1 of 32

AI in Healthcare: Opportunities and Requirements��

September 14th,2024

Presented by Oluwadara Adedeji

​

Health Rhythms

Health Rhythms

2 of 32

  • Education
    • Ph.D.(in view), Machine Learning, University College Dublin, Ireland
    • MSc., Engineering Artificial Intelligence, Carnegie Mellon University Africa
    • BSc. Electrical & Electronic Engineering, University of Ibadan, Nigeria
    • Duolingo dropout
  • Experience
    • Data Science/ Machine Learning Researcher, Health Rhythms
    • Research Associate, CyLab/ Upanzi Network
    • Research Assistant, CyLab/ Upanzi Network
  • Fun fact about me
    • I love memes. So you will see a few in my slides.

About Me

Health Rhythms

3 of 32

Table of Contents

  • AI Areas
  • AI Timeline
  • Opportunities: AI in Healthcare
  • Opportunities: Mental Health & AI
  • Requirements: Policies, Data, Privacy, Ethics
  • Requirements: Careers and Skills

Health Rhythms

4 of 32

What is AI?

​

Health Rhythms

5 of 32

A marketing gimmick?

What is AI?

​

Health Rhythms

6 of 32

What is AI? Definitions

Reference: Artificial intelligence and human development Smith & Neupane, 2018

Health Rhythms

7 of 32

Reference: https://medium.com/swlh/the-map-of-artificial-intelligence-2020-2c4f446f4e43

Health Rhythms

8 of 32

AI Areas

​

Hierarchy not included

Health Rhythms

9 of 32

What is AI? - Timeline

​

​

Kuipers, Martijn & Prasad, Ramjee. (2022). Journey of Artificial Intelligence. Wireless Personal Communications. 123. 10.1007/s11277-021-09288-0.

The beginning

Health Rhythms

10 of 32

What is AI? - Timeline

2014

2020

2022

2023

2024

2011

2012

IBM Watson competes in the quiz show, Jeopardy and defeats the show's all-time champion.  Later in 2017, neurologists use it to identify RNA-binding proteins altered in ALS

Geoffrey Hinton, Ilya Sutskever and Alex Krizhevsky introduced a deep CNN architecture that won the ImageNet challenge 

Ian Goodfellow and his team developed generative adversarial networks (GANs).

The University of Oxford created an AI tool called Curial to quickly detect COVID-19 in emergency room patients.

Open AI introduces GPT-4 and Elon Musk, Steve Wozniak, and thousands of other signatories called for a six-month halt on training "AI systems more advanced than GPT-4."

Open AI launches ChatGPT

OpenAI launches 01, a model with enhanced reasoning.

​

Now

To insert your own icons*:

​

Insert >> Icons

​

(*Only available to Office 365 subscribers)

Health Rhythms

11 of 32

Overpromised ?

​

Health Rhythms

12 of 32

Opportunities?�

[1] High-risk behaviour in hypomanic states Fletcher et. al 2013

Health Rhythms

13 of 32

Reference: Data Flair @https://data-flair.training/blogs/ai-in-healthcare-sector/

Health Rhythms

14 of 32

OPPORTUNITIES - Examples

Reference: Artificial Intelligence: Past, Present and Future by David Vernonz

  • Prediction of Alzheimer disease based on analysis of magnetic resonance imaging (MRI) scans
  • Diagnosis of benign and malignant breast nodules from ultrasound images
  • Robotic Surgery
  • Prediction of congestive heart failure and chronic obstructive pulmonary disease from patient clinical health records
  • Prediction of future clinical events from electronic health records
  • Classification of cancer from gene expression profiles
  • Prediction of malaria from microscopic smear images
  • Prediction of depression and suicide risk through Ecologocial Momentary Assessment
  • Prediction of mental health using behavioural data

​

Examples

Health Rhythms

15 of 32

OPPORTUNITIES – Mental Health

Behavioral Vital Signs and Passive Assessment

Image: https://maria-gilfoyle.medium.com/mapping-out-mental-health-tech-af47cb703225

Health Rhythms

16 of 32

My research ?�

Health Rhythms

17 of 32

BACKGROUND: BIPOLAR DISORDER

  • Affects about 40 million people in the world (WHO, 2019).
  • Characterized by two poles of mania/hypomania and depression.
  • In mania/hypomania,  elevated mood and energy, decreased need for sleep, impulsivity typically experienced. 
  • In a study [1], 70% of people with Bipolar Disorder reported overspending while hypomanic.

[1] High-risk behaviour in hypomanic states Fletcher et. al 2013

Snapshot from Wikipedia

Health Rhythms

18 of 32

CAN OBJECTIVE FINANCIAL DATA BE USED TO ASSESS MENTAL HEALTH?

In a research study of 46, 167 patients with Bipolar I Disorder(BPI) and Schizophrenia, and 65% percent having BPI, it was discovered that patients with BPI were more likely to experience any of the financial disruptive life events with odds ratio ranging from 1.15 to 1.50 [2]

​

[2] Assessment of Disruptive Life Events for Individuals Diagnosed With Schizophrenia or Bipolar I Disorder Using Data From a Consumer Credit Reporting Agency. Nau et. al. 2023

​

​

Quotes from people with bipolar disorder on themighty.com

“Finally, there is the impulsivity factor. With mania, there is no pause button to stop and reflect on whether or not I need the item. Get it and go, and maybe get another one”

“Long story short — I declared bankruptcy two years ago. It took a lot for me to take that step, and I carry a lot of shame about it.”

