Best of LinkedIn: Health Tech CW 38/ 39

00:00:00: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about health tech from CW-ThirtyEightandThirtyNine.

00:00:08: Frenness equips HealthTech providers with The Market Intelligence to identify which hospitals to target... ...and how to reach decision makers for hospital digitalization as a result of the Krakenhausuchum Schizetz.

00:00:19: You can find more info in description.

00:00:23: Imagine buying a multi million dollar medical AI system right?

00:00:27: And it aces its licensing-style benchmark exams with a ninety two percent success rate.

00:00:32: Wow, so theoretically It's flawless.

00:00:34: exactly the theoretical capability is totally flawless.

00:00:38: But the moment you drop that exact same AI into the clinical workflow of a real functioning hospital Its accuracy immediately plummets to forty five percent.

00:00:48: That Is just a jarring reality.

00:00:50: and honestly?

00:00:51: It's exactly what we're tearing in today.

00:00:53: It really is The biggest hurdle right now.

00:00:55: yeah.

00:00:55: And for those of you listening, whether your digital transformation leaders product managers or tech professionals mapping out Your infrastructure.

00:01:03: For the next five years this deep dive is for You.

00:01:05: we're skipping The fluff today.

00:01:07: no fluff just the practical impact Of the latest developments

00:01:11: right?

00:01:12: We're breaking down that messy reality of AI workflows the physical data Infrastructure you need to actually make these systems function.

00:01:19: how surgical robotics are scaling up and uh...the global market mechanics dictating where This Tech lands first.

00:01:25: It's a packed agenda.

00:01:26: It is!

00:01:27: So let's start with that massive drop in AI accuracy you mentioned?

00:01:31: Yeah, so that ninety-two to forty five percent drop... That's real statistic highlighted recently by Parminder Baudia and I think it perfectly encapsulates the current maturity phase of health tech.

00:01:42: Because its not like the AI just suddenly forgot medical science

00:01:46: right?!

00:01:47: No Not at

00:01:47: all!!

00:01:48: It fails because answering multiple choice medical question In sterile isolated testing environment Well, it's a fundamentally different computational problem than operating within a fragmented chaotic clinical workflow.

00:02:01: Yeah I mean if you think about software engineering that makes perfect sense.

00:02:04: It is like writing this absolutely flawless piece of code on your local machine but then trying to deploy onto some legacy production server

00:02:12: A server with a thousand conflicting dependencies right?

00:02:15: Exactly and outdated operating systems.

00:02:17: The Code isn't the Problem!

00:02:18: The Environment just rejects it.

00:02:20: That's great analogy.

00:02:21: And that hostile environment is changing how product development has to happen moving forward.

00:02:26: John Exario-Sillies actually brought a recent report from Define Ventures To Our Attention, which puts a hard number on this friction.

00:02:33: Oh yeah?

00:02:33: What did they find?

00:02:34: They found that eighty two percent of healthcare provider leaders now cite workflow integration not the underlying AI technology itself as their top barrier to adoption.

00:02:45: Wow!

00:02:46: Eighty Two Percent Yeah

00:02:48: And that specific metric jumped fifty one points in just a single year.

00:02:52: So for you listening, if you are designing a digital product roadmap right now That eighty two percent figure is your primary red flag?

00:03:00: Absolutely You cannot build the feature Just assuming that regulatory approval means victory.

00:03:05: Right?

00:03:06: Parthik Kodemuri and Austin Chang both pointed this out super clearly.

00:03:10: Getting an FDA clearance on new diagnostic tool does not automatically drive clinician adoption.

00:03:15: It really doesn't.

00:03:16: If your new feature requires a triage nurse to say open the second screen, copy and paste a value or break their routine for even ten seconds.

00:03:25: Your user adoption rate will flatline.

00:03:27: Ten Seconds is all it takes to kill product in hospital.

00:03:30: It's.

00:03:31: Joseph T. Shulin framed this shift brilliantly.

00:03:34: He noted that industry has officially moved past asking can AI do this?

00:03:39: To much harder operational question

00:03:42: Which?

00:03:43: how we actually deploy AI across health system

00:03:46: Exactly.

00:03:48: It requires turning complex, disparate information into immediate frictionless action for the end user.

00:03:55: But okay if that technology only proves its actual worth in the messy reality of The clinic how are regulators supposed to handle it?

