Michael Barbella, Managing Editor04.04.24
The countdown is on to healthcare’s reimagined future.
Only five years remain until modern medicine reaches the point of no return in its journey toward AI singularity.
Two hundred forty-six weeks, give or take a few.
One thousand, seven hundred thirty-five days.
There is no turning back.
There might be no way to turn back, actually. By 2029, artificial intelligence (AI) may be so vital to medical technology that it will be nearly impossible for the industry (or devices) to function effectively without it.
Should it reach such a crossroads, the medtech industry will need to work collectively to ensure that AI-powered solutions are safely, ethically, and efficiently executed. AI increasingly is becoming an integral part of healthcare, helping providers diagnose and treat diseases, and patients to live longer, healthier lives. Its boundless potential is already transforming the practice of medicine.
“The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone,” billionaire Microsoft co-founder and former chairman/CEO/president Bill Gates wrote last spring on his blog, GatesNotes. “It will change the way people work, learn, travel, get health care, and communicate with each other. Entire industries will reorient around it. Businesses will distinguish themselves by how well they use it.”
Many healthcare businesses are already distinguishing themselves with their AI use, employing it to detect cancer (GRAIL LLC), assess cognitive function (Linus Health), streamline radiology diagnoses (Enlitic Inc.), and measure/analyze atherosclerosis (Cleerly Inc.).
Pharmaceutical/biotech giants Moderna and Pfizer perhaps scored the best use case, leveraging both AI and machine learning (ML) to create a COVID-19 vaccine in record time. Pfizer deployed the technology to rapidly identify the most promising target vaccine compounds, and manage clinical trial data.
Similarly, Moderna’s AI-powered messenger RNA (mRNA) therapy program enabled the firm to analyze vast amounts of mRNA sequences and pinpoint the most potent virus-busting formula. The company also enlisted AI for reviewing pre-clinical data and identifying the mRNAs best suited for possible animal and human testing. Both applications of the technology enabled Moderna to produce a COVID-19 vaccine for human trials in just 42 days.
Moderna is now teaming up with computing giant IBM to advance and accelerate mRNA research and science through quantum computing and AI. Scientists from the two companies will use AI foundation model MoLFormer to predict molecular properties and better understand the characteristics of potential mRNA medicines. MoLFormer will help Moderna optimize lipid nanoparticles, which encapsulate and protect mRNA in its corporeal travels, and mRNA to design pharmaceutical therapies.
The remedies beget by this partnership will still need live test subjects for market approval but AI will eventually infiltrate that arena too, saving healthcare companies precious time and money. Computer scientist, inventor, and futurist Ray Kurzweil envisions biologically simulated humans replacing their flesh-and-blood counterparts in virtual clinical trials. These trials, however, will be far more extensive than current models, as they will involve all potential treatments, from concept to prototype.
“To cure cancer, for example, we’ll simply feed in every possible method that can detect cancer calls...we won’t evaluate them, we’ll just feed in all the ideas we have about each of these possibilities into the computer,” Kurzweil explained during a summit last spring in Los Angeles. “The computer will evaluate all of the many billions of sequences and provide the results. We’ll then test the final product with simulated humans, also very quickly, and we’ll do this for every major health predicament. It will be done 1,000 times faster than conventional methods. And based on our ability to do this, we should be able to overcome most significant health problems by 2029. That is my prediction for passing the Turing test. I came out with that in 1999, and people thought that was crazy...people thought we could do it, but they thought it would take 100 years.”
That timeline now seems crazy. In the quarter century since Kurzweil’s singularity prediction, AI has evolved from sci-fi concept to mind-blowing reality. In just the last half-dozen years or so, artificial intelligence has bested humans in image and speech recognition, reading comprehension, and test-taking.
And it is nowhere close to reaching its full potential.
Scholars believe AI to be in the “Wright brothers stage of its gestation, as its significance in human history has yet to be understood. “I would not believe anything that anybody said about predicting the future because I don’t think we have a good enough imagination to know where things are headed,” Terrence Sejnowski, distinguished professor in UC San Diego’s Neurobiology Department and holder of the Francis Crick Chair at the Salk Institute for Biological Studies, told UC San Diego Magazine last fall. “Whenever you have a new technology, it plays out in ways you can’t imagine.”
