1950-1961
February 1, 2022
Supervised Machine Learning
February 1, 2022Otorhinolaryngology-Ophthalmology
Digital otoscopy with AI diagnostic support: making the diagnosis of ear disease more accessible
The HearX group released a beta version of the HearScope device, which includes a smartphone-compatible video otoscope for capturing images and videos of the ear canal and tympanic membrane at a minimal price. In seconds, the AI algorithm classifies the image using the most prevailing ear disease categories. This implementation of AI is a world first. It appears to have the potential to improve access to diagnostic knowledge in areas where audiology and ENT facilities are scarce or nonexistent. The proposal also aims to diversify markets where audiology facilities are well-established by inexpensively introducing video otoscopy. The AI classification function also provides a favorable second opinion, cross-checking the clinician diagnosis for practitioners working independently and without access to referral networks.(1)
The HearX group released a beta version of the HearScope device, which includes a smartphone-compatible video otoscope for capturing images and videos of the ear canal and tympanic membrane at a minimal price. In seconds, the AI algorithm classifies the image using the most prevailing ear disease categories. This implementation of AI is a world first. It appears to have the potential to improve access to diagnostic knowledge in areas where audiology and ENT facilities are scarce or nonexistent.
The proposal also aims to diversify markets where audiology facilities are well-established by inexpensively introducing video otoscopy. The AI classification function also provides a favorable second opinion, cross-checking the clinician diagnosis for practitioners working independently and without access to referral networks.(1)
First ENT Navigation Technology Powered by Artificial Intelligence
Acclarent, Inc., a medical branch of the Johnson & Johnson Medical Devices Companies, and a leader in developing minimally-invasive Ear, Nose & Throat technologies, announced the launch of its first AI-powered technology to simplify surgical planning and provide real-time feedback. The new software components are TruSeg™ and TruPath™, for use with the TruDi® Navigation System, and a machine-learning algorithm to provide reliable and efficient image-guided preoperative planning and navigation for procedures like minimally invasive sinus surgery. TruDi® was designed for ENT surgeons and enabled software-guided navigation to deliver real-time 3D guidance, anatomical mapping, and surgical insights. TruSeg™ enables surgeons to apply automatic fragmentation of the patient’s anatomical structures based on CT scans. This feature allows precise labeling of anatomical structures to function as alert beacons when approaching the anatomical landmark during procedures. Lastly, TruPath™ calculates and presents the shortest valid path that does not cross bone.(2)
TruDi® was designed for ENT surgeons and enabled software-guided navigation to deliver real-time 3D guidance, anatomical mapping, and surgical insights. TruSeg™ enables surgeons to apply automatic fragmentation of the patient’s anatomical structures based on CT scans. This feature allows precise labeling of anatomical structures to function as alert beacons when approaching the anatomical landmark during procedures. Lastly, TruPath™ calculates and presents the shortest valid path that does not cross bone.(2)
AI-based Prediction Model for Obstructive Sleep Apnea Surgery
Seoul National University Hospital (SNUH) team has developed an artificial intelligence-based system that can predict the success rate of obstructive sleep apnea (OSA) surgery. The Professor Kim Hyun-Jik of the Department of Otolaryngology developed the platform by analyzing 163 patients who underwent sleep apnea surgery over nine years. The research team analyzed the surgery’s success rate based on different parameters, such as pre and postoperative polysomnography results, and compared it with the preoperative prediction of the AI program. The research team applied three AI models to the study (support vector machine, random forest, and gradient boosting) to account for various factors contributing to outcome prediction, like age, tonsil size, BMI, and sleep time. Among them, the gradient boosting model showed an accuracy of 70.8 percent, significantly higher than the existing models.(3)
Seoul National University Hospital (SNUH) team has developed an artificial intelligence-based system that can predict the success rate of obstructive sleep apnea (OSA) surgery. The Professor Kim Hyun-Jik of the Department of Otolaryngology developed the platform by analyzing 163 patients who underwent sleep apnea surgery over nine years. The research team analyzed the surgery’s success rate based on different parameters, such as pre and postoperative polysomnography results, and compared it with the preoperative prediction of the AI program.