00:04:03: It's a huge paradox For them

00:04:05: right because you can't just test an adaptive workflow tool and a pristine lab but You also obviously can't Just unleash untested AI on Actual patients.

00:04:14: well Regulators are actually being forced to adapt their mechanisms in real time, to solve that exact paradox.

00:04:20: David Lobajo Izcairdo highlighted a massive structural shift with the FDA's new Tempo program.

00:04:26: Okay, Tempo?

00:04:27: What is this focus there?

00:04:29: This initiative was designed to allow selected generative AI healthcare tools To gather real world evidence safely through deployment.

00:04:36: Wait, let's look at the mechanics of that for a second.

00:04:38: Are they running these models in a shadow mode alongside the clinicians or are they actively influencing patient care while still learning?

00:04:47: They're building a bridge between innovation and clinical practice.

00:04:51: It allows the tools to process The real-world complexities of the hospital

00:04:55: like the messy data inputs and all that

00:04:57: exactly the messy Data the unique patient populations.

00:05:01: they process all of that in a controlled manner learning from actual use rather than simulated data.

00:05:07: That's

00:05:07: fascinating!

00:05:08: And this isn't just an isolated US strategy either, Robert Z Phillips noted in the recent IMD-RF forum that regulators globally are moving toward reliance frameworks

00:05:19: Meaning they're leaning on each other?

00:05:20: Yes, leaning on one another methodologies because no single agency can keep pace with a software life cycle alone.

00:05:27: That makes sense, but if these models are gathering real-world evidence on the fly... The patient absolutely has to be factored into this loop.

00:05:34: Oh hundred percent!

00:05:35: I saw Donna Crier is starting a new weekly tracker specifically to keep patients informed of healthcare AI rules because historically, Patients ARE THE LAST ONES TO UNDERSTAND THE ALGORITHMS MAKING DECISIONS ABOUT THEIR CARE

00:05:46: And patients absolutely have to understand the infrastructure Because the systemic risks are massive here.

00:05:53: Kyar and Galacy brought up a huge structural vulnerability regarding the European Health Data Space, or EHDS.

00:05:59: The EHDS?

00:06:01: Okay so the core logic of that is facilitating data reuse across international borders to train better models right which you know sounds great on paper.

00:06:09: more data generally means more robust AI.

00:06:12: it does until you look at the mechanism of contamination.

00:06:16: She points out that if a single inherently biased dataset gets shared across this massive international infrastructure, The risk isn't just one flawed algorithm

00:06:25: That bias data get baked into foundation.

00:06:28: Exactly It's structural.

00:06:30: Basically it is a corrupted foundational code library.

00:06:33: If an open source library used by thousands of developers has hidden vulnerability Every application built on top of it instantly compromised.

00:06:42: That's a perfect way to look at

00:06:43: it.

00:06:44: So if the EHDS shares bias data, It becomes the flawed foundational code for Europe's entire medical AI ecosystem.

00:06:51: by the time someone notices The output is biased tracing it back To the original dataset across borders Is nearly impossible

00:06:58: which means you cannot fix the AI workflow integration gap or secure these models from systemic bias without A rock solid highly transparent data foundation.

00:07:09: The software layer is entirely dependent on the physical data infrastructure generating the inputs.

00:07:14: Absolutely Which actually a perfect transition into our second theme?

00:07:18: If the EHDS is code repository, the actual hardware generating medical data Is getting massive architectural upgrade.

00:07:25: right now

00:07:26: It really is!

00:07:27: Siemens Healthineers just opened a one hundred million euro Photon counting sensor factory in Forthheim Germany.

00:07:33: Yeah, Baron Montag, Baron Onesorga and Katarina Hitzels were all sharing developments on that.

00:07:39: They're producing cadmium telluride or CDTE sensors for next-generation CT scanners.

00:07:45: And to understand why this physical upgrade matters to the digital roadmap you have to look at a mechanism of how older CT scanders capture data versus new method.

00:07:55: Right because traditional scanners convert X-rays into visible light first then convert it in an electrical signal

00:08:02: Exactly

00:08:03: And you lose information and introduce noise at every single conversion step.

00:08:07: But these cadmium telluride crystals eliminate that middle step entirely, they convert individual X-ray photons directly into digital information.

00:08:16: It is literally digitizing the physical scan on a subatomic level

00:08:20: which delivers exponentially richer sharper diagnostic data while potentially lowering radiation dose the patient absorbs.