Indeed, AI’s progression is boggling the imagination, even at its nascent stage. Artificial intelligence has touched almost every aspect of modern living, from transportation (self-driving cars), travel (trip planning, flight forecasting), lifestyle (thank you, Alexa), customer service (chatbots), entertainment (music/video streaming), and finance (electronic payments).
In healthcare, AI is fostering the rise of precision medicine. ML algorithms now enable providers to create individualized treatment plans based on a patient’s genetic, behavioral, and environmental context.
British pharmatech firm Exscientia plc, for example, uses AI to precisely match patients with existing drugs, accounting for individual biological differences. While still in its early phase, the company’s tailor-made treatment approach has nevertheless helped an 82-year-old man beat blood cancer by exposing a small (live) tissue sample to various drug cocktails.
Exscientia expanded the scope of its precision medicine platform last summer to encompass solid tumors. The company also bolstered its new drug development prowess by opening a 50,000-square-foot laboratory in Vienna, Austria, where novel pharmaceutical candidates are tested on live patient tissue samples and analyzed for efficacy through an ML algorithm.
Exscientia’s approach to new drug development has beget four potential candidates, which began clinical trials in 2023. One of the possibilities is a CDK7 inhibitor that could improve outcomes in head and neck cancer, pancreatic cancer, non-small cell lung cancer, breast carcinoma, and ovarian cancer.
In addition to its own drug development program, Exscientia is lending its AI precision medicine expertise to Sanofi, Bristol Meyers Squibb, Merck, Sumitomo Pharma, and the MD Anderson Cancer Center.
The technology leverages large datasets and complex algorithms to identify patterns and abnormalities indicative of disease or injury. AI and machine learning algorithms enable computers to predict patterns, evaluate trends, calculate accuracy, and optimize processes. Studies have shown that current ML algorithms are more adept at interpreting chest radiographs, diagnosing skin cancer, and directing optimal treatment strategies for sepsis patients in intensive care.
“AI aids in disease detection, acting as a tireless second pair of eyes that complements human experts,” noted Ha Hong, artificial intelligence officer of Endoscopy at Medtronic. “Particularly in scenarios where small lesions might be overlooked towards the end of the day due to fatigue, physicians may find the support of AI invaluable. By increasing the sensitivity of tests without sacrificing specificity, AI enhances diagnostic efficacy, enabling more accurate treatment. A critical extension of this is the early detection of diseases, a decisive factor in improving prognosis. In general, there is ample clinical evidence demonstrating that well-validated AI systems can provide tangible clinical benefits, such as improved disease detection and diagnosis, without compromising specificity.”
Such clinical evidence exists for Medtronic’s GI Genius and AccuRhythm AI, two computer-aided solutions designed to improve cardiac- and gastroenterological-related diagnoses.
GI Genius is a computer-aided detection system that uses advanced AI to identify pre-cancerous and cancerous colorectal polyps (small clumps of cells). Granted U.S. Food and Drug Administration (FDA) de novo clearance in April 2021, GI Genius scans every visual frame taken during a colonoscopy in milliseconds and alerts physicians to the presence of lesions—including small, flat polyps that are easily missed.
U.S. clinical trial results found that GI Genius reduced missed colorectal polyps by 50% compared to a standard colonoscopy. The device also generated an absolute 14% adenoma detection rate (ADR) increase versus colonoscopy alone for both flat and polyploid lesions (a 42% and 36% spike, respectively), data show.
“GI Genius has been trained using millions of images that cover the entire spectrum of real-world polyps, derived from over 20 clinical centers worldwide,” Hong explained. “The detection of these polyps is crucial, as they can lead to the second most common cause of cancer-related death in the United States. However, if caught early, this type of cancer is among the most curable. This is where the GI Genius system truly proves its worth. The system is supported by substantial clinical evidence and has been shown to increase the Adenoma Detection Rate (ADR) by up to 14.4%. This is significant given that each 1% increase in ADR decreases the risk of interval cancer by 3%.”
Similar encouraging proof exists for Medtronic’s AccuRhythm AI algorithms. Approved by the FDA in July 2021, AccuRhythm AI is designed to improve the accuracy of heart rhythm data collected by the company’s LINQ II insertable cardiac monitor (ICM). The algorithms address the two most common ICM false alerts: atrial fibrillation (AF), an irregular or rapid rhythm in the heart’s upper chambers; and asystole, a long pause between heartbeats. Medtronic developed the AccuRhythm AI platform and initial algorithms using a diverse and debiased database of more than 1 million electrocardiogram heart rhythm episodes.
