Abstract Artificial intelligence for medical imaging is required to be accurate and interpretable to clinicians. However, current multimodal biomedical foundation models often prioritize performance over explainability. Here we present ConceptCLIP, an explainable biomedical foundation model that achieves state-of-the-art diagnostic accuracy while delivering human-interpretable explanations across diverse imaging modalities. We curate MedConcept-23M, a large-scale dataset comprising 23 million biomedical image–text–concept triplets. Leveraging this dataset, we pretrain ConceptCLIP via joint image–text and region–concept alignment for precise and interpretable medical image analysis. Across a large-scale benchmark covering 78 datasets in 10 imaging modalities, ConceptCLIP demonstrates superior diagnostic performance while providing human-understandable explanations. In a clinician user study spanning three modalities, the concept-based explanations provided by ConceptCLIP help clinicians verify model predictions and identify potential errors. As an explainable biomedical foundation model, ConceptCLIP represents a critical milestone towards the widespread clinical adoption of AI, thereby advancing trustworthy AI in medicine. This is a preview of subscription content, access via your institution Access options Access Nature and 54 other Nature Portfolio journals Get Nature+, our best-value online-access subscription $32.99 / 30 days cancel any time Subscribe to this journal Receive 12 digital issues and online access to articles $119.00 per year only $9.92 per issue Buy this article - Purchase on SpringerLink - Instant access to the full article PDF. USD 39.95 Prices may be subject to local taxes which are calculated during checkout Data availability This study incorporates a total of 79 datasets. Among these, the proposed MedConcept-23M dataset is used for training the ConceptCLIP model. The image part of the pretraining dataset, MedConcept-23M, is open-sourced and can be obtained directly from the publicly available PMC Open Access Subset (PMC-OA). The captions and concepts are available at https://huggingface.co/datasets/JerrryNie/MedConcept-23M. To enable reconstruction of the full dataset, we provide a dataset reconstruction script as part of
Aug 17, 2026 · via nature.com
Supermarket duopoly tests facial recognition
Coles and Woolworths have conducted tests of facial recognition technology that they could use in their Australian supermarkets to combat crime, although it is not clear whether they will implement it fully.
Dr Jason Pallant, a marketing expert at RMIT University, said the trials posed a “security versus privacy dilemma” (The Guardian).
Tom Sulston, the head of policy at Digital Rights Watch, flagged that facial recognition was “wildly inaccurate”, and also raised concerns that Coles and Woolworths wouldn’t be able to control the way their biometric data was used (The Guardian).
Both Coles and Woolworths say they have not decided whether to deploy the technology after the trials. “Keeping our team and customers safe is the most important thing we do, and we’ve put significant investment towards this,” a Woolworths spokesman said (AFR).
The testing revelations follow hardware chain Bunnings’ legal battle win earlier this year that allowed it to monitor its customers with artificial intelligence (ABC).
Aug 17, 2026 · via thesaturdaypaper.com.au
“The Federal Circuit concluded the reference ‘logically would have commended itself to an inventor’s attention in considering his problems.’” The U.S. Court of Appeals for the Federal Circuit (CAFC) issued a precedential decision on Friday, August 14, in The Nielsen Company (US), LLC v. TVision Insights, Inc., affirming a Patent Trial and Appeal Board (PTAB) final written decision that invalidated challenged claims of a Nielsen audience measurement patent as obvious. The court rejected arguments that the Board improperly relied on a scientific publication as analogous prior art and that the publication failed to disclose the claimed resolution reducing and facial recognition steps. TVision Insights, Inc. filed a petition for inter partes review (IPR) of U.S. Patent No. 11,470,243, owned by The Nielsen Company (US), LLC, titled “Methods and Apparatus to Capture Images.” The patent relates to systems for measuring and identifying the audience of a media presentation device such as a television. The specification describes a camera-based system that uses a “people counter” to detect audience members based on features such as heads and faces in low-resolution images. A “person identifier” then compares the detected faces against stored facial signatures using higher-resolution images. TVision’s petition challenged 14 claims, but Nielsen disclaimed the 3 independent claims at issue in its preliminary response, and the Board instituted review of 11 dependent claims, with claims 4 through 6 becoming the focus of the appeal. Claim 4 recites processor circuitry that reduces the resolution of a first image to obtain a reduced resolution image and determines head orientation based on that image. Claims 5 and 6 add a two-step process that generates a facial signature from a separate image corresponding to the head location identified in the reduced resolution image, then compares that signature against a database of stored signatures. The Board considered two
Aug 17, 2026 · via ipwatchdog.com
Offering a variety of advertising and sponsorship options for reaching influential specialists from targeted demographic splits.