The research team applied three AI models to the study (support vector machine, random forest, and gradient boosting) to account for various factors contributing to outcome prediction, like age, tonsil size, BMI, and sleep time. Among them, the gradient boosting model showed an accuracy of 70.8 percent, significantly higher than the existing models.(3)
AI hearing aid
Starkey Hearing Technologies, a Minnesota-based American private company, has unveiled Evolv AI, a ground-breaking technology created by years of refining research and AI algorithms to power very high-fidelity audio modeled for the human auditory system. Like the brain, Evolv AI Sound automatically suppresses background noise and increases speech audibility and intelligibility using machine learning technology. As the world leader in listening devices manufacturing, Starkey is proud to introduce the industry’s smallest 2.4 GHz Completely-in-the-canal (CIC) device as part of the Evolv AI line of hearing aids. Additional features of the Evolv AI product include a 40% reduction in energy consumption compared to previous technologies. Combining these uniques features offers users an unmatched hearing experience. Evolv AI delivers realistic and genuine sound quality in every environment, without extra adjustments or settings.(4)
Starkey Hearing Technologies, a Minnesota-based American private company, has unveiled Evolv AI, a ground-breaking technology created by years of refining research and AI algorithms to power very high-fidelity audio modeled for the human auditory system. Like the brain, Evolv AI Sound automatically suppresses background noise and increases speech audibility and intelligibility using machine learning technology. As the world leader in listening devices manufacturing, Starkey is proud to introduce the industry’s smallest 2.4 GHz Completely-in-the-canal (CIC) device as part of the Evolv AI line of hearing aids.
Additional features of the Evolv AI product include a 40% reduction in energy consumption compared to previous technologies. Combining these uniques features offers users an unmatched hearing experience. Evolv AI delivers realistic and genuine sound quality in every environment, without extra adjustments or settings.(4)
Starkey Hearing Technologies, a Minnesota-based American private company, has unveiled Evolv AI, a ground-breaking technology created by years of refining research and AI algorithms to power very high-fidelity audio modeled for the human auditory system. Like the brain, Evolv AI Sound automatically suppresses background noise and increases speech audibility and intelligibility using machine learning technology. As the world leader in listening devices manufacturing, Starkey is proud to introduce the industry’s smallest 2.4 GHz Completely-in-the-canal (CIC) device as part of the Evolv AI line of hearing aids. Additional features of the Evolv AI product include a 40% reduction in energy consumption compared to previous technologies. Combining these uniques features offers users an unmatched hearing experience. Evolv AI delivers realistic and genuine sound quality in every environment, without extra adjustments or settings.(4)
Artificial Intelligence Algorithm Can Rapidly Detect Severity of Common Blinding Eye
Specialists of the New York Eye and Ear (NYEE) Infirmary of Mount Sinai have created an artificial intelligence (AI) algorithm that can quickly and precisely determine age-related macular degeneration (AMD), the primary cause of vision loss in the United In AMD, the macula (focal space of the retina) deteriorates, producing foggy vision that can radically worsen over the long run. AI innovation may assist specialists in foreseeing AMD with a quick assessment showing the degree and progression of the disease, prompting patients to receive earlier clinical treatment to prevent further vision loss.(5)
Eye, robot: Artificial intelligence dramatically improves accuracy of classic eye exam
The classic eye test might be getting an update soon. Chris Piech, a computer scientist at Stanford University, and his associates have produced an AI-powered online vision test that delivers more precise conclusions than the 19th-century-devised Snellen chart. Clients initially align their screen size and distance by redimensioning a crate on a page to the size of a charge card and typing their distance from the screen. The test shows an “E” in one of four directions. Given the appropriate response, the calculation utilizes a program to forecast a dream score similar to recommendations in music platforms. The measure can make a more exact estimate of the score as the test goes on. The test poses 20 inquiries for each eye and takes a few minutes to finish.(6)
The classic eye test might be getting an update soon. Chris Piech, a computer scientist at Stanford University, and his associates have produced an AI-powered online vision test that delivers more precise conclusions than the 19th-century-devised Snellen chart. Clients initially align their screen size and distance by redimensioning a crate on a page to the size of a charge card and typing their distance from the screen. The test shows an “E” in one of four directions.