Studies have shown the AccuRhythm AI algorithms can significantly improve LINQ II ICM alert accuracy. Validation data indicate the AccuRhythm AF algorithm reduced LINQ II ICM false AF alerts by 74.1% and preserved 99.3% of true AF alerts, according to results presented at Heart Rhythm 2021. The data also demonstrate the Pause algorithm reduced LINQ II ICM false pause alerts by 97.4% and preserved 100% of true pause alerts.
“In general, there is ample clinical evidence demonstrating that well-validated AI systems can provide tangible clinical benefits, such as improved disease detection and diagnosis, without compromising specificity,” Hong said. “In today’s world where AI is transforming every industry, its potential in healthcare is especially remarkable. It’s not just about the technology, but the meaningful difference it can make in patient care and outcomes.”
Sometimes, though, it’s about both. Case in point: Eko Health Inc.’s digital stethoscopes and AI-powered analysis platform are not only improving cardiac-related disease detection, they also are giving a somewhat antiquated healthcare staple a 21st-century overhaul.
The digital stethoscopes developed by the Emeryville, Calif.-based firm are designed to help clinicians better detect conditions such as AF and heart murmurs during routine health exams. Its latest FDA-cleared product, the CORE 500, features a three-lead electrocardiogram (ECG), a full-color display, high-fidelity audio experience, and AI software. Able to flag abnormalities in seconds, the CORE 500 is equipped with up to 40x sound amplification, active noise cancellation, three audio filter modes, and Eko’s newest audio innovation, TrueSound.
The CORE 500 is compatible with Eko’s SENSORA cardiac disease detection platform, and is intended for use across various clinical specialties, including virtual care, home care, emergency medicine, urgent care, primary care, and structural heart.
“We think of our digital stethoscopes as an intelligent microphone put on a patient’s chest. We’re taking 15 seconds of heart sounds, sending it up to the Cloud, and in five seconds of processing time, determining whether the sounds the clinician is hearing are suspicious,” Eko co-founder Jason Bellet told MPO. “We were talking to cardiologists and they told us that so many heart conditions are being missed in patients because doctors are not hearing the murmurs or irregular heartbeats. So we thought, if we could build a smart stethoscope, we could change the way we detect heart disease.”
Eko’s stethoscopes are certainly changing cardiac disease detection methods, largely for the better. Its structural heart murmur algorithm has been found to more than double valvular heart disease detection compared to the standard method, 94.1% vs. 41.2%, according to the company. Likewise, Eko’s AF detection algorithm has been clinically validated to perform at 100% sensitivity and 96.2% specificity.
Clinical data also show that Eko’s digital stethoscope and its FDA-cleared structural murmur detection algorithm were twice as good as an analog device in identifying previously undiscovered valvular heart disease. The AI method performed at 94.1% sensitivity and 84.5% specificity, compared with the standard method of 41.2% sensitivity and 95.5% specificity. In addition, the AI method identified 22 patients with moderate or greater VHD who were previously undiagnosed, while the standard method identified eight previously undiagnosed patients.
“We often compare our AI to ‘Shazam for heartbeats,’” Bellet said. “We’re not replacing the physician. We’re helping them hear very quiet sounds they may otherwise miss. We have a pipeline of new detection algorithms in development over the next few years, including an algorithm for pulmonary hypertension which is often detected far too late.”
And that’s just the tip of the proverbial AI iceberg.
Healthcare’s increased digitization is fostering innovation in ways never before imagined. Medtech companies are racing to market with solutions that leverage big data, AI, and the Internet of Things to streamline clinical workflow, improve analytics, prolong patient monitoring, and boost diagnostic intellect.
GE HealthCare is outpacing its rivals with AI-powered remedies, having secured 58 such 510(k) clearances/authorizations from the FDA as of October 2023, according to the agency. Many of those innovations aim to improve diagnostic accuracy through image recognition technology.
The company is using AI algorithms, for example, to reduce magnetic resonance imaging (MRI) scan time by up to 83% (Sonic DL) and capture diagnostic-quality cardiac ultrasound images (Caption Guidance). Other deep-learning based computer programs are trained to automatically contour and label at-risk organs from computed tomography (CT) images (Auto Segmentation); and provide increased small, low-contrast lesion detectability compared to conventional Time-of-Flight PET/CT scanners (Precision DL).