Cureus provides an equitable, efficient publishing and peer reviewing experience without sacrificing publication times.
Generate broad awareness and deliver relevant, peer-reviewed clinical experiences directly to potential customers.
Dedicated Cranial Radiosurgery: Clinical Experience with New & Innovative SRS Technologies
Sponsored by Zap Surgical Systems
Real-Time Adaptive Motion Management on Helical and Robotic RT Platforms
Sponsored by Accuray, Inc.
Cureus Journal of Medical Science
Sponsored by Zap Surgical Systems
Sponsored by Accuray, Inc.
You can unsubscribe anytime.
By joining Cureus, you agree to our Privacy Policy and Terms of Use.
Aug 17, 2026 · via cureus.com
Researchers at the Institute of Fundamental and Frontier Sciences, University of Electronic Sciences and Technology of China have combined boson sampling, a quantum process with experimentally verified advantage over classical computers, with neural networks to improve machine learning classification. The team developed a hybrid framework where a neural network compresses data features onto a boson sampling circuit, generating quantum states that enhance support vector machine performance. Using four datasets with various classes, the model outperformed classical linear and sigmoid kernels, demonstrating the potential of boson sampling-based quantum kernels for practical quantum-enhanced machine learning. Hybrid Boson Sampling-Neural Network Architecture for Enhanced Classification The core innovation lies in a neural network’s ability to compress complex data features, preparing them for processing by a programmable boson sampling circuit. This approach addresses a significant hurdle in quantum machine learning: the high dimensionality of practical datasets. The team’s framework utilizes the neural network to reduce the number of features needed for analysis, bridging the gap between large, complex data and the limitations of current quantum hardware. The resulting quantum states, generated by the boson sampling circuit, span a high-dimensional space, enabling improved classification performance. The researchers tested their model against four distinct datasets, Ionosphere, Spambase, MNIST, and Fashion-MNIST, each containing various classes of data, and the hybrid model outperformed classical linear and sigmoid kernels in these tests. The researchers found that achieving enhanced accuracy depended on utilizing a sufficiently expressive boson sampling circuit, with expressivity controlled by both the number of modes and injected photons. This suggests a pathway to optimize the quantum component for specific classification tasks. Mohammad Sharifian explained in their published work that “the integrated architecture of classical neural network with quantum boson sampler enhances the accuracy of SVM image classification outperforming both classical linear and non-linear sigmoid kernels as well as
Aug 17, 2026 · via quantumzeitgeist.com
Meta files patent for AI facial recognition glasses that identify people around you and record automatically A newly published patent from Meta suggests smart glasses could recognize faces, detect laughter and more Imagine hosting a dinner party without stopping to pull out your phone. Later, your glasses present you with a collection of photos and video clips showing your partner across the table and your friends laughing together. That's the idea behind the system Meta has patented for it's Meta glasses. The patent filing describes camera-equipped glasses that use facial recognition, expression analysis and information about your relationships to identify the people around you and decide which moments you may want to keep. The glasses could even ask, “I’ve generated some highlights of tonight’s dinner party. Would you like to see them?” At a time when Meta is continuously under fire for it's "pervert glasses" the patent sounds convenient for some, but also a potential privacy nightmare. How Meta’s glasses could choose what to capture The patent, published Thursday and first reported by 404 Media, describes a system that looks for “points of interest” within the wearer’s field of view. At a dinner party, facial recognition could identify the wearer’s wife and place her in the center of the frame. Other guests could be cropped out or blurred in the background. The system could then analyze facial expressions. If two people start laughing, for example, the glasses could recognize that reaction and reposition the camera to focus on them. Who makes the final cut may depend on the wearer’s relationship with each person. According to the report, information from a user’s social graph could help the system determine which friends or family members are likely to be more interesting to them than other people in the room. Get instant access
Aug 17, 2026 · via tomsguide.com
