Wie sind die Öffnungszeiten?

Montag 09:00 - 17:00 Dienstag 09:00 - 17:00 Mittwoch 09:00 - 17:00 Donnerstag 09:00 - 17:00 Freitag 09:00 - 17:00

Digi-Bridges, Zürich (2026)

11/10/2026

Could the future of drug development involve fewer animal experiments and more research based on human biology?

An interesting article in IEEE Spectrum explores how technologies such as organs on chips, human tissue models, and AI simulations are creating new possibilities for drug development and safety testing.

Some of the figures are encouraging:

Around 92% of drugs entering clinical trials in the US never reach the market. This does not necessarily mean that animal testing is responsible, but it does raise questions about how well our current research methods predict what will happen in humans.

In one study highlighted in the article, computer models of human heart cells identified drugs associated with dangerous heart rhythm problems with 89% accuracy, compared with 75% for animal studies.

These results are promising, particularly when we consider the ethical concerns surrounding animal testing. But we should also be careful not to assume that newer technologies are automatically better.

Human biology is incredibly complex, and no single model can fully reproduce it. Many of these alternatives still need stronger evidence, consistent standards, and independent validation before they can be widely trusted. They may also complement rather than completely replace animal studies, at least for now.

What I found particularly interesting is that the challenge is not only technological, but also cultural and institutional.

Even when better alternatives become available, changing established research practices takes time. Researchers need training, regulators need reliable evidence, and funding bodies and scientific journals need to become more open to different approaches.

The encouraging news is that regulation is beginning to move in this direction, including recent developments in the US and Europe.

If newer methods can provide more reliable evidence while reducing animal suffering, there is a strong case for adopting them. But that case needs to be built on rigorous comparisons, not technological enthusiasm alone.

And perhaps this is where some of the hardest work lies: creating a research environment where established methods can be questioned, emerging approaches are evaluated fairly, and scientific decisions are guided by evidence rather than familiarity.

An interesting development to follow, with cautious optimism.

Article: https://spectrum.ieee.org/alternatives-to-animal-testing

05/10/2026

AI in healthcare is often discussed in extremes.

In some conversations, it is presented as a replacement for clinicians. In others, it is presented as a simple solution to access, cost, and quality challenges. The reality is more complex.

A recent McKinsey analysis argues that AI can already perform parts of what it calls "discursive care". In simple terms, this means the nonphysical parts of care, such as intake, triage, diagnosis support, explaining results, referrals, and follow up.

Some of the figures are striking. The analysis estimates that AI could already perform 16-22 % of US outpatient care, equal to about 13-19 % of outpatient healthcare spending. It also estimates that AI can perform a substantial portion of core tasks for 11 million out of 19 million healthcare workers.

This makes the role of clinicians more important, especially in the areas where trust, judgment, empathy, context, and accountability matter most. Listening to patients, making sense of uncertainty, discussing trade offs, and providing hands on care are central parts of healthcare.

It is also important to look at why many patients are already turning to AI. The article points to major access gaps: 92.3 million people living in primary care shortage areas, an average 31 day wait for a physician appointment, and 36% of adults delaying or skipping care because of cost. For many people, AI is becoming a workaround for a system that is often hard to access.

This is where careful implementation becomes essential. AI may improve access, support clinicians, and make some parts of care more consistent. At the same time, rushed implementation or a narrow focus on cost savings could create new risks. These include unsafe advice, bias, privacy concerns, unclear accountability, overreliance, deskilling, and new forms of low value care.

The focus therefore needs to include both capability and conditions for use:
Where can AI safely add value?
Who remains accountable?
How are patients protected?
How are healthcare professionals supported rather than bypassed?

For me, the future of AI in healthcare should be framed around strengthening care, especially where the current system is failing patients and overburdening professionals.

https://www.mckinsey.com/industries/healthcare/our-insights/the-ai-powered-future-of-care #/

20/09/2026

A thoughtful example of what healthcare AI looks like once it leaves the lab.

In rural India, AI is being used to screen for tuberculosis through cough analysis, monitor newborn growth with a smartphone, and support X ray interpretation where specialist radiologists are scarce.

India has around 1.02 million ASHA community health workers (frontline workers who connect rural communities with the public health system). Each typically serves about 1 000 people and already manages a wide range of responsibilities. If an AI tool adds another device, another workflow, or another administrative task, it can easily make an overstretched system more complicated rather than less.

Some of the examples in this article show a more practical approach. The newborn monitoring tool works on low end Android phones, can function offline, and has been integrated into an existing government platform. In Goa, AI supported X ray screening is linked to patient navigation and follow up rather than stopping at an automated alert.