Philips N.V. has developed a host of AI-powered imaging solutions as well. Like its rival, the company has developed a CT reconstruction solution to simultaneously reduce dose, lower noise, and improve low-contrast detectability with fast reconstruction speed. Precise Image uses a supervised learning process and convolutional neural networks to reproduce the CT image appearance traditionally associated with high-dose filtered back-projection reconstruction.
“We have a heritage of using AI for a long time. Without question, everyone tends to think of AI algorithms for diagnosis and treatment,” stated Shez Partovi, chief innovation and strategy officer at Philips. “To diagnose MS, for example, you have to look at an MRI scan and find lesions. We use AI embedded in software to support doctors to find those lesions. It’s an advanced visualization aid that can help increase the chances of finding [MS] lesions by 44%. Somewhere today, a radiologist is looking for lesions on an MRI and artificial intelligence is helping them find it faster.”
Philips’ AI solutions are also speeding up patient positioning during CT scans (CT Precise Position) and provide suggestions based on the protocol used the most for a given clinical indication (MR AI Protocol Assistant).
In addition, Philips has developed software for its Ingenia MR-Radiation Therapy simulation platform, which enables MR-based radiotherapy planning for head, neck, and brain tumors without the need for CT (MRCAT Brain, MRCAT Head and Neck).
The EPIQ CVx and CVxi, on the other hand, are cardiac ultrasound solutions for diagnostic and interventional echo exams across a wide range of patients. 3D Auto RV uses anatomical models to provide semi-automated measurements for right ventricle volume, ejection fraction, and other relevant calculations for cardiac anatomical and functional quantification. The parameters for the models were trained using AI technologies. 3D AutoRV was clinically validated on a typical patient population whose cardiac images covered a range of heart sizes and image quality.
“A lot of our AI solutions the patient may never see. We have a clinical application we can use with our MRI scanners that enables clinicians to complete scanning three times faster,” Partovi said.
“What used to take nine minutes now takes three minutes. Nobody may ever know about that...it’s an embedded feature that helps improve productivity and throughput, and extends the life and value of the hardware. Diagnosis is an important aspect of AI solutions, but embedded features are being incorporated more, too.”
Those features are helping to democratize healthcare and enhance surgery. In 2021, Johnson & Johnson MedTech teamed up with the West Virginia University (WVU) Cancer Institute and WVU Medicine to improve lung cancer screening access via the mobile Lung Cancer Screening Unit System (LUCAS), the only fully mobile AI-enabled CT unit in the U.S. for low-dose lung cancer screening that does not require an outside power source. Housed in an 18-wheeler truck, LUCAS can screen up to 20 people daily.
In the surgical theater, J&J is working with Silicon Valley tech firm NVIDIA to better integrate AI in surgical procedures. The partnership seeks to enhance real-time analysis and broaden the use of AI algorithms in surgical decision-making, education, and OR collaboration.
J&J MedTech will use NVIDIA’s IGX edge computing platform and Holoscan edge AI to create infrastructure for deploying AI-powered software applications in the OR. The company will also leverage AI technologies to extend its open ecosystem for surgery. NVIDIA’s purpose-built solutions are designed to help accelerate innovation throughout the surgical ecosystem and speed development and deployment of AI-driven applications in a secure, scalable manner.
“One area where we see a lot of opportunity for AI is in its ability to create a better surgical experience,” noted Shan Jegatheeswaran, Global Head of Digital, MedTech, at J&J. “We’re hearing that surgeons increasingly want real-time critical decision support before, during, and after surgery so they can take the best possible action for the patient. We can now use AI to deliver preoperative planning guidance, intraoperative decision guidance, and postoperative case review, training, and education. For example, AI can be used in real-time to augment surgical video for clinical guidance, remove information from surgical images to reduce risk, or inform postoperative patient guidance for improved healing.”
“There is a nascent opportunity that requires the right technologies, collaboration, regulatory guidance, and close change management and partnership within hospitals to understand necessary infrastructure and support,” he continued. “Ultimately, the near-term goal is to offer more personalization and more insights in secure environments suited for surgery.”