00:08:27: it's incredible tech but hardware alone doesn't solve this problem.

00:08:31: Having a space-age CT scanner generating incredibly rich data files is totally useless if it bottlenecks the radiologist.

00:08:39: Chrissy Hall pointed out that scaling something like cardiac CT requires physical technology to be built in lockstep with operational tools.

00:08:47: So you need integrated patient routing, revised clinical order sets

00:08:51: Yeah and AI scoring algorithms.

00:08:53: build into system natively not just bolted on after fact.

00:08:58: By the way, if you are finding this deep dive useful for tracking how these massive physical and digital systems are converging make sure to subscribe so that you catch our future additions.

00:09:07: Definitely!

00:09:08: Okay but I want look closely at a study Auntie Helman shared because it perfectly illustrates this convergence.

00:09:14: She highlighted her studies showing how combining compressed sensing accelerated MRA with four D flow MRI allows clinicians comprehensively image complex aortic dissections.

00:09:24: It essentially combines morphological assessment with hemodynamic quantification in one single exam,

00:09:31: which is a massive leap forward for cardiovascular diagnostics.

00:09:34: Really?

00:09:35: But what

00:09:36: hold on hemo dynamic quantification alongside morphological assessments?

00:09:42: For those of us on the IT and digital transformation side who aren't radiologists What does that actually mean for the patient lying inside that machine And the data it generates?

00:09:52: let's break down.

00:09:53: Morphology is just the physical shape, getting a high-resolution three D picture of the aorta to see if it's bulging or torn.

00:10:00: Okay HEMO dynamics is the actual fluid dynamics measuring the precise speed pressure and direction of the blood flowing through that tear.

00:10:09: Got it!

00:10:10: And historically getting both required multiple time consuming highly stressful tests right?

00:10:15: Exactly this new approach captures the physical structure and the realtime fluid dynamic simultaneously in one short exam.

00:10:23: That is incredible for the patient.

00:10:25: But I have to push back here.

00:10:27: on the data management side, we are talking about insanely rich multi-dimensional imaging data from these new photon counting CTs and four D MRI's plus all the real time EHR data genomics wearables...

00:10:40: It's a lot of data!

00:10:41: ...it's a mountain of data with massive tech companies aggregating all this across entire health systems.

00:10:47: aren't we just creating a much bigger haystack instead?

00:10:52: That is the central challenge of digital transformation right now.

00:10:56: If you just dump that data into a dashboard, You paralyze the clinician!

00:11:00: Right?

00:11:01: This is why data platforms have to evolve from passive storage repositories To active curation engines and we are seeing Oracle making A massive play-to be that engine Seema.

00:11:12: Verma highlighted their new AI powered oracle health patient portal Which actively translates in curates clinical data so patients can actually understand it

00:11:21: And the scale they are deploying this at is just aggressive.

00:11:24: Dan Spellman noted that Quorum Health is standardizing eleven acute care hospitals on Oracle's platform.

00:11:30: Yeah, and over in the UAE Nick Barnes, Amer Hawaidi & Jamie Blomert highlighted that M-forty two health as adopting oracle health data intelligence specifically to power precision healthcare across their network.

00:11:43: But to address your pushback about the haystack, The curation is where the real value lies.

00:11:48: Joe Corvea shared that Oracle's clinical AI agent has already saved physicians over four hundred thousand hours.

00:11:55: It isn't just aggregating data it is synthesizing.

00:11:58: I

00:11:58: am still a little skeptical of that mechanism.

00:12:01: honestly If an AI curates my complex medical history To save a doctor time and decides to hide A specific data point it deems irrelevant aren't we introducing a massive clinical blind spot?

00:12:13: That's a valid concern.

00:12:14: How does the physician trust an AI agent's curation when their liability is on-the

00:12:18: line?".

00:12:19: They trust it.

00:12:19: through traceability, A Clinical AI Agent isn't allowed to just present final conclusion and hide its work When it surfaces at risk – say flagging patient for early sepsis.

00:12:31: The UI must provide immediate clickable audit trail back to exact lab results, vital sign trends & clinical notes that triggered the flag.

00:12:39: So it doesn't delete the haystack.

00:12:41: It highlights the needles and provides a map showing exactly how it found them

00:12:45: Precisely, it creates a verifiable chain of logic.

00:12:49: That makes sense.

00:12:50: And once that curated data services the right diagnosis you actually have to treat the patient.

