Best of LinkedIn: Health Tech CW 40/ 41

00:00:00: This episode is provided by Timus Allgeier and Frennus, based on the most relevant LinkedIn posts on health tech in CW-Forty and Forty-One.

00:00:08: Frennus equips product and strategy teams with market and competitive intelligence to navigate the health tech landscape.

00:00:15: Welcome to the deep dive.

00:00:17: Our mission today.

00:00:19: To really cut through the noise, giving the most relevant strategic health tech insights from LinkedIn these past couple of weeks, calendar weeks, forty and forty-one.

00:00:28: If you need that shortcut to what product and strategy folks are actually talking about, well, you're in the right place.

00:00:33: Yeah, we saw some really interesting stuff.

00:00:34: What's kind of fascinating looking across the board is the shift.

00:00:37: It feels like we're moving past just talking about cool tech.

00:00:39: Right,

00:00:40: the hype phase.

00:00:40: Exactly.

00:00:41: Yeah.

00:00:41: Now it's squarely focused on you know, pragmatic adoption.

00:00:45: The big theme is execution quality and how well these things actually integrate into clinical operations.

00:00:50: That's what's making the difference in the real world, not just how fancy the tech is underneath.

00:00:55: Makes sense.

00:00:55: It's all about making it work for the people using it, clinicians, patients.

00:01:00: So let's kick off.

00:01:02: where there's often the most friction.

00:01:04: EHRs, workflows, and that constant battle against admin, Bergen, and burnout.

00:01:09: Absolutely.

00:01:10: And that's where we saw probably the biggest story, certainly from CW-Forty, Oracle Health, Seema Verma posted about their new AI-powered EHRs, specifically for ambulatory clinics.

00:01:21: And what's key here is... This isn't just layering AI onto the old Cerner system.

00:01:26: Oh, right.

00:01:26: This is post-acquisition, built differently.

00:01:29: Yeah, built differently, seemingly from the ground up.

00:01:31: They

00:01:31: threw around some pretty heavy tech terms in that announcement.

00:01:34: Purpose-built, agentic AI, knowledge graphs, semantic index.

00:01:39: How should we, you know, translate that for folks who are in deep in data science?

00:01:42: What does it actually do?

00:01:43: Okay, think of it like this.

00:01:44: The agentic AI is sort of like a super specialized assistant.

00:01:48: It listens in ambiently to the doctor-patient conversation using voice tech.

00:01:52: captures the data naturally.

00:01:53: Then the knowledge graph.

00:01:55: That's like this deep map of clinical information, how symptoms, treatments, outcomes all connect.

00:02:00: It gives the AI real clinical understanding.

00:02:03: Ah, so it gets the context, not just the words.

00:02:06: It's not just glorified transcriptions.

00:02:07: Precisely.

00:02:08: And the semantic index.

00:02:09: That helps it grasp the meaning, the intent behind the conversation.

00:02:13: So it can summarize accurately, maybe suggest next steps, fill out the chart automatically.

00:02:18: The goal... like Verma put it, is clear.

00:02:20: Let doctors spend more time to heal than chart, shifting that load.

00:02:26: That's

00:02:26: the big picture.

00:02:27: But I also loved seeing those really practical, almost granular fixes getting attention.

00:02:32: Like Craig Joseph's post about Stanford Children's Health fixing that weirdly persistent EHR flaw, assigning phone numbers to babies.

00:02:39: It sounds almost funny, doesn't it?

00:02:41: Yeah.

00:02:41: But that tiny data error causes real problems down the line.

00:02:44: Yeah.

00:02:45: Like what?

00:02:46: Well, think about it.

00:02:47: patient portal invites going to the wrong place.

00:02:49: Or worse, adolescent confidentiality getting broken because the system defaults to a parent's contact info.

00:02:55: It really undermines trust and safety.

00:02:57: So Stanford fixed it with, what do they call it, age aware rules?

00:03:01: Exactly.

00:03:01: Basically just smart logic.

00:03:03: Art stops preventing phone or email entries for infants, making sure there are distinct contact fields for teens versus parents.

00:03:10: It's a great reminder that, you know, basic data hygiene and smart rules are just as vital as fancy AI for safety and getting people to actually use the platform.

00:03:20: And that solid data foundation is what enables the more advanced AI stuff, right?

00:03:24: Which ties back to what Mayank Kumar shared about EHR Foundation models.

00:03:27: Right.

00:03:28: These aren't just summarizing notes.