Health Rhythms

19 of 32

RESEARCH GAP

  • Prior work has explored assessing Bipolar Disorder, mental health and finance through surveys and longitudinal studies [3][4].
  • Limitations such as self-bias, non-quantitative evaluation, and more.
  • No tool to objectively detect mental health changes and predict mood episodes (based on financial behaviour).

​

[3] Financial difficulties in bipolar disorder part 2: psychological correlates and a proposed psychological model Richardson et. al.

[4] Longitudinal relationships between financial difficulties and eating attitudes in undergraduate students Richardson et. al.

�

Health Rhythms

20 of 32

RESEARCH AREAS

�E & F

+

S & ML

Mental Health

�MH�+

E & F

�MH

+

S & ML

MY�RESEARCH

FINANCIAL DATA

  • Time Series Analysis
  • Data Anonymization
  • Privacy- Enhancing Technologies
  • Privacy-Preserving Machine Learning

MENTAL HEALTH MODELING

  • Statistical analysis of mood using financial data
  • Features from participatory studies
  • Mental health assessment from financial behavior

MENTAL HEALTH & MONEY

  • Financial Behavior
  • Financial difficulty, stress, worry
  • Socioeconomic factors e.g income, education
  • Economic factors e.g. Recession

Economics and Finance

Statistics and Machine Learning

To insert your own icons*:

​

Insert >> Icons

​

(*Only available to Office 365 subscribers)

Health Rhythms

21 of 32

Requirements?

[1] High-risk behaviour in hypomanic states Fletcher et. al 2013

Health Rhythms

22 of 32

REQUIREMENTS – Policies

https://www.dentons.com/en/insights/articles/2024/june/13/ai-regulation-and-policy-in-africa

https://www.europarl.europa.eu/topics/en/article/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence

Health Rhythms

23 of 32

REQUIREMENTS - Data

  • Nice idea, but no data.
  • Data Quality: Missing or Duplicate data etc. Garbage in, Garbage Out
  • Data Instrumentation and Annotation
  • Data Privacy and Security
  • Data Minimization
  • Data and Concept drift
  • Need for Validation               �

Health Rhythms

24 of 32

Your data should represent what you are modeling

Health Rhythms

25 of 32

  • Privacy
  • Data Sharing Preferences [5]
  • Privacy-Enhancing Technologies e.g. Differential privacy, Anonymization
  • Privacy-Preserving Machine Learning e.g. FHE, Federated Learning
  • Measurement Bottleneck and Trust

​

  • Ethics
    • What is ethical?
    • Institutional reviews, regulatory bodies e.g ICO.
    • Informed consent.

​

REQUIREMENTS: Privacy and Ethics

[5] J. Brozena et al., “Supportive Fintech for Individuals with Bipolar Disorder: Financial Data Sharing Preferences to Support Longitudinal Care Management,” Feb. 28, 2024. doi: 10.1145/3613904.3642645.

Health Rhythms

26 of 32

Reference: https://www.sprintzeal.com/blog/career-opportunities-in-ai

Health Rhythms

27 of 32

  • Mathematical Fundamentals: Statistics, Probability, Linear Algebra
  • Programming: Python, R, Scala
  • Frameworks: Scikit-learn, Pytorch, Tensorflow
  • Data Engineering e.g. Airflow, SQL, dbt, Spark
  • Cloud platforms e.g. AWS, GCP, Azure etc.
  • MLOps e.g. MLflow, Grafana, CI/CD etc.
  • Visualization e.g. Tableau, Power BI
  • Deployment e.g. Docker, Kubernetes, etc.
  • Degree: Msc, PhD or not?
  • Domain knowledge e.g. Healthcare
  • Communication
  • Also passion!

​

REQUIREMENTS: ML Skills

Health Rhythms

28 of 32

REQUIREMENTS: Staying Abreast

  • MOOCs e.g. Data Talks Club courses
  • Portfolio of Projects
  • Community Engagement e.g. NSK.AI events
  • Newsletters
  • Podcasts
  • Publications and Conferences
  • LinkedIn
  • Twitter/X

Health Rhythms

29 of 32

REQUIREMENTS: Getting Started

Get the fundamentals right!

Health Rhythms

30 of 32

RECAP

  • AI Areas
  • History of AI
  • Opportunities: AI in Healthcare
  • Opportunities: Mental Health and AI
  • Requirements: Policies, Data, Privacy, Ethics
  • Requirements: Careers and Skills

​

​

​

Health Rhythms

31 of 32

Thanks for listening

  • .

Give yourself time to make mistakes and grow

Health Rhythms

32 of 32

CONTACT

LinkedIn: Oluwadara Adedeji

Twitter/X : darasiemi

Email(s):oluwadara.adedeji@ucdconnect.ie, adedejioluwadara7@gmail.com

Health Rhythms