That difference matters.

X rays that previously took 7 to 10 days to be reviewed could be flagged within minutes. But the bigger improvement came from reorganising what happened afterwards. With patient navigation and clearer follow up pathways, a process that had taken three to four weeks was reportedly reduced to around three to four days.

This also highlights a limitation that is easy to overlook. Finding more possible cases does not help much if the health system has no capacity to deal with them. AI can remove one delay while creating pressure somewhere else.

Trust is another part of the picture. A patient who feels healthy may not act on an AI generated warning without someone they trust explaining why further testing matters. That keeps community health workers and clinicians central, even as the technology becomes more capable.

For me, the most useful questions are practical ones: Does the technology reduce work or add to it? Can people act on what it finds? Does it fit into the existing care pathway? And does it make the overall process better for patients and healthcare workers?

Technology can cerainly help, but the surrounding system determines how much difference it actually makes.

https://www.channelnewsasia.com/asia/coughs-x-rays-how-ai-changing-indias-healthcare-frontline-6356946

17/09/2026

Foundation models are moving into physiological data, and Research’s new GlucoFM is an interesting example of where this could lead.

GlucoFM is a foundation model for continuous glucose monitoring (CGM) data. What caught my attention is not simply that it predicts metabolic outcomes, but how it approaches the signal itself: separating slower glucose trends from short-term deviations rather than treating a CGM trace as one continuous stream.

The early results are impressive. Across 14 cohort–task evaluations, GlucoFM improved average PR-AUC from 54.7 to 58.8 compared with the strongest CGM-specific baseline trained on the same data. It also performed well when transferred across datasets and when only small amounts of labelled data were available. For post-meal glucose prediction, it achieved a mean absolute error of 21.88 mg/dL compared with 22.90 mg/dL for the best baseline.

That matters because labelled clinical data are difficult and expensive to obtain. If models can learn useful physiological representations from large amounts of unlabelled sensor data, they could potentially make much better use of the growing volume of data generated by wearables and medical devices.

But this is also where I think we need to keep the excitement in perspective.

The model was pre-trained on an impressive 109,066 hours of CGM data, but these came from only 477 participant/session records. The authors acknowledge that the population remains modest and that metabolic responses vary considerably between people, cohorts and sensor devices.

More importantly, better prediction performance is not yet the same as better healthcare.

We still need to understand how well these models perform prospectively in diverse real-world populations, how well calibrated their predictions are, whether performance is consistent across important patient subgroups, and what happens when they encounter the reality of different devices, behaviours, comorbidities and care settings.

And ultimately, the question is not only whether a model can detect metabolic patterns.

It is whether those predictions provide information that clinicians and patients can act on, improve decisions or outcomes, fit into existing workflows, and do so without creating new risks or inequalities.

This is therefore an exciting development. But perhaps its most interesting contribution is that it pushes us closer to a much bigger question: how do we turn increasingly rich streams of continuous physiological data into information that is not only predictive, but genuinely useful in healthcare?

That transition from technical performance to clinical value is where the harder work begins.

https://research.google/blog/glucofm-foundation-model-for-continuous-glucose-monitoring/

13/09/2026

“Superhuman” AI that can spot heart disease from an ECG in under two seconds certainly makes for a striking headline. But for me, the more interesting question is not how fast the algorithm is. It is what happens after those two seconds.

The technology described here has been trained on millions of ECGs and was evaluated in a trial involving around 67,000 patients. It identified up to 81% of patients with heart failure and up to 90% of those with heart valve disease.

That is potentially important because ECGs are already one of the most widely used diagnostic tests in medicine, with around one billion performed globally each year. If information already contained in a routine ECG can help identify patients who should receive an echocardiogram more urgently, AI could potentially help shorten the path to diagnosis and treatment.

But this is also where some nuance is needed.

The AI does not diagnose heart failure or valve disease on its own, nor can it reliably rule them out. Its value is in identifying people who may need further investigation. That makes it less of an autonomous diagnostic tool and more of a potentially powerful triage and decision-support mechanism.

And the real-world impact will depend on much more than accuracy or speed. How many false alerts will clinicians need to manage? Will already stretched echocardiography services be able to absorb additional referrals? How should thresholds be set for different patient groups? Does performance remain consistent across hospitals and populations? And, perhaps most importantly, does introducing the tool actually lead to earlier treatment and better outcomes?

This is why implementation matters as much as algorithmic performance.