00:03:29: They're more like LLMs trained on huge amounts of longitudinal health data years of patient records.

00:03:34: They're learning the actual language of patient care, the patterns.

00:03:38: So they're kind of natively fluent in clinical context.

00:03:41: Okay, let's switch gears.

00:03:42: Next big theme, AI in imaging and how it's actually getting deployed in clinics.

00:03:47: It feels like the focus shifted from just building massive models to making them integrate more smartly.

00:03:52: integration is definitely the word.

00:03:54: Ryan Fukushima highlighted this with the Flexinesis toolkit.

00:03:57: I mean, imagine being a health system trying to manage maybe fifty different AI tools.

00:04:02: One for radiology, another for pathology, another for genomics.

00:04:06: Yeah,

00:04:06: that sounds like a nightmare for clinicians.

00:04:08: Just overwhelming.

00:04:09: Totally.

00:04:10: So Flexinesis is trying to solve that.

00:04:12: It's built for integrating data from multiple omics sources, genomics, proteomics, imaging, etc.

00:04:18: The idea is one adaptable model.

00:04:20: can handle various tasks instead of juggling dozens of separate ones.

00:04:24: It cuts down the management headache and should speed up adoption because the workflow is unified, makes AI actually usable.

00:04:31: And we're seeing real results when hardware and software work together, especially in cardiac imaging.

00:04:36: Atul Gupta mentioned this, pointing to posts from Manny Vembar and Mathias Goyen about Phillips advancements in cardiac CT.

00:04:43: Absolutely.

00:04:44: Two really cool things happening there simultaneously.

00:04:46: First, spectral CT.

00:04:48: This is huge because it captures high and low energy x-ray data at the same time.

00:04:52: Why did that matter?

00:04:53: It helps cut through the blooming artifact you get from dense calcium and

00:04:56: arteries.

00:04:57: Ah, so you can see past the calcium blockage more clearly.

00:05:00: Exactly.

00:05:01: It's a potential game changer for diagnosis.

00:05:04: And then you have the AI algorithms cleaning things up further.

00:05:07: Right.

00:05:08: Like what?

00:05:08: Well, they mentioned precise cardiac that uses AI to correct for the heart constantly moving so you don't get blurry images.

00:05:16: And then precise image reduces noise, which is great because it means you can get sharp images with a lower radiation dose.

00:05:22: Safer and more effective.

00:05:23: Nice.

00:05:24: A perfect example.

00:05:25: smart hardware plus smart software.

00:05:27: And the specificity of some AI tools is just incredible.

00:05:31: Yol Baca has pointed out Mayo Clinic's Radon GPT.

00:05:34: That's an AI agent specifically designed for radiation oncology planning.

00:05:39: Super niche.

00:05:39: Yeah.

00:05:40: These aren't generalists.

00:05:41: They're becoming deep experts.

00:05:42: Yeah.

00:05:42: And speaking of efficiency, Yuanfang Li shared work on a reinforcement learning algorithm called ORAPO.

00:05:48: This is really interesting because historically, medical AI needed massive amounts of training data.

00:05:54: which can be hard to get, especially high quality data.

00:05:57: Exactly.

00:05:58: But or appeal seems to get around that.

00:06:00: It's achieving top tier performance, like generating complex chest x-ray reports.

00:06:05: But using get this two to three orders of magnitude less training data.

00:06:10: Wow.

00:06:11: So like a hundred to a thousand times less data.

00:06:14: Something like that.

00:06:15: Think about what that means for developing AI for rare diseases or specialized areas where you just don't have huge data sets.

00:06:22: It lowers the barrier significantly.

00:06:24: Okay, let's pivot again.

00:06:25: Theme three, precision medicine, genomics, and just the sheer scale needed for truly personalized care.

00:06:32: Starts with this idea of the digital twin.

00:06:34: Sounds sci-fi, but it's real.

00:06:36: It is.

00:06:37: Jai Tamar had a great post-breaking it down.

00:06:39: Essentially, it's a dynamic virtual model of a patient, maybe even just an organ.

00:06:43: Constantly fed data scans, genomics, real-time stuff from wearables.

00:06:47: So what are the practical uses right now?

00:06:48: Where is it actually adding value today?

00:06:50: Two

00:06:50: big areas seem to be emerging.

00:06:52: Drug development.

00:06:53: Sanofi's using digital twins to test how drugs might work, side effects, etc.

00:06:58: on virtual patients before human trials saves time and money.