260814-N-CF730-1023 U.S. Navy Capt. Brian Bungay, commanding officer, Naval Base San Diego (NBSD), left, presents awardees with a challenge coin during the Presidential Physical Fitness Award presentation at Murphy Canyon Youth Center, Aug. 14, 2026. The Presidential Physical Fitness Award is a national recognition program designed to encourage physical fitness, strength, and healthy habits among youth. NBSD recognize excellence in physical fitness as one of the select bases across enterprise to pilot the revived program. Established in 1922, NBSD is the largest West Coast naval installation and principal homeport of the Pacific Fleet, supporting more than 60 combatant and auxiliary surface ships and more than 250 shore commands. (U.S. Navy photo by Interior Communications Electrician 2nd Class Ulrika Mendiola) | Date Taken: | 08.14.2026 | | Date Posted: | 08.17.2026 14:41 | | Photo ID: | 9872859 | | VIRIN: | 260814-N-CF730-1023 | | Resolution: | 4519x3008 | | Size: | 2.16 MB | | Location: | SAN DIEGO, CALIFORNIA, US | | Web Views: | 1 | | Downloads: | 0 | This work, Naval Base San Diego presents Presidential Physical Fitness Awards [Image 4 of 4], by PO2 Ulrika Mendiola, identified by DVIDS, must comply with the restrictions shown on https://www.dvidshub.net/about/copyright.
Aug 17, 2026 · via dvidshub.net
This is an uncorrected proof. Figures Abstract Predicting the neural response to natural images in the visual cortex requires extracting relevant features from the images and relating those feature to the observed responses. In this work, we optimize the feature extraction in order to maximize the information shared between the image features and the neural response across voxels in a given region of interest (ROI) extracted from the BOLD signal measured by functional magnetic resonance imaging (fMRI). We adapt contrastive learning (CL) to fine-tune a convolutional neural network, which was pretrained for image classification, such that a mapping of a given image’s features are more similar to the corresponding fMRI response than to the responses to other images. We exploit the Natural Scenes Dataset as organized for the Algonauts Project, which contains the high-resolution fMRI responses of eight subjects to tens of thousands of naturalistic images. We show that CL fine-tuning creates feature extraction models that enable higher encoding accuracy in both early and higher visual ROIs as compared to the features from the pretrained network. Quantitatively, performance is similar to baseline approach that directly uses a regression loss at the output of the network to tune it for fMRI response encoding. We investigate inter-subject transfer of the CL fine-tuned models, including subjects from the Natural Object Dataset, another lower-resolution dataset with 9 subjects. We also pool subjects for fine-tuning, which further improves encoding performance in early ROIs. Finally, we examine the performance of the fine-tuned models on common image classification tasks, explore the landscape of ROI-specific models by applying dimensionality reduction on the Bhattacharya dissimilarity matrix created using the predictions on those tasks, show that these landscapes match those based on representational similarity analysis. Finally, we generate images via Stable Diffusion based on vector-space prompts created by aligning the
Aug 17, 2026 · via journals.plos.org
Machine Vision Insights from Industry Experts on Three Decades of Progress Key Highlights - Machine vision has evolved from proprietary, expensive components to standardized, affordable solutions driven by advances in CMOS sensors and imaging protocols. - Key technological milestones include the rise of CMOS cameras, LED lighting, SWIR imaging, and the integration of AI, transforming automated inspection and industrial applications. - Experts predict continued hardware commoditization, smarter systems with on-the-job learning, and broader adoption of SWIR imaging fueled by AI and emerging sensor technologies. Vision Systems Design is celebrating a major milestone in 2026: 30 years of covering the machine vision and imaging industry. VSD began as a spin-off of sister publication Laser Focus World. While machine vision may seem like a specialty market to some, it is in fact a robust, growing, multi-billion-dollar industry that clearly merits a publication dedicated exclusively to its coverage. Still, 30 years is a long time, and too often the present can seem a little slow—even static—until it becomes the past. A quick AI search will instantly generate a reasonably accurate timeline of machine vision’s evolution. In brief, research into image processing and pattern recognition—the foundations of applications such as automated inspection—began in the 1950s and 1960s. Industrial adoption accelerated in the 1980s and 1990s, while neural networks and deep learning expanded machine vision’s capabilities by 2010. Today, AI-powered platforms continue to push applications to new levels. Such an exercise is helpful, if for no other reason than it illustrates that the sheer power, not to mention convenience, of technology has advanced, almost unimaginably so, from just a few years ago. But a much more interesting exercise is to see the past, present, and future through the knowledgeable eyes, deep experience, and informed viewpoint of people who have seen the timeline unfold—first person and
Aug 17, 2026 · via vision-systems.com
It's not too long ago that buying something meant handing over cash, or at least a physical card.