There is something very compelling about using AI to extract additional clinical information from tests that are already routinely performed. But the measure of success should ultimately not be whether an AI can analyse an ECG in two seconds. It should be whether those two seconds lead to the right patients receiving the right care sooner.

https://www.theguardian.com/technology/2026/aug/31/superhuman-ai-tool-spots-heart-disease

31/07/2026

What people ask a chatbot about health may tell us something about the healthcare system around them, not only about their interest in AI.

A new study in Nature Health analysed 1.7 million health-related conversations with Microsoft Copilot across 109 countries and regions. One of its most interesting findings was that countries with lower confidence in hospitals tended to have a higher proportion of AI conversations about health.

The types of questions also differed. In lower-income countries with younger populations, conversations were more often about broad health information and education. In wealthier countries with older populations, people were more likely to ask about symptoms, care pathways and how to navigate the healthcare system. Countries with more structured health systems also had more questions about medical paperwork.

This is a useful reminder that people do not use health technology in a vacuum. They use it in response to the access, trust, complexity and gaps they experience in their local healthcare environment.

At the same time, the findings need careful interpretation. This was a country-level analysis, so it cannot show that the individuals who distrust hospitals are the same people turning to AI. It covers one platform, whose users may not represent the wider population in each country. It also identifies associations rather than causes. A dataset can be very large and still be affected by who is - and is not - represented in it.

Perhaps the most important question is therefore not whether conversational AI is replacing or supporting healthcare. It may be doing both, depending on the person and the context.

For some people, it could improve health literacy, help them prepare for an appointment or make a confusing system easier to navigate. For others, particularly where trust or access is low, it could become a substitute for professional care—even when the information provided is incomplete or wrong.

That makes context-sensitive design essential. Health chatbots need reliable safety boundaries, culturally and linguistically appropriate information, clear routes to human care and ongoing monitoring of how they are used across different populations.

The study offers an important starting point. The next step is to connect these broad usage patterns with individual experiences and health outcomes, so we can understand where AI is genuinely closing gaps, and where it may quietly be creating new ones.

https://www.nature.com/articles/s44360-026-00174-2

17/07/2026

Google’s launch of a $9.99-per-month AI health coach is another sign that health support is moving beyond simply counting steps or tracking sleep.

The new service promises to combine information about fitness, sleep, nutrition, menstrual cycles, mental well-being, environmental factors and, where users permit it, medical records. In principle, this could help people make better sense of health information that is currently scattered across different apps and devices.

The appeal is understandable. Many of us collect far more health data than we can meaningfully interpret. A conversational coach that connects the dots, adapts to personal circumstances and offers practical guidance could make this information more useful.

But the important question is not only how personalized the advice feels. It is whether that advice is reliable, safe and genuinely helpful over time.

How will the system respond when wearable data are incomplete or misleading? Will it recognise when a person needs professional medical care rather than another lifestyle suggestion? How transparent will it be about why it has reached a particular conclusion? And how carefully will highly sensitive information about health, reproductive cycles, mental well-being and medical history be governed?

There is also a risk that a confident and friendly AI coach may appear more authoritative than the evidence behind its recommendations justifies. Personalization can increase relevance, but it can also increase trust, even when the underlying advice remains uncertain.

The subscription model raises another question. If these tools become an increasingly important route to preventive guidance and health literacy, will their benefits mainly reach people who already have compatible devices, digital confidence and the ability to pay?

This development has real potential. But success should not be measured only through engagement, subscription numbers or how often users return to the app. We also need evidence that the coach improves meaningful health outcomes, supports informed decisions, protects sensitive data and knows where its role should end.

The future of digital health may well include AI coaches. The challenge is ensuring that they become trustworthy companions rather than simply persuasive ones.

https://techcrunch.com/2026/05/07/googles-9-99-per-month-ai-health-coach-launches-may-19/

08/07/2026

A recent survey found that 57% of healthcare professionals had either encountered or used unauthorized AI tools at work. That figure is concerning, but perhaps not entirely surprising.

Healthcare professionals are under constant pressure to work faster, manage growing administrative demands, and make sense of increasingly complex information. When approved tools are unavailable, difficult to access, or do not fit everyday workflows, people may turn to familiar public AI tools to fill the gap.

This does not make “shadow AI” harmless. Using unauthorized tools can create serious risks around patient confidentiality, data security, inaccurate outputs, and unclear accountability, particularly when AI is used to support direct patient care.

However, simply introducing stricter rules or blocking access is unlikely to solve the underlying problem. Unofficial use may also be a signal that staff needs are not being adequately addressed.