00:07:01: And the other is surgical planning.

00:07:04: Institutes like SGPGI in India are using them for reversals, simulating complex operations, predicting how different approaches might turn out.

00:07:12: The potential seems enormous, but Tamar also brought up the tough questions, didn't she?

00:07:16: Ethics, access.

00:07:18: Yeah, the unavoidable ones.

00:07:19: Precision medicine is already often out of reach for what, eighty percent of people?

00:07:24: If... Digital twins need all this constant, detailed data.

00:07:28: Could they make that gap even wider?

00:07:30: And the data itself, who actually owns your virtual self?

00:07:33: That continuous stream of deeply personal health information.

00:07:37: Massive questions for regulators, for sure.

00:07:40: But that continuous data stream is also the key to the proactive care vision.

00:07:44: Jan Bigger noted this common thread between Google's personal health agent idea and WH Health operating system.

00:07:50: Let's dig into that Google PHA, the personal health agent.

00:07:53: Sounds like an AIT.

00:07:54: team dedicated to one person.

00:07:55: That's

00:07:56: a pretty good way to put it.

00:07:57: UC Matthias' research showed it's not just one big AI trying to do everything.

00:08:01: It's a team, a data science agent, a domain expert agent, a health coach agent, all working together, managed by an orchestrator.

00:08:09: Like a clinical team structure, but with AI agents.

00:08:12: Exactly.

00:08:13: And apparently, this multi-agent approach worked way better than a single AI trying to juggle it all.

00:08:18: The specialization mirrors real world teamwork.

00:08:21: Interesting.

00:08:22: And staying on diagnostics, Oscar Bologna has shared results from that French Defidiog study.

00:08:28: What was the key finding

00:08:29: there?

00:08:30: You really underscored the power of whole genome sequencing, WGS, for diagnosing rare diseases.

00:08:36: The study showed a significantly higher diagnostic yield, nearly forty-two percent, compared to the usual, often slower, step-by-step diagnostic process.

00:08:45: So getting answers faster for patients with rare conditions, that's huge.

00:08:49: Definitely a capability leap.

00:08:51: Yeah.

00:08:51: But then you see the contrast with Theodora's as is research on racial disparities, specifically in key TRCA diagnosis.

00:08:58: The equity issue again.

00:08:59: Right.

00:09:00: Her work highlighted that Afro-Crubean patients often face significant delays and worse outcomes for this condition.

00:09:05: It's a stark reminder that even with better tools like WGS, we absolutely need tailored approaches and better awareness to make sure everyone benefits.

00:09:13: Technology alone isn't enough.

00:09:15: Okay.

00:09:15: Moving to our fourth theme, med tech, surgical robotics and usability.

00:09:21: Sometimes the biggest wins are the simplest, right?

00:09:24: Nishant Nair pointed out the shift to electronic instructions for use EIFUs in MedTech.

00:09:29: It sounds so mundane, doesn't it?

00:09:30: Electronic manuals.

00:09:32: But think about the logistics.

00:09:33: Those huge paper manuals, expensive, hard to update, a pain in sterile environments, EIFUs, cut costs, simplify the supply chain, and critically, they integrate into hospital digital systems.

00:09:47: So the clinician gets the right up-to-date instructions right when they need them, where they need them.

00:09:51: It's a big usability win.

00:09:53: Makes sense.

00:09:54: And on the more cutting edge side, robotics.

00:09:56: Mike Marinuro announced Medtronic's US clinical study for the Hugo robot embrace gynecology.

00:10:02: Yeah, focusing on procedures like hysterectomies.

00:10:04: The aim, as with most robotic surgery, is less invasive procedures, faster recovery times for patients.

00:10:10: And Medtronic's also pushing hard and cardiac treatments with pulse field ablation, PFA.

00:10:15: Rebecca Seidel highlighted their progress there.

00:10:17: They seem to be leading the charge.

00:10:18: First company with two FDA approved PFA systems, pulse select and AFRA.

00:10:23: Okay, quick explainer, what is PFA?

00:10:25: For those of us not in electrophysiology.

00:10:27: So traditional ablation for heart arrhythmia is often used as extreme heat or cold to destroy the tissue causing the problem.

00:10:34: PFA is different, it's non-thermal.

00:10:36: It uses very quick bursts of high voltage electricity.

00:10:39: Okay.

00:10:40: The idea is it precisely targets just the problematic heart tissue, creating tiny pores in the cells to ablate them, but minimizing damage to surrounding structures like the esophagus or nerves.

