- AI Grand Rounds
- Posts
- Exciting New Addition to AI Grand Rounds
Exciting New Addition to AI Grand Rounds
Empowering individuals with knowledge to make the right decision!
Fellow Healthcare Champions,
Are you overwhelmed by all the fluff and hype around AI and not sure how to identify meaningful information? We get it. As busy clinicians ourselves, our newsletter, "AI Grand Rounds," is here to provide clinically relevant AI information.
No matter who you are—a healthcare provider, inventor, investor, or curious reader—we PROMISE one thing: you will always find information that is relevant, meaningful, and balanced.
Let’s join our journey to share clinically meaningful and relevant knowledge on healthcare AI.
Your fellow Physicians,
Table of Contents
🚨 Exciting New Addition 🚨
CME/CNE Certified AI Education
Over the last many months, we have been working closely with colleagues and friends on educating clinically meaningful use of AI, harnessing its power, and learning how it can lead to new threats. To help share the knowledge and teach colleagues, the AI Grand Rounds team is developing a CME/CNE accredited Journal Club focusing on AI in healthcare.
As clinicians, we know that every minute of our day counts, so doctors and nurses can get FREE CME/CNE credits for attending the meeting.
Please join us for our inaugural AI Grand Rounds journal club with Professor Jiapeng Huang, MD, PhD, FASA, FASE, D.ABA, and D.NBE. We will discuss pivotal clinical studies on the use of AI in Pre-operative care.
Register here ➡ ➡ ➡ https://lnkd.in/gpMyZjyG
🚨 Pulse of Innovation 🚨
Breaking news in the healthcare AI
Generative AI-powered Robots in Healthcare: Are They the Future?
The 2024 Nobel Prize in Physics has been awarded to John Hopfield and Geoffrey Hinton for their pioneering work in artificial intelligence, recognizing their foundational discoveries that enabled machine learning with artificial neural networks. Their groundbreaking research in the 1980s laid the groundwork for today's AI technologies, from image recognition to language processing.
John Hopfield's Contributions
In 1982, John Hopfield introduced the Hopfield network, a type of recurrent artificial neural network capable of storing and reconstructing patterns in data. This model applied principles from condensed matter physics to create a neural network that could hold memory and recognize patterns. Hopfield's innovative approach demonstrated how computer circuits could mimic brain functions, laying the foundation for future advancements in machine learning and artificial intelligence.
Geoffrey Hinton's Innovations
Geoffrey Hinton developed the Boltzmann machine in 1985, a groundbreaking deep learning model that could autonomously recognize patterns in data. Hinton's contributions extended beyond the Boltzmann machine, as he later published influential papers on backpropagation, a crucial learning process used in modern machine learning systems. His work significantly advanced the field of artificial neural networks, earning him the moniker "godfather of AI".
Impact on Modern AI
The pioneering work of Hopfield and Hinton has had a profound impact on modern artificial intelligence, enabling advancements across various scientific disciplines. Their neural network models form the building blocks of machine learning, with applications ranging from particle physics to astrophysics. The Boltzmann machine, developed by Hinton, laid the groundwork for current generative AI models like ChatGPT. Their contributions have revolutionized image recognition, language processing, and pattern analysis, leading to breakthroughs in fields such as material science and computer vision. The Nobel Committee emphasized that their discoveries have provided humanity with a powerful new tool, highlighting the far-reaching implications of their research on contemporary technology and scientific exploration.
Hinton's AI Concerns
Geoffrey Hinton, a prominent figure in AI, has raised concerns about the rapid development of AI and its potential risks. In 2023, he left Google to openly address these issues, particularly the potential inability to distinguish AI-generated content from reality. Hinton opposes AI use in military applications and has expressed some regret about his life's work. He believes that AI's impact will surpass our intellectual capabilities and has highlighted the transformative potential and ethical challenges of advanced AI systems, comparing it to the Industrial Revolution after receiving the Nobel Prize.
🧑🏼🔬 Bench to Bedside👨🏽🔬
Developments in healthcare AI research and innovations
This AI-powered Artificial tongue can tell you your morning coffee blend!
Imagine knowing definitely about how watered down your milk is, how fresh your food is or even what coffee you are consuming in the morning? Scientists from Penn State developed an electronic tongue that can make it possible.
This electronic tongue contains a graphene based ion sensitive fieldeffect transistor, which is a conductive device that can detect chemical ions. Each sensor can detect different types of chemicals as opposed to the human tongue where each receptor is specific for one taste sensation.
This device is linked to artificial neural networks that imitate the gustatory cortex in the human brain and can detect food samples. The researchers published their results in Nature.
Key Findings:
The neural network provided 20 specific parameters to assess how a sample liquid interacts with the sensors’ electrical properties. This device could identify differences in similar liquids with over 80% accuracy in about a minute.
