Digital Health

GPT-4o now outperforms human radiologists at complex CT protocoling A new study suggests that artificial intelligence may soon handle the complex administrative tasks that currently bog down radiologists. Researchers tested GPT-4o on its ability to assign the correct scanning protocols for abdominal and pelvic CTs—a critical step that determines how the machine is set up for each patient. The AI achieved an impressive 96.2% accuracy rate significantly outperforming the human radiologists who scored 88.3%. The model was particularly effective when provided with specific clinical context demonstrating that large language models are now capable of understanding nuanced medical instructions. This finding points to a future where AI handles the workflow logistics allowing human doctors to focus entirely on image interpretation and diagnosis. Read the original article at: https://radiologybusiness.com/topics/artificial-intelligence/gpt-4o-outperforms-radiologists-ct-protocoling ...

 The ECRI Institute, a global authority on medical safety, has designated the misuse of AI chatbots as the number one health technology hazard for 2026. The report warns that the rapid deployment of patient-facing AI tools is creating a dangerous credibility trap. These chatbots often produce coherent nonsense with answers that sound authoritative and grammatically perfect but are factually incorrect.  Because the output looks professional, patients are less likely to question it, leading to potential medication errors or delayed care. The institute urges healthcare organisations to implement strict oversight and visible disclaimers before letting these tools interact directly with patients. Read the original article at: https://www.healthcareinfosecurity.com/healthcare-chatbots-provoke-unease-in-ai-governance-analysts-a-30483 Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

 While AI is excelling at administrative tasks it still struggles to beat the human eye in complex cancer diagnosis. A comparative study pitted three major language models—ChatGPT-4o, Claude 3 Opus, and DeepSeek-R1—against human radiologists in analyzing mammography reports for breast cancer risks (BI-RADS 4). The results were clear: human radiologists significantly outperformed all three AI models. While the AI tools demonstrated high sensitivity meaning they were good at flagging potential issues they suffered from low specificity leading to a high volume of false alarms. The study concludes that for now these tools are best used as "safety net" assistants to ensure nothing is missed rather than as independent diagnostic agents. Read the original article at: https://medinform.jmir.org/2025/1/e80182 Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

 In a surprising upset for proprietary tech giants an open source AI model has taken the lead in patient communication. A 2026 study evaluated how well various AI models could answer patient inquiries about atherosclerotic cardiovascular disease (ASCVD). The open source model DeepSeek R1 achieved a 96% "good response" rate outperforming both OpenAI's GPT-4o and Google's Gemini. The study highlights that specialized open source tools can be just as effective as expensive commercial models for patient education. However researchers noted a critical flaw across the board: all models, including DeepSeek, struggled to accurately provide specific treatment regimens reinforcing that while AI is an excellent educator it is not yet ready to be a prescriber. Read the original article at: https://medinform.jmir.org/2026/1/e81422 Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

A new study suggests that artificial intelligence may soon handle the complex administrative tasks that currently bog down radiologists. Researchers tested GPT-4o on its ability to assign the correct scanning protocols for abdominal and pelvic CTs—a critical step that determines how the machine is set up for each patient. The AI achieved an impressive 96.2% accuracy rate significantly outperforming the human radiologists who scored 88.3%. The model was particularly effective when provided with specific clinical context demonstrating that large language models are now capable of understanding nuanced medical instructions. This finding points to a future where AI handles the workflow logistics allowing human doctors to focus entirely on image interpretation and diagnosis. Read the original article at: https://radiologybusiness.com/topics/artificial-intelligence/gpt-4o-outperforms-radiologists-ct-protocoling Follow us on Instagram , Twitter , and Facebook to stay up to date with wh...

  A comprehensive new forecast identifies ten critical trends that will define the medical technology landscape this year. The report predicts a shifting environment where companies must navigate new financial strategies and complex regulations to succeed. Investment is bouncing back as AI and a stronger stock market encourage venture capital spending Buyers are using creative payment plans involving stocks and performance bonuses to reduce their financial risk Companies are choosing partnerships and licensing deals instead of full buyouts to share development costs Large corporations are spinning off smaller business units to focus resources on high growth areas like robotics Regulators are watching closely to prevent monopolies on patient data and artificial intelligence tools New US rules now block the transfer of sensitive health and genetic data to countries like China and Russia Rising tariffs an...

  Northern Ireland deploys AI that acts as a second pair of eyes, spotting missed fractures instantly Health services in Northern Ireland have rolled out a region wide artificial intelligence initiative designed to reduce diagnostic errors in emergency departments. The system integrates directly into existing radiology workflows acting as a digital safety net that reviews X-rays for signs of fractures that human eyes might miss during busy shifts. By flagging potential breaks in real time the tool ensures that patients receive immediate and accurate treatment preventing the complications that arise from missed injuries. This deployment represents a major step in using AI not just for complex diseases but for improving the speed and accuracy of everyday trauma care across an entire national health system. Read the original article at: https://www.digitalhealth.net/2026/01/northern-ireland-deploys-ai-tool-through-sectra-to-spot-fractures/ Deep learning now instantly grades...

