The future of AI in medicine | Conor Judge | TEDxGalway
TLDRConor Judge's TEDxGalway talk explores the future of AI in medicine, focusing on multimodal AI that processes various data types. He discusses the potential of AI to improve healthcare efficiency, using examples like 'Chest Link' for X-ray triage and 'Med-PaM' for medical question answering. Judge emphasizes the importance of trust, explainability, and clinical trials for safe AI implementation and envisions a future where AI supports doctors, enhancing patient care and access to specialized healthcare globally.
Takeaways
- 🕵️♂️ Conor Judge, a medical consultant and lecturer, draws a parallel between his past as a young detective in a play and his current role in solving medical mysteries.
- ⏱️ The imbalance in healthcare delivery, with 70% of time spent on data collection and only 30% on decision-making and patient communication, is a global issue exacerbated by electronic health records.
- 🤖 The introduction of multimodal AI in medicine, which processes various data forms like text, images, and numbers, offers a potential solution to the information overload faced by healthcare professionals.
- 🔎 AI has advanced in medical imaging, with systems like Chest Link capable of autonomously triaging X-rays and identifying abnormalities, thus sharing tasks with human radiologists.
- 👁️ An AI model developed by researchers in University College London can diagnose eye diseases and even predict conditions like Parkinson's disease from retina images, highlighting AI's ability to see beyond human capabilities.
- 🗣️ Large language models, such as Med-PaM by Google, have demonstrated the ability to answer medical questions, with Med-PaM 2 scoring at an expert level on the US medical licensing exam.
- 📱 The latest multimodal AI models are accessible through smartphones, exemplified by OpenAI's chat GPT, which can analyze medical images and provide follow-up advice.
- 🔒 Trust, explainability, and randomized clinical trials are essential for the safe implementation of multimodal AI in healthcare to ensure patient safety and acceptance.
- 🔍 Explainable AI is crucial for understanding the decision-making process behind AI models, which is vital for medical professionals to make informed decisions.
- 🧩 Randomized clinical trials are necessary to test AI models in healthcare, providing evidence-based validation similar to drug testing protocols.
- 🌐 The future of healthcare with multimodal AI envisions a more efficient, personalized, and accessible system, especially beneficial for remote and low-income regions lacking specialized care.
- 💖 The importance of maintaining compassion and the human touch in healthcare, ensuring that AI serves as a tool to enhance, not replace, the doctor-patient relationship.
Q & A
What was Conor Judge's role 26 years ago when he last stood on the stage at the town hall theater?
-Conor Judge was a 12-year-old boy participating in a drama competition for schools, playing the role of a detective in a play written by his best friend, trying to solve a mystery of who robbed a fictional hotel called Hotel El Chipo.
How does Conor describe his current professional role, and how does it relate to his past experience as a detective in a play?
-Conor is now a medical consultant and a senior lecturer in applied clinical data analytics. His role has evolved from solving a fictional mystery to diagnosing the cause of illness in real patients, with a similar emphasis on gathering information and making decisions based on that information.
What is the common challenge Conor identifies in healthcare delivery that he faced both as a young actor and a medical professional?
-The common challenge is the imbalance of time spent on collecting information (70%) versus making decisions and communicating with patients (30%), which has been exacerbated by the introduction of electronic health records.
What is the term for AI that Conor Judge introduces as a potential solution to the challenges in healthcare delivery?
-Conor introduces 'multimodal AI' as a potential solution, which is AI that takes in data in many different forms, such as text, images, and numbers, similar to how a doctor assesses a patient in a hospital.
What is the difference between multimodal AI and single model AI according to Conor's explanation?
-Multimodal AI processes and analyzes data from multiple types of inputs, whereas single model AI focuses on a single type of data, such as images, text, or numbers.
Can you provide an example of single model AI in healthcare mentioned by Conor?
-One example is Chest Link by OxyPit, a medical AI triage system that can autonomously report on chest X-rays, identifying 75 abnormalities and facilitating task sharing between AI and human radiologists.
What is unique about the AI model developed by researchers at University College London for diagnosing eye diseases?
-The AI model is trained on 1.6 million pictures of the retina and can diagnose eye diseases and predict outcomes from conditions like macular degeneration. Remarkably, it can also predict Parkinson's disease years before symptoms develop by analyzing the retina.
What is Med-PaM, and how did it perform on the US medical licensing exam?
-Med-PaM is a medical large language model developed by Google. It passed the US medical licensing exam with a score of 67% in its first version and improved to an 86% score in the second version, which is considered expert level.
How does Conor envision the future integration of multimodal AI in healthcare, especially considering the patient's context?