One result I found particularly important was that administrators were three times more likely than healthcare providers to be involved in developing AI policies. Policies designed without sufficient clinical input may look strong on paper while remaining disconnected from how work is actually carried out.

The survey itself should also be interpreted cautiously. It included 518 professionals, relied on self-reported behaviour, and was commissioned by a company that provides healthcare AI solutions. It nevertheless raises a useful question: are organisations creating AI policies around their staff, or with them?

Responsible AI adoption requires more than access controls. It requires clear guidance, approved tools that genuinely support real work, practical training, safe opportunities to experiment, and meaningful involvement from the people expected to use these systems.

The goal should not be to eliminate curiosity. It should be to create an environment where curiosity can be explored safely, transparently, and with appropriate safeguards.

https://www.fiercehealthcare.com/digital-health/nearly-fifth-healthcare-professionals-use-unauthorized-ai-tools-work

23/05/2026

Smart pills that can diagnose, monitor, and even treat disease from inside the body sound like science fiction. But research into ingestible electronics is steadily moving this idea closer to clinical reality.

The promise is easy to understand. For people living with gastrointestinal conditions, diagnosis can be slow, invasive, uncomfortable, and expensive. A swallowable capsule that could detect inflammation, monitor gut health, deliver treatment locally, or even collect a small tissue sample could make care more precise and less burdensome.

But as exciting as this is, the article is also a reminder that innovation in healthcare is rarely just about what is technically possible.

For these devices to improve care, they will need to prove not only that they work in controlled settings, but that they are safe, reliable, clinically meaningful, acceptable to patients, trusted by clinicians, and accessible within real healthcare systems. Questions around power, biocompatibility, data quality, device retention, regulatory approval, reimbursement, and responsibility for interpreting the data are not small details. They will shape whether these technologies become useful tools or remain impressive prototypes.

There is also a broader point here. The future of digital and AI-enabled health will not only be software-based. Some of the most interesting developments may come from the combination of sensors, microelectronics, materials science, robotics, and medicine. These systems could generate new kinds of health data from places we previously could not easily access.

That is both exciting and challenging.

The real test will be whether these technologies are developed around clear clinical needs, meaningful patient benefit, and careful evaluation rather than technological possibility alone. A pill that can “sense and act” inside the body is a remarkable achievement. Making sure it acts in ways that genuinely improve care is the harder, and more important, task.

https://spectrum.ieee.org/ingestible-electronics

21/05/2026

The more AI becomes part of everyday health information seeking, the more we need to be honest about both its promise and its limits.

A recent article in The Guardian discusses a study in Nature Medicine evaluating ChatGPT Health in simulated medical scenarios. The findings are concerning: in many cases where urgent care was needed, the system reportedly failed to recommend immediate medical attention. It also appeared inconsistent in recognising suicidal ideation when additional contextual information, such as normal lab results, was included.

This does not mean AI has no place in health. Quite the opposite. AI tools can help people prepare for consultations, understand medical terminology, organise questions, and access information more easily. For many people, especially those who feel dismissed, overwhelmed, or unsure where to start, this can be genuinely valuable.

But there is a major difference between supporting understanding and making triage decisions.

The risk is not only that AI may give wrong advice. It is that it may give advice with confidence, in moments when people are anxious, vulnerable, or looking for reassurance. A false sense of security can be dangerous, especially when symptoms require urgent medical assessment.

This is why independent evaluation, clear safety standards, transparency, and strong guardrails are not optional extras. They are essential if AI is going to be used responsibly in health contexts.

The discussion should not be reduced to “AI is good” or “AI is dangerous”. The more useful question is: what should AI be allowed to do, under what conditions, with what evidence, and with what safeguards?

For now, I think the key message remains: AI can support health literacy, but it should not replace clinical judgement, emergency care pathways, or human support in moments of crisis.

https://www.theguardian.com/technology/2026/feb/26/chatgpt-health-fails-recognise-medical-emergencies

Digi-Bridges

11/10/2026

05/10/2026

20/09/2026

17/09/2026

13/09/2026

31/07/2026

17/07/2026

08/07/2026

23/05/2026

21/05/2026

Adresse

Öffnungszeiten

Website

Benachrichtigungen

Verknüpfungen

Teilen

Kategorie

Antrag auf Entfernung des Eintrags

Montag	09:00 - 17:00
Dienstag	09:00 - 17:00
Mittwoch	09:00 - 17:00
Donnerstag	09:00 - 17:00
Freitag	09:00 - 17:00