Artificial Intelligence in Healthcare

Artificial Intelligence in healthcare

Carly Noel, DO, MPH Clinical Informatics Fellow

Start

Disclosure

Dr. Noel is the President of the AMIA Clinical Informatics Fellows under the American Medical Informatics Association

Goals and Objectives

Learn the fudamental definitions and uses of AI and machine learning

Identify the problems that healthcare providers face and how machine learning can augment the solution

Determine how AI affects patient safety, care quality, and research

Apply the building blocks of AI to help understand emerging technologies

Introduction

2.

1.

Click on the icons to learn more!

4.

3.

The Evolution of AI in Healthcare

2015

2012

2019

2017

2010

Medical imaging

Predictive analytics

Robotic surguries

AI-driven administrative tools

Personalized medicine

2023

2023

2021

Real-time diagnostics

Virtual health assistants

Telemedicine

Key Definitions

Artificial Intelligence

AI refers to systems that perform tasks requiring human intelligence—like learning, decision-making, and problem-solving.

Machine Learning

Machine learning uses algorithms and data to help systems learn and make predictions without explicit programming.

Example

Key Definitions

Title

Title

Use this side to give more information about a topic.

Use this side to give more information about a topic.

Subtitle

Subtitle

Flip to reveal image - click on image to enlarge

Flip to revel definition

Flip to revel definition

(Swipe to reveal image!)

Deep Learning trains LLMs with neural networks to understand patterns in text and code.

Deep Learning

LLMs are AI systems trained on massive text and code datasets to generate and understand language.

Large Language Model (LLM)

Title

Title

Use this side to give more information about a topic.

Use this side to give more information about a topic.

Subtitle

Subtitle

Flip to revel definition

Flip to revel definition

Benefits of AI in Healthcare

Benefits of AI in Healthcare

REFERENCES

Click on the icons below for more info

Enhanced Clinial Care

Operational Efficiency

Personalized & Precise Treatment

Accelerated Research & Innovation

Real-World Examples of AI in Healthcare

Examples of AI in Healthcare

• Calvin is a 65-year-old male with a 30-pack-year smoking history who is presenting to the oncology clinic after a suspicious lung lesion was seen on routine low dose CT imaging
• You are a medical student rotating in the clinic, and you have been interested to learn more about how Artificial Intelligence is being used in cancer diagnosis and treatment
• As we move forward, let’s think about how artificial intelligence can be a part of his care

Meet Our Patient

Real-World Examples

Some AI models now match or outperform radiologists in spotting certain lesions

Detects pulmonary nodules, breast, and colon cancers.

AI in Medical Imaging

• Lung nodule characterization involves analyzing a nodule’s size, volume, and density to determine malignancy
• AI algorithms have been shown to measure these variables and accurately track the growth in screening, as well as diagnose
• Relevant studies:
• Ardila et al, 2019: showed diagnosis of lung cancer using deep learning
• Delzell et al., 2019: verification of nodules as benign or malignancy
• **Aydin et al., 2021: differentiate squamous cell vs adenocarcinoma vs small cell carcinoma

Imaging for Calvin

AI in Medical Imaging

Real-World Examples

Supports providers with transcription, EHR updates, prior authorizations, and billing guidance.

Helps patients with scheduling and routine questions.

Virtual Health Assistant (VHA)

• Oncology patients often face medical journeys that require complex care coordination
• They often see multiple specialists, require frequent monitoring, have constantly changing medications, and require more communication
• Calvin his been able to use the clinics VHA for the following tasks:
• Coordinating communication between him and his healthcare providers
• Assisting in medical billing
• Scheduling appropriate follow up appointments
• Answering straightforward medical questions, such as “what is the dose of my medication”

VHA for Calvin

Virtual Health Assistant

Real-World Examples

Uses data to predict future health outcomes

Enables earlier disease detection

Predictive Analysis

Can forecast disease outbreaks with large datasets

Supports personalized treatment plans

Predictive Analysis for Calvin

Based on Calvin's history, he is able to receive personalized care through an AI algorithm to determine the best course of treatment for his cancer

This can improve his outcomes and decrease unnecessary treatments

Predictive Analysis

For providers, models can also be used for cancer survival prediction based on a patient's unique demographic, historical, and medical circumstances

Real-World Examples

Learns from demonstration to guide robotic instruments.

Reduces human error and automates tasks like suturing.

Detects abnormal anatomy and potential obstacles.

Shortens operation time.

Robotic Surgery

Collects surgical data to improve future AI models.

Robotic Surgery for Calvin

• Between the oncologist and the radiologist, Calvin’s lung cancer was determined to be Stage 2 and operable
• Minimally invasive thoracic surgery (MITS) was chosen for treatment, which uses a robotic system to assist in the removal of lung cancer
• Calvin benefitted from a faster surgery, small incision, and shorter recovery time

Robotic Surgery

Knowledge Check

Multiple Choice Quiz

Challenges and Limitations of AI in Healthcare

Challenges and Limitations of AI in Healthcare

REFERENCES

Click on the icons below for more info

Data Privacy & Security

Bias in Algorithms

Lack of Transparency

Regulatory & Legal Hurdles

Integration

Ethical Considerations

Flip each card for the definition of the term.