00:10:51: That's the big safety advantage.

00:10:53: And Medtronic having two approved systems used in over a hundred thousand cases already that shows how fast this tech is taking off.

00:11:00: Wow, yeah, that's rapid adoption.

00:11:01: We also saw innovation reaching into chronic conditions too.

00:11:04: Tobias Ludwig mentioned in Germany performing the first cryoblation for endometriosis using Boston Scientific Tech.

00:11:11: That's significant.

00:11:12: Endometriosis can be incredibly painful, debilitating, and often takes years to diagnose properly.

00:11:18: Using cryoblation offers a less invasive treatment option compared to traditional surgery for some patients.

00:11:23: All right, let's wrap up with our final theme.

00:11:25: Market shifts, cost, and... maybe most importantly, trust.

00:11:30: We have to start with that huge number from PwC that Steven G shared.

00:11:35: Yeah, it was eye-popping.

00:11:36: PwC projecting that by twenty thirty-five, a trillion dollars in annual U.S.

00:11:41: health care spending could shift.

00:11:43: A trillion dollars moving away from traditional models.

00:11:46: Towards what?

00:11:46: Towards digital first AI-driven models.

00:11:49: Think virtual care, AI-powered diagnostics, remote monitoring, home-based care.

00:11:54: It implies a massive reallocation of resources for incumbent players.

00:11:58: They need to shift capital away from bricks and mortar and fragmented admin and towards these new digital platforms.

00:12:04: The risk of getting left behind is very real.

00:12:06: And

00:12:07: Nisha Chellam's post really highlighted why that shift is so necessary by looking at the current payment system.

00:12:12: Her example was a simple pap smear.

00:12:14: Right.

00:12:15: She showed how, okay, the reimbursement might be a hundred eighty two hundred dollars, which sounds okay.

00:12:19: But the system fundamentally pays for the documentation clicking the boxes in the EHR, not necessarily for the doctor's time, their expertise, or that crucial personalized conversation with the patient.

00:12:29: So it rewards process over outcomes, essentially.

00:12:31: Pretty much.

00:12:33: It covers the quick, templated, fifteen-minute visit, but struggles to pay for the thirty minutes needed to really explain complex results or develop a personalized lifestyle plan.

00:12:43: That structure actively works against the proactive, personalized care everyone says they want.

00:12:48: Which brings us to the people challenge.

00:12:50: Simon Philip Rust argued AI literacy is basically the new currency in healthcare.

00:12:54: Yeah,

00:12:54: and not just for techies.

00:12:56: He stressed it's needed at all levels.

00:12:59: Executives needed to make smart investments.

00:13:01: Clinicians needed to understand the tools, trust them, know their limits.

00:13:05: Even patients need some level of understanding to engage effectively.

00:13:09: Without that broad literacy, he argued, transformation just stalls.

00:13:13: It's not a tech problem, it's a people and trust problem.

00:13:16: And finally, circling back to innovation itself, Myeongchol had that great point about the GLP-I market as epic would go via that whole class worth over a hundred billion dollars now.

00:13:25: Right, and it didn't start with an AI crunching data.

00:13:27: No, it started with human intuition, right?

00:13:29: Yeah.

00:13:29: Someone noticing something about Gila monster venom, a creative leap.

00:13:34: Exactly, a moment of serendipity, which forces that really interesting question, can AI automate that?

00:13:40: Can you code for serendipity?

00:13:42: Or will we always need that human spark, that unexpected connection, for the truly groundbreaking discoveries?

00:13:48: That really sums up the tension, doesn't it?

00:13:50: The push of amazing technology versus the essential human element.

00:13:54: So looking back at these weeks, the clear direction is towards care that's more proactive, more personalized, more integrated, using all this advanced tech.

00:14:03: For sure, the tech capability... It's largely there or rapidly getting there, but the real roadblocks right now feel softer.

00:14:11: It's about building literacy, fostering trust among clinicians and patients, especially with more autonomous systems, and crucially fixing those financial models that still reward ticking boxes over actual patient care and outcomes.

00:14:23: If you enjoyed this episode, new episodes drop every two weeks.

00:14:27: Also check out our other editions on ICT and tech insights, defense tech, cloud, digital products and services, artificial intelligence and sustainability and green ICT.

00:14:36: Thanks for diving deep with us today.

00:14:38: Make sure you subscribe so you don't miss our next look at the evolving health tech landscape.