00:12:55: But here is the major structural shift.

00:12:57: The actual physical intervention —the surgery—is moving onto the exact same kind of integrated digital platforms we just discussed.

00:13:05: We are seeing a massive shift from traditional isolated surgical tools into an augmented robotic ecosystem and the breadth of this shift across different medical disciplines is staggering.

00:13:15: Just looking at the updates over these two weeks, The common denominator isn't specific body part.

00:13:20: it's the move toward smart co-pilot model.

00:13:24: Yeah we saw Mark Stoffel detail in ARPH award for Philips to advance AI enabled robotics stroke thrombectomy care

00:13:32: And Ryan Fukushima highlighted another ARPA-H award for Tempest to build an Autonomous AI Agent For Heart Failure.

00:13:39: Plus, Narendra Gogna noted stealth access trained over thirty surgeons in cranial and spine robotics...in just a single month!

00:13:46: And Emily Ellswick celebrated the Altavisa's device scaling up therapy for urge urinary incontinence —the entire surgical landscape is digitizing…

00:13:55: It really is—and what is crucial for adoption here ...is that manufacturers are finally understanding workflow integration.

00:14:02: There's that theme again.

00:14:03: Exactly, Jonathan Hatwell shared Medtronix new partnership with Cornerstone Robotics instead of trying to force a hospital use one proprietary robotic system for everything.

00:14:13: they are expanding Surgeon Choice supporting open surgery laparoscopic and robotic platforms under unified ecosystem

00:14:21: which significantly lowers the barrier entry.

00:14:24: Matt Anderson share.

00:14:25: great example.

00:14:26: this FDA just cleared ligature RAS.

00:14:30: This is an energy device that surgeons have already trusted in over four point five million open and laparoscopic procedures.

00:14:37: Now they're putting the exact same-trusted tool onto a robotic arm for U.S.

00:14:41: market!

00:14:42: That's exactly how you drive adoption, You don't force surgeon to learn new tools interacts with human tissue... ...you just give them more precise robotic interface from the tool they know

00:14:53: And clinical milestones are really reflecting this strategy.

00:14:57: Jonathan Sayle posted about Whittington Health performing the UK's first Medtronic robotic mini-bypass and their first robotic myomectomy.

00:15:05: Yeah, and Mary JoLed M also reflected on a Hugo Rass showcase drawing the direct line from rigorous quality system engineering to these tangible patient impacts.

00:15:15: Wait when we talk about this smart co-pilot ecosystem how scalable is this really?

00:15:21: Are cost constrained hospitals supposed to buy different multi million dollar robot for every single procedure type?

00:15:27: No, the economics of that would collapse immediately.

00:15:30: The scale only happens when the platforms are versatile and augment human expertise rather than trying to overwrite it with hyper-specialized automation.

00:15:39: Derek Monogh, Suzanne Van Eindhoven & Hogan Vatansavan all noted a staggering metric.

00:15:45: What was it?

00:15:46: Medtronics healthcare technologies are now impacting more than seventy nine million people every year.

00:15:52: That is roughly two people every single second.

00:15:54: Yeah

00:15:55: You don't reach a footprint of two people per second by replacing clinicians with incredibly niche, expensive robots.

00:16:11: But here is the hard truth about these robotic ecosystems and AI platforms.

00:16:15: You can engineer the smartest surgical co-pilot in the world, but if the local hospital's financial reimbursement model doesn't actually pay for preventative efficiency...

00:16:25: That advanced robot will sit at a warehouse?

00:16:27: Exactly!

00:16:28: Which brings us to our final theme – The global market mechanics dictating where this technology goes

00:16:34: And the center of gravity in medtech is absolutely shifting right now.

00:16:38: Deng Wu Kim pointed out that while North America has historically been dominant, the Asia-Pacific region is rapidly emerging as the fastest growing MedTech region globally.

00:16:47: And what's the mechanism driving this?

00:16:49: Is it purely a population scale or are there structural reasons?

00:16:53: they're adopting technology faster?

00:16:55: It is deeply structural – these emerging markets are heavily rewarding regulatory agility.

00:17:01: Furthermore… They prioritize proven economic value over legacy scale!

00:17:07: If a digital health startup can prove their technology is cost-effective and improves outcomes quickly, these health systems move incredibly fast to integrate them.

00:17:16: We are seeing the financial mechanics of that adoption firsthand.

