic bias.
• Preprocessing and Integration: Clean and standardize the data, integrate it from multiple sources (e.g., EHRs, wearables, patient reports), and resolve inconsistencies or missing information. AI tools like NLP and ML can assist in this process. It's also important to establish data governance frameworks at this stage.

• Preparation for AI Model Training (often involves Annotation): Prepare the cleaned data for training by labeling it, especially for supervised learning tasks (e.g., marking anomalies in medical images or outcomes in patient data). While annotation is critical, specific methods may not always be detailed in sources.

Ensure Robust Privacy and Security:

Throughout the entire process, it is paramount to implement stringent guidelines for data-sharing and robust security measures to protect patient data. Adherence to regulations like the Health Insurance Portability and Accountability Act (HIPAA) in the United States and the General Data Protection Regulation (GDPR) in Europe is crucial. AI can even be used to enhance data security by detecting unusual activities.

Trust

Action Steps you could take:

• Focus on building trust with patients and providers regarding AI systems.
• Ensure transparency and explainability in AI decision-making processes where possible and appropriate.
• Account for patient education, concerns, and comfort levels during AI integration planning.
• Prioritize responsible AI practices that emphasize safety, robustness, reliability, responsibility, privacy, transparency, and fairness.
• Communicate strategically and implement training endeavors to champions AI literacy organization-wide.

Patient-physician trust is vital in improving patient care and treatment effectiveness. Building a relationship based on trust is imperative for AI-based healthcare delivery systems. Transparency and responsible AI practices are linked to trust. Patient education, concerns, and comfort levels should be accounted for when planning for AI integration. Leading by building trust is critical, as without it, consumers, clinicians, and organizations will not maximize Generative AI solutions.

Data Privacy and Security

Action Steps you could take:

• Strictly adhere to healthcare regulations such as HIPAA and GDPR to safeguard patient data.
• Actively minimize the risk from malicious behavior that could disrupt operations and compromise trust.
• Implement safeguards to prevent unintentional exposure of confidential information by AI models or protect against cybersecurity threats.
• Implement stringent guidelines for data-sharing, especially for confidential or personally identifiable information.
• Deeply investigate and keep up to date with cybersecurity and the cyber risk landscape of healthcare systems
• Develop and implement comprehensive cybersecurity strategies.

Robust security measures and adherence to healthcare regulations are crucial to maintaining patient information confidentiality and developing trust. Regulatory frameworks like the Health Insurance Portability and Accountability Act (HIPAA) in the United States and the General Data Protection Regulation (GDPR) in Europe are in place to safeguard individual privacy, though GDPR has implemented extensive data protection law within the EU, creating a significant global shift in data protection. Ensuring data exchanged between systems is accurate and reliable is also vital.

Assess Data Readiness (Collection & Preprocessing Assessment):

Conduct a comprehensive assessment of your data landscape. This assessment looks at three key areas:

• Data Availability: Do you have the necessary data to support your identified AI use case(s)? This involves understanding what data sources exist (e.g., EHRs, medical images, sensor data, patient reports) and if they contain the information required.
• Data Access: Are your data systems integrated, or are they stored in isolated "silos"? Healthcare data often comes in various formats and is stored in disparate systems. Assessing access identifies the challenge of integrating these sources.
• Data Quality: Is your data clean, consistent, and reliable? Errors, inconsistencies, or missing information in healthcare data can have severe consequences. This assessment identifies issues that need to be fixed.

Identify and Plan to Address Data Gaps and Infrastructure Needs:

Based on the data readiness assessment, you will identify specific data gaps, quality issues, and infrastructure limitations that need to be addressed. This requires planning for investments to upgrade and optimize your data infrastructure and develop strategies to fill data gaps.