But one company is looking to take digital payment a step further, introducing an option that allows you to make a purchase not with your device — but your face.
Paying "with a smile" is a fair way off being commonplace in Australia, and it would be an opt-in service. However, facial recognition becoming more commonplace across in our daily lives. So what's around the corner?
Guests:
- Lauren Perry, responsible technology policy specialist at the University of Technology Sydney
- Kirsten Drysdale, independent journalist and host of The Internet Reviewed
Credits
Image Details
Technology, Society, Data Privacy
Aug 17, 2026 · via abc.net.au
Supermarket giants trial facial recognition technology Updated . First published at Coles and Woolworths are exploring the use of facial recognition technology to use in its stores in a bid to crack down on retail crime. The potential use of the technology is the latest step the supermarket giants are taking to protect its staff and prevent theft in its stores. Coles told nine.com.au it had not conducted an outright trial of the technology, instead saying it had completed a “small, one-off, controlled proof-of-concept test of the technology.” “[It] did not use customer or team member information or data,” a Coles spokesperson said. “No decision has been made about the deployment of facial recognition technology. “We will always explore technology that will help us keep our team and customers safe.” Nine.com.au understands the test used volunteers, and did not use the data of any customers or staff. Despite previous reports, Woolworths has denied conducting a trial of the same technology across its New Zealand stores. It is understood the company conducted a small trial of facial recognition technology in New Zealand, but not at a store with customers or staff. Nine.com.au understands the company has no formal plans to roll out the technology in Australia. Facial recognition technology has proved controversial, with its use in Australian stores only being approved earlier this year after Westfarmers, owner of hardware chain Bunnings, won a legal battle that reversed an initial ruling that blocked its rollout. The company had used the technology between 2018 and 2021, and insisted the images of a “vast majority of people [were] processed and deleted in 0.00417 seconds”. Bunnings managing director Mike Schneider insisted the technology would only be used to protect its staff and customers. “Our intent in trialling this technology was to help protect people from
Aug 17, 2026 · via nine.com.au
Amazon Rekognition faces scrutiny in false arrest lawsuit Amazon Web Services (AWS) was added as a defendant in a federal lawsuit brought by a St. Louis, Missouri man who spent 17 months in jail after police used facial recognition on a poor quality image of a masked suspect and then built a case around the photo. Christopher Gatlin alleges in an amended complaint filed last Thursday that his arrest and prosecution resulted from poor police work, inadequate training, misconduct and a defective facial recognition system developed by Amazon. The new filing names AWS and identifies Amazon Rekognition as the technology involved in the identification that ultimately led police to Gatlin. The addition of Amazon significantly broadens this civil rights case that until now has focused largely on how St. Louis and county police used facial recognition and what they did after it identified Gatlin as a possible match. The case began with the December 7, 2020, assault of a MetroLink security guard at the St. Charles Rock Road station in north St. Louis County. The victim suffered a traumatic brain injury and repeatedly told police he could not remember the men who attacked him, according to allegations recounted in a federal court ruling. Detectives obtained surveillance video showing the attackers. One investigator eventually submitted an image of one of them to the St. Louis Mugshot Recognition Technology system. The federal court record describes the image as grainy, blurry, taken from a distance and from above the suspect’s face. A hood covered part of his forehead and a medical mask obscured another portion of his face. The system nevertheless returned possible candidates, including Gatlin. Before the facial recognition search, investigators had no information pointing to him as a suspect. The Washington Post subsequently found that the regional St. Louis facial recognition
Aug 17, 2026 · via biometricupdate.com