00:17:19: Dr Trinh Hoang Ha pointed out that Vietnam is undergoing massive structural transition.

00:17:24: right now They're moving from traditional fee for service model To DRG based reimbursement model Diagnosis related groups.

00:17:31: That's fundamental rewiring of healthcare economy.

00:17:35: It really does.

00:17:36: Let's break that economic logic down for the listeners, because it explains The Tech Boom perfectly.

00:17:40: Go for it!

00:17:41: Under a traditional fee-for-service model A hospital generates revenue by doing more things More blood tests More MRI scans More days keeping patient in bed.

00:17:49: But

00:17:50: volume over value

00:17:51: Exactly.

00:17:51: but a DRG Model means Hospital gets paid single flat rate For specific diagnosis Like knee replacement.

00:17:58: Regardless of how many resources they use

00:18:01: Efficiency becomes paramount.

00:18:03: Suddenly, a piece of health tech that makes the surgery more precise and gets the patient discharged two days earlier isn't a loss-of-bed revenue.

00:18:09: it is pure profit margin.

00:18:11: Wow

00:18:12: That flip in financial incentives Is the exact mechanism driving advanced tech adoption In these regions.

00:18:17: You can see the immediate results Of those shifting incentives too.

00:18:20: Martin Pantow shared that Indonesia just successfully went live with its first somatom propulse installation bringing advanced cost-effective diagnostics into regional communities.

00:18:31: Amazing!

00:18:32: When you combine these shifting payment models with digital adoption, the global financial projections are massive.

00:18:38: Sujit Dokey reported that the Global MedTech market is currently projected to hit

00:18:46: With Asia accelerating.

00:18:47: like that, Europe is finding itself in a very delicate balancing act.

00:18:51: Martin Ferrer and Elizabeth Stoudinger noted the European med tech sector as massive economic engine.

00:18:57: Oh yeah!

00:18:57: It supports roughly one million jobs... ...and generates EURs in revenue.

00:19:02: But

00:19:02: they implicitly recognized to maintain leadership against global competition….

00:19:07: …Europe desperately needs stronger more agile partnerships between policymakers industry and healthcare providers themselves

00:19:14: Because the competitive landscape is just not slowing down.

00:19:18: Simon Philip Ross highlighted that time just named GE Healthcare and Inteler AI among world's top health tech companies for twenty-twenty six.

00:19:27: The companies at Wynn will be ones who can navigate these fragmented global frameworks.

00:19:31: Exactly!

00:19:32: We saw Sonia Weasley, Burt Van Meurs & Karen Kettleis reflecting on recent ESC Congress in Munich.

00:19:39: Cardiovascular disease remains an overwhelming global burden, obviously.

00:19:43: But they noted that collaborative policy frameworks like the EU Safe Hearts Plan are actual mechanisms driving how new cardiovascular innovations get deployed across patient pathway.

00:19:54: So, synthesizing all of this... For digital transformation professionals listening right now looking at their product and infrastructure roadmaps for next three to five years.

00:20:03: what is ultimate takeaway from these shifts?

00:20:06: The overarching lesson is that commercial victory over the next decade will not belong to the engineering team with the flashiest algorithm or most complex robotic arm.

00:20:15: It belongs to workflow.

00:20:17: it belongs to organizations they can seamlessly integrate into hostile legacy clinical workflows, it belongs those who navigate complex multi-jurisdictional regulations while proving clear immediate economic value of hospitals transitioning to DRG models.

00:20:33: whether you're building AI agents deploying photon counting data infrastructure or scaling surgical robotics, the only metric that matters is a measurable scalable patient outcome.

00:20:44: It all comes back to the foundation you build on.

00:20:46: we started this deep dive exploring the messy reality of AI integration and the massive structural risk of bias data spreading across international borders.

00:20:54: so as you look at your own infrastructure projects this week I want to leave you with a final thought to mull over

00:20:59: what's that

00:21:00: if an AI model learns perfectly but operates on an inherently flawed international data sharing network.

00:21:07: Who is ultimately responsible when it makes a critical error?

00:21:11: Is it the developer who wrote The Perfect Code, the local hospital that deployed into their workflow or the International Regulator?

00:21:25: Also, check out our other editions on Cloud Insights and Sovereignty, Digital Products & Services, AI & Agenetic Systems, Green ICT & Sustainable AI, ICT and Tech Insights, and DefenseTech.