Given the appropriate response, the calculation utilizes a program to forecast a dream score similar to recommendations in music platforms. The measure can make a more exact estimate of the score as the test goes on. The test poses 20 inquiries for each eye and takes a few minutes to finish.(6)
Teaching AI algorithms to identify corneal pathology
The research team at the School of Medicine and Health Sciences in Monterrey, Mexico, conducted a study to train different AI algorithms to discriminate between healthy and diseased corneas by evaluating spectral-domain optical coherence tomography (SD-OCT) corneal images. Their experimental, comparative pilot study included a control group comprised of 71 images of healthy corneas and an experimental group comprised of 22 SD-OCT images of random corneal pathologies. Some include corneal ectasia, active and inactive herpes simplex keratitis, as well as exposure keratopathy. For the algorithm’s training phase, investigators input the definitive diagnosis after complete imaging processing. Analysis indicated that all four methods—random forest, convolutional neural network, transfer learning-support vector machines, and transfer learning-random forest—achieved high levels of sensitivity, specificity, and accuracy.(7)
The research team at the School of Medicine and Health Sciences in Monterrey, Mexico, conducted a study to train different AI algorithms to discriminate between healthy and diseased corneas by evaluating spectral-domain optical coherence tomography (SD-OCT) corneal images. Their experimental, comparative pilot study included a control group comprised of 71 images of healthy corneas and an experimental group comprised of 22 SD-OCT images of random corneal pathologies. Some include corneal ectasia, active and inactive herpes simplex keratitis, as well as exposure keratopathy.
For the algorithm’s training phase, investigators input the definitive diagnosis after complete imaging processing. Analysis indicated that all four methods—random forest, convolutional neural network, transfer learning-support vector machines, and transfer learning-random forest—achieved high levels of sensitivity, specificity, and accuracy.(7)
Commercial AI system enables detection of vision-threatening diabetic retinopathy
An artificial intelligence (AI) system that can diagnose diabetic retinopathy without physician assistance has outperformed expectations in a clinical trial, including vision-threatening advanced cases. Researchers team at the Lundquist Institute for Biomedical Innovation, part of Harbor–UCLA Medical Center, in collaboration with other US centers, conducted a clinical trial to evaluate the safety of the EyeArt Automated DR (diabetic retinopathy) Detection System. The objective was to detect the disease type using dilated and non-dilated eye imaging databases. The results are exceptional; with the undilated imaging protocol, the AI system can exceed predetermined endpoints for diagnostic, with sensitivity (greater than 90%) and specificity (greater than 82.5%) of DR detection.(8)
An artificial intelligence (AI) system that can diagnose diabetic retinopathy without physician assistance has outperformed expectations in a clinical trial, including vision-threatening advanced cases. Researchers team at the Lundquist Institute for Biomedical Innovation, part of Harbor–UCLA Medical Center, in collaboration with other US centers, conducted a clinical trial to evaluate the safety of the EyeArt Automated DR (diabetic retinopathy) Detection System.
The objective was to detect the disease type using dilated and non-dilated eye imaging databases. The results are exceptional; with the undilated imaging protocol, the AI system can exceed predetermined endpoints for diagnostic, with sensitivity (greater than 90%) and specificity (greater than 82.5%) of DR detection.(8)
References