Only five years remain until modern medicine reaches the point of no return in its journey toward AI singularity.
Two hundred forty-six weeks, give or take a few.
One thousand, seven hundred thirty-five days.
There is no turning back.
There might be no way to turn back, actually. By 2029, artificial intelligence (AI) may be so vital to medical technology that it will be nearly impossible for the industry (or devices) to function effectively without it.
Should it reach such a crossroads, the medtech industry will need to work collectively to ensure that AI-powered solutions are safely, ethically, and efficiently executed. AI increasingly is becoming an integral part of healthcare, helping providers diagnose and treat diseases, and patients to live longer, healthier lives. Its boundless potential is already transforming the practice of medicine.
“The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone,” billionaire Microsoft co-founder and former chairman/CEO/president Bill Gates wrote last spring on his blog, GatesNotes. “It will change the way people work, learn, travel, get health care, and communicate with each other. Entire industries will reorient around it. Businesses will distinguish themselves by how well they use it.”
Many healthcare businesses are already distinguishing themselves with their AI use, employing it to detect cancer (GRAIL LLC), assess cognitive function (Linus Health), streamline radiology diagnoses (Enlitic Inc.), and measure/analyze atherosclerosis (Cleerly Inc.).
Pharmaceutical/biotech giants Moderna and Pfizer perhaps scored the best use case, leveraging both AI and machine learning (ML) to create a COVID-19 vaccine in record time. Pfizer deployed the technology to rapidly identify the most promising target vaccine compounds, and manage clinical trial data.
Similarly, Moderna’s AI-powered messenger RNA (mRNA) therapy program enabled the firm to analyze vast amounts of mRNA sequences and pinpoint the most potent virus-busting formula. The company also enlisted AI for reviewing pre-clinical data and identifying the mRNAs best suited for possible animal and human testing. Both applications of the technology enabled Moderna to produce a COVID-19 vaccine for human trials in just 42 days.
Moderna is now teaming up with computing giant IBM to advance and accelerate mRNA research and science through quantum computing and AI. Scientists from the two companies will use AI foundation model MoLFormer to predict molecular properties and better understand the characteristics of potential mRNA medicines. MoLFormer will help Moderna optimize lipid nanoparticles, which encapsulate and protect mRNA in its corporeal travels, and mRNA to design pharmaceutical therapies.
The remedies beget by this partnership will still need live test subjects for market approval but AI will eventually infiltrate that arena too, saving healthcare companies precious time and money. Computer scientist, inventor, and futurist Ray Kurzweil envisions biologically simulated humans replacing their flesh-and-blood counterparts in virtual clinical trials. These trials, however, will be far more extensive than current models, as they will involve all potential treatments, from concept to prototype.
“To cure cancer, for example, we’ll simply feed in every possible method that can detect cancer calls...we won’t evaluate them, we’ll just feed in all the ideas we have about each of these possibilities into the computer,” Kurzweil explained during a summit last spring in Los Angeles. “The computer will evaluate all of the many billions of sequences and provide the results. We’ll then test the final product with simulated humans, also very quickly, and we’ll do this for every major health predicament. It will be done 1,000 times faster than conventional methods. And based on our ability to do this, we should be able to overcome most significant health problems by 2029. That is my prediction for passing the Turing test. I came out with that in 1999, and people thought that was crazy...people thought we could do it, but they thought it would take 100 years.”
That timeline now seems crazy. In the quarter century since Kurzweil’s singularity prediction, AI has evolved from sci-fi concept to mind-blowing reality. In just the last half-dozen years or so, artificial intelligence has bested humans in image and speech recognition, reading comprehension, and test-taking.
And it is nowhere close to reaching its full potential.
Scholars believe AI to be in the “Wright brothers stage of its gestation, as its significance in human history has yet to be understood. “I would not believe anything that anybody said about predicting the future because I don’t think we have a good enough imagination to know where things are headed,” Terrence Sejnowski, distinguished professor in UC San Diego’s Neurobiology Department and holder of the Francis Crick Chair at the Salk Institute for Biological Studies, told UC San Diego Magazine last fall. “Whenever you have a new technology, it plays out in ways you can’t imagine.”
Indeed, AI’s progression is boggling the imagination, even at its nascent stage. Artificial intelligence has touched almost every aspect of modern living, from transportation (self-driving cars), travel (trip planning, flight forecasting), lifestyle (thank you, Alexa), customer service (chatbots), entertainment (music/video streaming), and finance (electronic payments).