Patient Consent

Transparency and Accountability

Human Oversight vs. Automation

• De-identified data doesn’t require consent, but patients may be unaware it’s used

• Robust security measures are essential to protect patient information

• AI recommendations should be explainable
• Models should be continuously evaluated for improvement

Use this side of the card to provide more information about a topic. Focus on one concept. Make learning and communication more efficient.

Use this side of the card to provide more information about a topic. Focus on one concept. Make learning and communication more efficient.

Use this side of the card to provide more information about a topic. Focus on one concept. Make learning and communication more efficient.

• Always include human oversight to ensure ethical, appropriate use

• Providers should not rely solely on AI for decisions

Title

Patient Consent

Title

Transparency and Accountability

Title

Human Oversight vs. Automation

Write a brief description here

Write a brief description here

Write a brief description here

Knowledge Check

Multiple Choice Quiz

Future Outlooks

Job Impact

Public Health Benefits

• AI enhances public health efforts
• Supports pandemic response and population-level health planning

• Fears of AI replacing physicians are largely unfounded
• AI will change workflows but cannot replace expertise, judgment, or human connection

Genomics & Research

Personalized Medicine

• AI supports improved personalized and precision medicine
• Helps tailor treatments to individual patients

• Enables applications in clinical genomics and pharmacogenomics
• Helps advance research and improve health outcomes

Conclusion

AI is a powerful technology that is rapidly advancing

It is not a replacement for physician guided clinical care

Any AI tool requires responsible oversight, appropriate governance, and security for data privacy

Overall, AI is here to stay and will undoubtedly revolutionize healthcare

Where to Learn More

• Introduction to Artificial Intelligence (AI) in Health Care – Learning Series
• https://edhub.ama-assn.org/change-med-ed/interactive/18827029

• American Board of Artificial Intelligence in Medicine (ABAIM) courses
• https://abaim.org
• AMIA For Your Informatics Podcast: Episode 35 – AI in Medicine and Healthcare
• https://amia.org/news-publications/podcasts/for-your-informatics/your-informatics-episode-35-ai-medicine-and

Thank you!

Carly Noel Carly.Noel@CCHMC.org

Personalized & Precise Treatment

• Uses patient data (genetics, history, labs) for targeted care
• Identifies patterns in large datasets

• Minimizes false positives

• Optimizes imaging and lab use

Integration with Existing Systems and Workflows

• Integration can be costly, especially with legacy systems
• Users may resist workflow changes → requires proper training

• Interoperability issues can affect data standardization and quality

Regulatory & Legal Hurdles

• HIPAA compliance can be hard to verify, especially with large datasets
• Legal responsibility for AI errors is complex

• Rapid AI evolution makes regulation difficult

• Need for frameworks ensuring AI decisions are explainable and understandable

The Impact

AI helps improve outcomes, cut costs, and enhance care quality.

Emerging Relationship

AI is rapidly transforming healthcare and every clinician needs to understand it.

Correct!

The 21st Century Cures Act prohibited information blocking. This means that anything that hinders the access, exchange, or use of health information is prohibited. This led to more data sharing and better coordination of care. Healthcare systems as well as electronic health record vendors were mandated to follow these changes.

Lack of Transparency

• Many AI models are “black boxes,” hiding how conclusions are made
• Makes validation and oversight difficult

• Hallucinations: AI generates false or misleading information (e.g., ChatGPT, CoPilot)

• Risks: reduced trust and potential medical errors

Accelerated Research & Innovation

• Optimizes clinical trial design
• Analyzes large datasets to find new insights

• Reduces administrative burden in research

• Supports faster, data-driven discoveries

Correct!

The 21st Century Cures Act prohibited information blocking. This means that anything that hinders the access, exchange, or use of health information is prohibited. This led to more data sharing and better coordination of care. Healthcare systems as well as electronic health record vendors were mandated to follow these changes.

Powering Progress: Big Data

Most breakthroughs come from Big Data — vast health info from:

• Electronic records
• Genomics

• Wearables 
• Medical devices

Enhanced Clinical Care

• Supports clinical decisions: diagnosis, prognosis, treatment
• Improves diagnostic accuracy (sometimes surpassing humans)

• Enables earlier disease detection

• Reduces errors and unnecessary tests

Changing the Pace

Healthcare has been slow to adopt tech, but AI is the exception.

Bias in Algorithms

• AI can reflect human and systemic biases in training data
• This may lead to unequal treatment or outcomes for marginalized groups
• Mitigation Strategies:

• Use diverse, representative data

• Follow ethical guidelines

• Involve providers, patients, and communities

• Conduct thorough validation and quality checks

Data Privacy & Security

• AI needs lots of patient data, raising privacy concerns
• Risks include re-identification, lack of transparency, and uncontrolled access

• Requires strict governance: informed consent, encryption, access controls, and ongoing monitoring

Operational Efficiency

• Automates repetitive tasks: billing, scheduling, claims
• Transcribes and summarizes patient records

• Answers routine patient portal messages

• Streamlines patient management and reduces wait times