Meta smart glasses patent reignites facial recognition debate Meta’s latest smart glasses patent offers a glimpse of a future in which cameras do more than record what their wearer sees. They could identify who is there, interpret what is happening and determine which people or moments matter most to the wearer. The patent describes meaningful privacy and data-security controls, but it also raises questions Meta may need to answer if it wants users—and the people on the other side of the glasses—to trust the technology. Published August 13, US 2026/0238876 A1, “Smart Cameras Enabled by Assistant Systems,” describes an AI assistant combining cameras with capabilities including facial recognition, expression analysis, gaze and object recognition. The application is a continuation of a 2022 filing, which itself continues an application first filed in 2019. When the camera decides what matters The patent’s dinner-party example makes the concept tangible. A user wearing AR glasses is surrounded by several people. The assistant could use facial recognition to determine that one person is the user’s wife and assign her greater “interestingness.” The camera could then center on her while other people are excluded from the frame or blurred. Alternatively, facial-expression recognition could determine that two people laughing are more interesting subjects and focus the capture on them. The camera, in other words, is no longer simply capturing a scene. It is interpreting it. Privacy is built into the patent Importantly, Meta does not ignore privacy. Facial recognition and facial-expression recognition can be subject to privacy settings. The identity-resolution system can also respect restrictions preventing another person’s identity from being searchable. The patent describes controls allowing biometric information to be limited to specific purposes while restricting its use by other applications or sharing with third parties. Some assistant functions, including speech processing, reasoning and memory, can
Aug 17, 2026 · via biometricupdate.com
USSOCOM advances smartphone iris biometrics toward field testing U.S. Special Operations Command (USSOCOM) is preparing competitive testing of technology that could turn commercial Android phones and tablets into contactless iris biometric collection devices compatible with the Defense Department’s Automated Biometric Identification System (DoD ABIS). A series of events to complete a study of mobile devices will be held September 30, although SOFWERX’s current events calendar lists September 29. Requests to participate are due August 28 and are limited to U.S. citizens. The effort, being run by USSOCOM’s Program Executive Office for Tactical Information Systems (PEO-TIS) with SOFWERX, seeks to determine whether cameras and lighting already built into mobile devices can capture iris images of sufficient quality for biometric matching without a separate optical attachment. A previous study found mobile cameras could locate irises, but USSOCOM wants testing across 500 to 1,000 people to establish whether the approach is viable. Testing is expected to examine matching scores, differences among individual cameras, image distortion and sensitivity to near-infrared wavelengths between 805 and 1,000 nanometers. The objective is to eliminate dedicated iris scanners by using cameras and lighting already built into commercial mobile devices, reducing the size, weight and power burden carried by special operations personnel while maintaining interoperability with DoD ABIS. Companies, universities and national laboratories whose technologies perform well could advance to negotiations for prototype or other agreements, including Other Transaction Authority arrangements under 10 U.S.C. 4022. Successful prototype projects can, when statutory requirements are met, lead to follow-on production without another competition. Article Topics Android | biometrics | data collection | iris biometrics | SOFWERX | U.S. Army | U.S. Government | USSOCOM Comments
Aug 16, 2026 · via biometricupdate.com
Coles and Woolworths have conducted early testing of facial recognition technology as the country’s two supermarket giants consider ways to stem the growth in crime and violent incidents in their stores.
The trials are part of broader moves across the retail industry to assess how to use the technology to protect staff, particularly in Victoria where major chains have reported rising rates of theft and other crimes.
Loading...
Aug 16, 2026 · via afr.com
JavaScript is disabled in your browser.
Please enable JavaScript to proceed.
A required part of this site couldn’t load. This may be due to a browser extension, network issues, or browser settings. Please check your connection, disable any ad blockers, or try using a different browser.
Aug 16, 2026 · via mysanantonio.com
Explore the things
you love
.
Log into Facebook
Email or mobile number
Password
Log in
Forgot password?