In healthcare, AI is fostering the rise of precision medicine. ML algorithms now enable providers to create individualized treatment plans based on a patient’s genetic, behavioral, and environmental context.
British pharmatech firm Exscientia plc, for example, uses AI to precisely match patients with existing drugs, accounting for individual biological differences. While still in its early phase, the company’s tailor-made treatment approach has nevertheless helped an 82-year-old man beat blood cancer by exposing a small (live) tissue sample to various drug cocktails.
Exscientia expanded the scope of its precision medicine platform last summer to encompass solid tumors. The company also bolstered its new drug development prowess by opening a 50,000-square-foot laboratory in Vienna, Austria, where novel pharmaceutical candidates are tested on live patient tissue samples and analyzed for efficacy through an ML algorithm.
Exscientia’s approach to new drug development has beget four potential candidates, which began clinical trials in 2023. One of the possibilities is a CDK7 inhibitor that could improve outcomes in head and neck cancer, pancreatic cancer, non-small cell lung cancer, breast carcinoma, and ovarian cancer.
In addition to its own drug development program, Exscientia is lending its AI precision medicine expertise to Sanofi, Bristol Meyers Squibb, Merck, Sumitomo Pharma, and the MD Anderson Cancer Center.
‘Shazam for Heartbeats’
Along with its potential as a powerful new drug development tool, artificial intelligence also is proving its worth in disease detection and diagnosis.The technology leverages large datasets and complex algorithms to identify patterns and abnormalities indicative of disease or injury. AI and machine learning algorithms enable computers to predict patterns, evaluate trends, calculate accuracy, and optimize processes. Studies have shown that current ML algorithms are more adept at interpreting chest radiographs, diagnosing skin cancer, and directing optimal treatment strategies for sepsis patients in intensive care.
“AI aids in disease detection, acting as a tireless second pair of eyes that complements human experts,” noted Ha Hong, artificial intelligence officer of Endoscopy at Medtronic. “Particularly in scenarios where small lesions might be overlooked towards the end of the day due to fatigue, physicians may find the support of AI invaluable. By increasing the sensitivity of tests without sacrificing specificity, AI enhances diagnostic efficacy, enabling more accurate treatment. A critical extension of this is the early detection of diseases, a decisive factor in improving prognosis. In general, there is ample clinical evidence demonstrating that well-validated AI systems can provide tangible clinical benefits, such as improved disease detection and diagnosis, without compromising specificity.”
Such clinical evidence exists for Medtronic’s GI Genius and AccuRhythm AI, two computer-aided solutions designed to improve cardiac- and gastroenterological-related diagnoses.
GI Genius is a computer-aided detection system that uses advanced AI to identify pre-cancerous and cancerous colorectal polyps (small clumps of cells). Granted U.S. Food and Drug Administration (FDA) de novo clearance in April 2021, GI Genius scans every visual frame taken during a colonoscopy in milliseconds and alerts physicians to the presence of lesions—including small, flat polyps that are easily missed.
U.S. clinical trial results found that GI Genius reduced missed colorectal polyps by 50% compared to a standard colonoscopy. The device also generated an absolute 14% adenoma detection rate (ADR) increase versus colonoscopy alone for both flat and polyploid lesions (a 42% and 36% spike, respectively), data show.
“GI Genius has been trained using millions of images that cover the entire spectrum of real-world polyps, derived from over 20 clinical centers worldwide,” Hong explained. “The detection of these polyps is crucial, as they can lead to the second most common cause of cancer-related death in the United States. However, if caught early, this type of cancer is among the most curable. This is where the GI Genius system truly proves its worth. The system is supported by substantial clinical evidence and has been shown to increase the Adenoma Detection Rate (ADR) by up to 14.4%. This is significant given that each 1% increase in ADR decreases the risk of interval cancer by 3%.”
Similar encouraging proof exists for Medtronic’s AccuRhythm AI algorithms. Approved by the FDA in July 2021, AccuRhythm AI is designed to improve the accuracy of heart rhythm data collected by the company’s LINQ II insertable cardiac monitor (ICM). The algorithms address the two most common ICM false alerts: atrial fibrillation (AF), an irregular or rapid rhythm in the heart’s upper chambers; and asystole, a long pause between heartbeats. Medtronic developed the AccuRhythm AI platform and initial algorithms using a diverse and debiased database of more than 1 million electrocardiogram heart rhythm episodes.