-Conor envisions a future where a picture or video of the patient is also fed into the multimodal model, allowing for a more personalized and efficient healthcare system that can provide insights even in remote areas with limited access to specialized care.
What are the three key elements Conor believes are necessary for the safe implementation of multimodal AI in healthcare?
-The three key elements are trust, explainability, and randomized clinical trials. Trust is essential to overcome patient anxiety about AI in healthcare, explainability helps to understand the AI's decision-making process, and randomized clinical trials ensure the AI's effectiveness and safety.
What does Conor emphasize as the missing piece in the integration of AI with healthcare, and why is it important?
-Conor emphasizes the importance of the 'eyeball test' or the initial assessment of the patient's context before interpreting any medical data. This human element, which includes compassion and understanding, is crucial for building a relationship between AI and healthcare professionals to enhance patient care.
Outlines
🎭 From Childhood Drama to Medical AI
The speaker reminisces about performing in a drama competition 26 years ago, where he played a detective with a stammer, and draws a parallel to his current role as a medical consultant and lecturer. He discusses the imbalance in healthcare delivery, where doctors spend 70% of their time collecting patient information and only 30% making decisions. The speaker introduces the concept of multimodal AI, which processes various data forms like text, images, and numbers, and contrasts it with single-model AI, such as large language models, machine learning, and computer vision. He emphasizes the potential of AI to improve healthcare efficiency and patient interaction.
🤖 Cutting-Edge Single-Model AI in Healthcare
The speaker provides three examples of single-model AI applications in healthcare. The first is Chest Link, an AI system that autonomously reports on chest X-rays, identifying 75 abnormalities and serving as a triage tool. The second is an AI model developed at University College London that diagnoses eye diseases and predicts outcomes, even detecting conditions like Parkinson's disease from retinal images. Lastly, the speaker mentions Med-PaM, a large language model by Google that passed a US medical licensing exam, showcasing AI's ability to answer medical questions and its rapid improvement over time.
🔮 The Future of Multimodal AI in Medicine
The speaker discusses the implementation of multimodal AI in healthcare, emphasizing the need for trust, explainability, and randomized clinical trials. He addresses concerns about patients' anxiety regarding AI in healthcare and the importance of understanding AI's decision-making process. The speaker also highlights the importance of clinical trials to test AI models, comparing them to the gold standard of medical testing. He envisions a future where multimodal AI, including patient images or videos, enhances healthcare efficiency and accessibility, especially in underserved areas. The speaker concludes by stressing the importance of compassion and the human-AI relationship, allowing doctors to spend more time with patients and improve their health outcomes.
Mindmap
Keywords
💡AI in medicine
💡Multimodal AI
💡Medical data analytics
💡Electronic health record
💡Task sharing
💡Chest X-ray
💡Retina analysis
💡Parkinson's disease
💡Medical question answering
💡ECG
💡Randomized clinical trials
💡Explainable AI
Highlights
Conor Judge's experience as a medical consultant and senior lecturer in applied clinical data analytics shapes his perspective on the future of AI in medicine.
The imbalance in healthcare delivery, with 70% of time spent on data collection and only 30% on decision-making and patient communication.
The impact of electronic health records on the reduction of face-to-face time with patients due to increased administrative workload.
The potential of multimodal AI in medicine, which processes various data forms like text, images, and numbers, similar to human intelligence.
The introduction of 'Chest Link' by OxyPit, an AI system for autonomous triage of chest X-rays, identifying abnormalities and assisting radiologists.
AI's ability to diagnose eye diseases and predict outcomes from retinal images, even detecting conditions like Parkinson's disease before symptoms appear.
The importance of using AI in conjunction with healthcare professionals, emphasizing that AI should not replace compassionate care.
Google's 'Med-PaM', a medical large language model that passed the US medical licensing exam, showcasing AI's capability in medical knowledge.
The release of the multimodal version of Chat GPT by Open AI, which can analyze medical images and provide patient care advice.
The need for trust, explainability, and randomized clinical trials to safely implement multimodal AI in healthcare.
Survey results indicating public anxiety about healthcare workers relying on AI and the fear of rapid integration without understanding the risks.
The significance of explainable AI in understanding the decision-making process behind AI recommendations in medicine.
The necessity of randomized clinical trials for AI models to ensure their effectiveness and safety in healthcare.
The role of the 'eyeball test' in medicine, emphasizing the importance of patient context in medical diagnosis.
Conor Judge's vision of a future where multimodal AI makes healthcare more efficient, personalized, and accessible globally.
The call to prioritize compassion and understanding in the integration of AI with human healthcare providers.
The conclusion that AI should be a tool to assist doctors in spending more time with patients, enhancing health and happiness.