 Researchers at MIT have developed a revolutionary non invasive tool to probe the mysteries of human consciousness. Using low intensity focused ultrasound the team can precisely target and stimulate deep brain structures such as the thalamus without the need for surgery. This technique allows scientists to effectively touch specific neural circuits to see how they influence awareness and arousal states. The goal is to map the physical source of consciousness and develop new treatments for patients with severe brain injuries or disorders of consciousness like comas. By manipulating these deep brain regions safely the study opens new frontiers in understanding how the brain generates the subjective experience of reality. Read the original article at: https://medicalxpress.com/news/2026-01-tool-consciousness.html Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

As artificial intelligence tools flood the mental health market a new ethical framework called Humane Intelligence has been proposed to protect vulnerable patients. The report argues that while AI can assist in monitoring symptoms or providing administrative support it must never cross the line into replacing the therapeutic human relationship. The framework establishes strict safety protocols to ensure that algorithmic decisions are transparent and that human clinicians remain the ultimate authority in care. It warns that relying on automated systems for psychological support without adequate guardrails risks eroding patient trust and potentially causing psychological harm. The initiative calls for a human in the loop approach where technology serves to strengthen rather than sever the bond between doctor and patient. Read the original article at: https://medicalxpress.com/news/2026-01-humane-intelligence-framework-safer-patient.html Follow us on Instagram , Twitter , and Facebook...

 A new study published in a leading scientific journal demonstrates how deep learning is modernizing dental radiography. Researchers developed an AI model that instantly evaluates the quality of dental X-rays at the moment they are taken. The system checks for issues like poor positioning or incorrect exposure grading the image quality in real time. This immediate feedback loop allows technicians to correct errors instantly or confirms that the scan is usable preventing the need for patients to be called back for retakes. By reducing the number of repeated scans the technology significantly lowers the cumulative radiation dose patients receive while ensuring dentists have high quality images for accurate diagnosis. Read the original article at: https://www.nature.com/articles/s41598-026-35100-9 Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

 Health services in Northern Ireland have rolled out a region wide artificial intelligence initiative designed to reduce diagnostic errors in emergency departments. The system integrates directly into existing radiology workflows acting as a digital safety net that reviews X-rays for signs of fractures that human eyes might miss during busy shifts. By flagging potential breaks in real time the tool ensures that patients receive immediate and accurate treatment preventing the complications that arise from missed injuries. This deployment represents a major step in using AI not just for complex diseases but for improving the speed and accuracy of everyday trauma care across an entire national health system.  Read the original article at: https://www.digitalhealth.net/2026/01/northern-ireland-deploys-ai-tool-through-sectra-to-spot-fractures/ Follow us on Instagram , Twitter , and Facebook to stay up to date with what's new in healthcare all around the world.

  Memories are not snapshots; they are dynamic files that scientists are now learning to rewrite. New insights from neuroscience challenge the long held belief that memories are fixed recordings of the past. Research reveals that recalling a memory is actually a reconstructive process where the brain reactivates specific cells and releases chemicals that can alter the original information. This biological mechanism means that every time we remember something we are potentially rewriting it based on our current emotions and environment. Scientists have successfully manipulated this process in mice to erase or modify specific memories opening the door to potential treatments for humans suffering from PTSD or dementia. While the concept of memory editing raises ethical questions the goal is strictly therapeutic. The aim is to use these mechanisms to soften traumatic memories or reinforce positive ones providing relief for mental health conditions that are currently difficult to trea...

 A powerful new artificial intelligence model named TRisk is setting a new standard for predicting survival rates in heart failure patients. By analyzing the electronic health records of over 400,000 individuals the model was able to forecast outcomes with significantly higher accuracy than existing risk scores. Unlike older methods that rely on a limited set of variables TRisk evaluates a broad range of data points to identify both known and previously overlooked risk factors such as specific liver diseases or cancers. The tool proved robust across different demographics including age and gender suggesting it is less biased than traditional models. This capability allows doctors to pinpoint high risk patients using data that is already available in their medical records. It offers a practical way to personalize treatment plans and intervene earlier for those with the poorest prognoses. Read the original article at: https://www.nature.com/articles/s41746-025-02296-5 Follow us on...

 Detecting the early signs of cognitive decline often requires time consuming tests but new research suggests that artificial intelligence can spot these signals directly in routine clinical documentation. Researchers developed a system using multiple automated AI agents to analyze medical notes for subtle language patterns indicative of cognitive issues. When pitted against systems guided by human experts the automated agents performed with remarkable accuracy even surpassing human led models in fine tuning results. The study found that the AI could sift through vast amounts of unstructured text to identify patients at risk who might otherwise be overlooked during brief office visits. While the model faced some challenges when applied to diverse real world populations it demonstrates the potential for AI to act as a scalable screening tool that expands access to early diagnosis without adding to the administrative burden of clinicians. Read the original article at: https://www....