Studies have shown the AccuRhythm AI algorithms can significantly improve LINQ II ICM alert accuracy. Validation data indicate the AccuRhythm AF algorithm reduced LINQ II ICM false AF alerts by 74.1% and preserved 99.3% of true AF alerts, according to results presented at Heart Rhythm 2021. The data also demonstrate the Pause algorithm reduced LINQ II ICM false pause alerts by 97.4% and preserved 100% of true pause alerts.
“In general, there is ample clinical evidence demonstrating that well-validated AI systems can provide tangible clinical benefits, such as improved disease detection and diagnosis, without compromising specificity,” Hong said. “In today’s world where AI is transforming every industry, its potential in healthcare is especially remarkable. It’s not just about the technology, but the meaningful difference it can make in patient care and outcomes.”
Sometimes, though, it’s about both. Case in point: Eko Health Inc.’s digital stethoscopes and AI-powered analysis platform are not only improving cardiac-related disease detection, they also are giving a somewhat antiquated healthcare staple a 21st-century overhaul.
The digital stethoscopes developed by the Emeryville, Calif.-based firm are designed to help clinicians better detect conditions such as AF and heart murmurs during routine health exams. Its latest FDA-cleared product, the CORE 500, features a three-lead electrocardiogram (ECG), a full-color display, high-fidelity audio experience, and AI software. Able to flag abnormalities in seconds, the CORE 500 is equipped with up to 40x sound amplification, active noise cancellation, three audio filter modes, and Eko’s newest audio innovation, TrueSound.
The CORE 500 is compatible with Eko’s SENSORA cardiac disease detection platform, and is intended for use across various clinical specialties, including virtual care, home care, emergency medicine, urgent care, primary care, and structural heart.
“We think of our digital stethoscopes as an intelligent microphone put on a patient’s chest. We’re taking 15 seconds of heart sounds, sending it up to the Cloud, and in five seconds of processing time, determining whether the sounds the clinician is hearing are suspicious,” Eko co-founder Jason Bellet told MPO. “We were talking to cardiologists and they told us that so many heart conditions are being missed in patients because doctors are not hearing the murmurs or irregular heartbeats. So we thought, if we could build a smart stethoscope, we could change the way we detect heart disease.”
Eko’s stethoscopes are certainly changing cardiac disease detection methods, largely for the better. Its structural heart murmur algorithm has been found to more than double valvular heart disease detection compared to the standard method, 94.1% vs. 41.2%, according to the company. Likewise, Eko’s AF detection algorithm has been clinically validated to perform at 100% sensitivity and 96.2% specificity.
Clinical data also show that Eko’s digital stethoscope and its FDA-cleared structural murmur detection algorithm were twice as good as an analog device in identifying previously undiscovered valvular heart disease. The AI method performed at 94.1% sensitivity and 84.5% specificity, compared with the standard method of 41.2% sensitivity and 95.5% specificity. In addition, the AI method identified 22 patients with moderate or greater VHD who were previously undiagnosed, while the standard method identified eight previously undiagnosed patients.
“We often compare our AI to ‘Shazam for heartbeats,’” Bellet said. “We’re not replacing the physician. We’re helping them hear very quiet sounds they may otherwise miss. We have a pipeline of new detection algorithms in development over the next few years, including an algorithm for pulmonary hypertension which is often detected far too late.”
And that’s just the tip of the proverbial AI iceberg.
Healthcare’s increased digitization is fostering innovation in ways never before imagined. Medtech companies are racing to market with solutions that leverage big data, AI, and the Internet of Things to streamline clinical workflow, improve analytics, prolong patient monitoring, and boost diagnostic intellect.
GE HealthCare is outpacing its rivals with AI-powered remedies, having secured 58 such 510(k) clearances/authorizations from the FDA as of October 2023, according to the agency. Many of those innovations aim to improve diagnostic accuracy through image recognition technology.