Create new account
English (US)
Español
Français (France)
中文(简体)
العربية
Português (Brasil)
Italiano
More languages…
Sign Up
Log In
Messenger
Facebook Lite
Video
Meta Pay
Meta Store
Meta Quest
Ray-Ban Meta
Meta AI
Instagram
Threads
Privacy Policy
Privacy Center
About
Create ad
Create Page
Developers
Careers
Cookies
Ad choices
Terms
Help
Contact Uploading & Non-Users
Meta © 2026
Aug 16, 2026 · via facebook.com
Most first year emergency medicine residents have at least basic facility with point of care ultrasound (POCUS). Many have skills far beyond what would have been considered adequate for a graduating EM resident 20 years ago. We ultrasound everything from optic nerves to metatarsal joints, but does it actually improve patient care? Let's start with the good news: what we know POCUS does well. Ultrasound often cuts out the middleman. Instead of waiting for a lab test or a read by radiology, we get answers in seconds at the bedside. In patients with concern for pregnancy complications, POCUS shortened the time to diagnosis of intrauterine pregnancy by 72 minutes when compared with radiology-performed ultrasound. A diaphoretic patient with chest pain can be a lot of things but a quick ultrasound that identifies a dilated aortic root can help focus care quickly, limiting time chasing other diagnoses. Ultrasound improves procedural success. POCUS efficacy for placement of central lines is well known, but it also aids peripheral intravenous (IV) placement, improving overall success, first attempt success, and decreased overall time. A 2025 meta-analysis of randomized controlled trials in pediatric emergency departments (EDs) demonstrated that POCUS improved both first attempt and overall procedural success, including IV access and lumbar puncture. Unfortunately, there has been less demonstration of morbidity or mortality benefit. Notably, the SHOC-ED trial (covered in Speed of Sound in January 2019) found that adding POCUS to the evaluation of undifferentiated shock did not add much in the terms of survival, length of stay, or several other metrics. Designing studies that meet stringent criteria to establish a mortality benefit can be difficult. A critical review from 2025 found most POCUS studies focus on accuracy and less on outcome, preventing a full assessment of its impact. It also highlighted other issues impeding the
Aug 16, 2026 · via ovid.com
A monthslong fight over an ID-scanning system that photographs arriving patrons at Castro gay bars has come to a head, with all three venues known to use it — Badlands, Toad Hall, and most recently The Mix — halting the practice under pressure from privacy advocates. But the retreat from the ID-check tool, called PatronScan, hasn’t settled the question at the heart of the fight: whether the added security is worth the privacy trade-off. On one side are patrons and digital-rights groups warning that ID-scanning kiosks amount to surveillance in LGBTQ+ spaces at a moment when many in the community feel exposed at a moment of heightened federal scrutiny of queer and trans people. On the other side are bar owners and staff citing documented incidents — including a brutal assault (opens in new tab) on a Mix employee — as the reason they wanted the extra layer of security. Evan Greer, head of Fight for the Future, which spearheaded a boycott campaign against the bars, celebrated the news as “a huge step in the right direction.” The Mix hasn’t stopped using PatronScan altogether — it paused only the camera function of the company’s Guard+ kiosk, which photographs each patron. The bar still scans government IDs to verify age and authenticity, the way it did before. Mix manager Nick Kealy said he didn’t even want to suspend the photo-taking portion of PatronScan, and neither did his staff who credit the tool with making the bar safer after a brutal attack on a former staffer. The assault happened on March 9, 2025. A doorman ousted three unruly patrons from The Mix earlier in the day, according to the Bay Area Reporter, including one who bit him on the finger. That evening, the bouncer was hunted down and beaten a couple of
Aug 16, 2026 · via sfstandard.com
Meta Files Patent for Facial Recognition Camera System
Meta's patent for a facial recognition camera system.
The United States Patent and Trademark Office has published a patent application from Meta Platforms Technologies for a system that recognizes faces in a frame, identifies their actions, and automatically edits tagged video clips. The document uses a dinner party as an example: a person wearing smart glasses observes guests, and an AI assistant then offers a selection of the evening’s best moments.
Found a mistake in the text? Select it and press CTRL+ENTER
Aug 16, 2026 · via forklog.com