The researchers used Shapley additive explanations, which allowed them to understand the neural network—what it was thinking after making a decision—hence enhancing the transparency in its functioning.
The main limitation of this tongue is that it can only sense the data on which it is trained.
Implications of the Artificial Tongue:
AI has made significant strides in enhancing vision, touch, and now taste. This research has major implications for ensuring food quality and safety.
Graphene’s use as a chemical sensor was long known; however, its tiny variations limited its use. The team behind the artificial tongue enabled AI to correctly identify liquids despite these variations.
The electronic tongue comprises a graphene-based ion-sensitive field-effect transistor linked to an artificial neural network, trained on various datasets. This is located in the top right of the device. Credit: Provided by the Das Lab
🧑🏽⚕️ AI in Clinic 🏥
Developments in healthcare AI research and innovations
AI - Jargon Buster
Data
Data is any information that can be analysed to gain insights. It can be in the form of numbers, text, symbols or multimedia such as images, videos, sounds and maps.
In healthcare, data can encompass patient records, clinical studies and real-time health monitoring outputs.
Big data
Extremely large and rapidly growing collections of diverse data types, including structured and unstructured, which are so complex that traditional data processing software cannot handle them.
In healthcare, it can include genetic data, medical history, and lifestyle factors to support personalized medicine.
Structured data
Data is organized and formatted in a specific way, making it easily readable and understandable by humans and machines. This allows viewers to recognize the type of data they are looking at immediately.
An EHR that includes fields for name, age, blood pressure, and diagnosis codes is structured data.
Synthetic data
Artificially generated data produced by computer algorithms or simulations, designed to mimic the patterns and characteristics of real-world data and often used as an alternative to actual data.
Test data
This is a final check of an unseen dataset to confirm that the ML algorithm was trained effectively and validate that the model can make accurate predictions.
Training data
The data is used to train machine learning models. Curated training datasets are fed to machine learning algorithms to teach them how to make predictions or perform a desired task.
Unstructured data
Data that does not have a predefined structure. It is unsorted and a vast information collection.
Medical notes, audio recordings of patient interactions, and images from various diagnostic procedures from unstructured data.
Validation data
Data not included in the model's training set allows data scientists to evaluate how well the model makes predictions based on new data previously unseen by the model when it was being trained.
🤖 Patient First, AI Second🤖
Ethical and Regulatory Landscape of Healthcare AI
Addressing Reporting Gaps in FDA- Approved AI/ML Medical Devices: A call for Transparency, Inclusivity, and Equality in Healthcare.
The FDA launched the Medical Device Development Tools program with the goal of " facilitating device development, timely evaluation of medical devices, and promoting innovation.” However, there are reporting gaps in FDA-approved AI Medical Devices.
With AI being clinically integrated, the end users need to evaluate the applicability of these devices to their settings and assess bias and risk due to these reporting gaps.
An article from Nature Digital reviews data related to 692 FDA-approved AI/ ML devices from 1995 to 2023 to examine safety reporting, transparency, and socio-demographic representation.
According to this review, despite an increasing number of FDA approvals, the data is inconsistent regarding transparency and safety issues.
3.6% reported race or ethnicity
99.1% reported no socioeconomic data
86.1% did not report the age of the study subjects with paediatric and geriatric age groups being severely underrepresented. The risk of generalizing findings in clinical settings can be risky in these age groups.
1.9% included a link to a scientific publication with safety and efficacy data.
9% contained a prospective study for postmarked surveillance.
Radiology, Neurology, and Cardiology were disproportionately represented compared to the others. A balanced representation among these specialities is needed to promote clinical equality.
20.5% of devices provided statements on potential risks to end users, only 13 approval documents included a corresponding published validation study.
The following recommendations are to be considered by the FDA and other regulatory authorities to ensure safety and reduce algorithmic bias and health disparity.
Clear representation standards include race, ethnicity, age, gender, and socio-economic status of the studied populations.
Inclusive validation criteria for diverse and representative populations.
Transparency requirements for the functions of the software.
Post-market surveillance for equity.
Inter-disciplinary review channels within the FDA with experts in healthcare, data science, and ethics departments.
Incentives for inclusivity in these devices.
Disclaimer: This newsletter contains opinions and speculations and is based solely on public information. It should not be considered medical, business, or investment advice. This newsletter's banner and other images are created for illustrative purposes only. All brand names, logos, and trademarks are the property of their respective owners. At the time of publication of this newsletter, the author has no business relationships, affiliations, or conflicts of interest with any of the companies mentioned except as noted. ** OPINIONS ARE PERSONAL AND NOT THOSE OF ANY AFFILIATED ORGANIZATIONS!
Reply