 Medication non adherence is a massive healthcare challenge that costs billions and endangers lives but engineers have developed a high tech solution in the form of a smart pill. This new ingestible device contains a biodegradable radio antenna made from zinc and cellulose which are safe for the body. Once swallowed the pill dissolves in the stomach and transmits a wireless signal to a nearby receiver confirming that the patient has taken their medicine. This system effectively tracks adherence in real time without requiring the patient to log anything manually. It is particularly promising for patients with critical conditions like tuberculosis or organ transplants where skipping a dose can have fatal consequences. The technology offers a reliable way for doctors to monitor treatment plans remotely ensuring that life saving therapies are actually being delivered as prescribed. Read the original article at: https://news.mit.edu/2026/pills-communicate-from-stomach-could-improve-m...

 New insights from neuroscience challenge the long held belief that memories are fixed recordings of the past. Research reveals that recalling a memory is actually a reconstructive process where the brain reactivates specific cells and releases chemicals that can alter the original information. This biological mechanism means that every time we remember something we are potentially rewriting it based on our current emotions and environment. Scientists have successfully manipulated this process in mice to erase or modify specific memories opening the door to potential treatments for humans suffering from PTSD or dementia. While the concept of memory editing raises ethical questions the goal is strictly therapeutic. The aim is to use these mechanisms to soften traumatic memories or reinforce positive ones providing relief for mental health conditions that are currently difficult to treat with standard medication or talk therapy alone. Read the original article at: https://medicalx...

Doctors defeated by AI. A Chinese AI outperformed human experts in diagnosing complex medical cases. In a dramatic public showdown in China, a locally developed AI model has reportedly outperformed a team of senior human physicians in diagnosing complex medical conditions. The competition, designed to test the limits of the "MedGPT" Large Language Model, pitted the AI against doctors from top-tier hospitals. Both sides were tasked with analyzing real-world patient cases, diagnosing the illness, and recommending treatment plans. The results were striking: the AI not only diagnosed cases faster but also achieved a higher accuracy rate as judged by a panel of independent experts. While human doctors still hold the edge in empathy and physical examination, this event marks a significant milestone in the capabilities of medical AI. It demonstrates that for data-heavy diagnostic tasks—where pattern recognition in symptoms and history is key—algorithms are rapidly closing the ga...

 A new study serves as a reality check for the booming mHealth sector. Researchers analyzed adherence rates among older adults undergoing mobile-based cardiac rehabilitation, where patients were asked to wear accelerometers to track their recovery. The findings revealed a "digital drop-off": while the technology worked perfectly, human behavior did not. Adherence to wearing the devices plummeted over time, with a significant portion of patients failing to use the monitoring tools consistently enough to generate useful data. The study identifies the "weak link" in digital health: the patient's willingness to engage. Factors such as comfort, technical literacy, and perceived value of the data played huge roles in whether a patient stuck with the program. The authors argue that simply giving a patient a high-tech device is not a solution in itself. Future mHealth interventions must prioritize "human-centric design", making devices invisible, automatic, ...

 As the line between consumer smartwatches and medical devices blurs, Healthcare IT Today explores the rigorous journey wearables must undergo to earn the trust of the medical community. The article highlights that for a wearable to transition from a "fitness gadget" to a "clinical tool," it must survive a battery of validation tests that go far beyond step counting. These include verifying sensor accuracy across diverse skin tones, ensuring consistent data transmission under movement, and proving that the device's battery life can support continuous medical monitoring without data gaps. The piece emphasizes that the "future" of wearables lies in this validation phase. Hospitals are eager to adopt remote monitoring to reduce readmissions, but they cannot risk liability on unproven tech. Manufacturers are now partnering with clinical research organizations earlier in the development cycle, subjecting their devices to the same scrutiny as traditional ...

A major new study addresses the "Wild West" of medical AI by introducing a rigorous new benchmark for Large Language Models (LLMs). Dubbed CSEDB (Clinical Safety-Effectiveness Dual-Track Benchmark), this framework was developed by a coalition of 32 specialists across 26 clinical departments. Unlike previous tests that only measured how well an AI could answer medical exam questions, CSEDB evaluates two critical real-world factors: safety (does the AI recommend dangerous treatments?) and effectiveness (does the advice follow standard clinical guidelines?). Testing prominent models like GPT-4 and localized medical LLMs, the researchers found a concerning gap. While many models are "knowledgeable" and can pass exams, they often fail on safety protocols, occasionally hallucinating non-existent treatments or missing critical contraindications. This new benchmark serves as a necessary "stress test" for the industry, providing a standardized way to ensure that ...