The company is using AI algorithms, for example, to reduce magnetic resonance imaging (MRI) scan time by up to 83% (Sonic DL) and capture diagnostic-quality cardiac ultrasound images (Caption Guidance). Other deep-learning based computer programs are trained to automatically contour and label at-risk organs from computed tomography (CT) images (Auto Segmentation); and provide increased small, low-contrast lesion detectability compared to conventional Time-of-Flight PET/CT scanners (Precision DL).
Philips N.V. has developed a host of AI-powered imaging solutions as well. Like its rival, the company has developed a CT reconstruction solution to simultaneously reduce dose, lower noise, and improve low-contrast detectability with fast reconstruction speed. Precise Image uses a supervised learning process and convolutional neural networks to reproduce the CT image appearance traditionally associated with high-dose filtered back-projection reconstruction.
“We have a heritage of using AI for a long time. Without question, everyone tends to think of AI algorithms for diagnosis and treatment,” stated Shez Partovi, chief innovation and strategy officer at Philips. “To diagnose MS, for example, you have to look at an MRI scan and find lesions. We use AI embedded in software to support doctors to find those lesions. It’s an advanced visualization aid that can help increase the chances of finding [MS] lesions by 44%. Somewhere today, a radiologist is looking for lesions on an MRI and artificial intelligence is helping them find it faster.”
Philips’ AI solutions are also speeding up patient positioning during CT scans (CT Precise Position) and provide suggestions based on the protocol used the most for a given clinical indication (MR AI Protocol Assistant).
In addition, Philips has developed software for its Ingenia MR-Radiation Therapy simulation platform, which enables MR-based radiotherapy planning for head, neck, and brain tumors without the need for CT (MRCAT Brain, MRCAT Head and Neck).
The EPIQ CVx and CVxi, on the other hand, are cardiac ultrasound solutions for diagnostic and interventional echo exams across a wide range of patients. 3D Auto RV uses anatomical models to provide semi-automated measurements for right ventricle volume, ejection fraction, and other relevant calculations for cardiac anatomical and functional quantification. The parameters for the models were trained using AI technologies. 3D AutoRV was clinically validated on a typical patient population whose cardiac images covered a range of heart sizes and image quality.
“A lot of our AI solutions the patient may never see. We have a clinical application we can use with our MRI scanners that enables clinicians to complete scanning three times faster,” Partovi said.
“What used to take nine minutes now takes three minutes. Nobody may ever know about that...it’s an embedded feature that helps improve productivity and throughput, and extends the life and value of the hardware. Diagnosis is an important aspect of AI solutions, but embedded features are being incorporated more, too.”
Those features are helping to democratize healthcare and enhance surgery. In 2021, Johnson & Johnson MedTech teamed up with the West Virginia University (WVU) Cancer Institute and WVU Medicine to improve lung cancer screening access via the mobile Lung Cancer Screening Unit System (LUCAS), the only fully mobile AI-enabled CT unit in the U.S. for low-dose lung cancer screening that does not require an outside power source. Housed in an 18-wheeler truck, LUCAS can screen up to 20 people daily.
In the surgical theater, J&J is working with Silicon Valley tech firm NVIDIA to better integrate AI in surgical procedures. The partnership seeks to enhance real-time analysis and broaden the use of AI algorithms in surgical decision-making, education, and OR collaboration.
J&J MedTech will use NVIDIA’s IGX edge computing platform and Holoscan edge AI to create infrastructure for deploying AI-powered software applications in the OR. The company will also leverage AI technologies to extend its open ecosystem for surgery. NVIDIA’s purpose-built solutions are designed to help accelerate innovation throughout the surgical ecosystem and speed development and deployment of AI-driven applications in a secure, scalable manner.
“One area where we see a lot of opportunity for AI is in its ability to create a better surgical experience,” noted Shan Jegatheeswaran, Global Head of Digital, MedTech, at J&J. “We’re hearing that surgeons increasingly want real-time critical decision support before, during, and after surgery so they can take the best possible action for the patient. We can now use AI to deliver preoperative planning guidance, intraoperative decision guidance, and postoperative case review, training, and education. For example, AI can be used in real-time to augment surgical video for clinical guidance, remove information from surgical images to reduce risk, or inform postoperative patient guidance for improved healing.”
“There is a nascent opportunity that requires the right technologies, collaboration, regulatory guidance, and close change management and partnership within hospitals to understand necessary infrastructure and support,” he continued. “Ultimately, the near-term goal is to offer more personalization and more insights in secure environments suited for surgery.”