If you've been a patient at a New England hospital in the last couple of years, there's a decent chance AI played some role in your care — and you probably had no idea. That's kind of the point. The best implementations aren't flashy. They're just... working, quietly, in the background.
New England has always punched above its weight in healthcare innovation. We've got Mass General, Dartmouth Health, Yale New Haven, and a dozen other world-class systems packed into a relatively small geographic footprint. So it shouldn't surprise anyone that this region is one of the more interesting places to watch when it comes to AI in medicine.
What's Actually Being Deployed (Not Just Piloted)
Let's skip past the hype and talk about what's genuinely in production. Radiology is probably the furthest along. AI tools from companies like Aidoc and Nuance are being used at several Boston-area hospitals to flag critical findings in CT scans — things like pulmonary embolisms or intracranial hemorrhages that need immediate attention. The AI doesn't replace the radiologist. It just makes sure the urgent stuff bubbles to the top of the queue faster.
Dartmouth Health, which serves a lot of rural New Hampshire and Vermont, has been particularly interesting to watch. Rural healthcare has this brutal problem where specialist access is limited, so getting a second opinion or a specialist read on an imaging study can take days. AI-assisted triage helps compress that timeline significantly. For someone in a small town two hours from the nearest academic medical center, that matters a lot.
Ambulatory care is another area seeing real movement. Epic, which powers the EHR systems at most major New England health systems, has been rolling out predictive models built directly into clinical workflows. These models flag patients who are at elevated risk for things like sepsis, hospital readmission, or deterioration overnight. Nurses get an alert. They go check on the patient. Sometimes it's nothing. Sometimes it genuinely catches something early.
The Mental Health Angle — And Why It's Complicated
One of the more unexpected applications getting attention in this region is AI for behavioral health. New Hampshire, as most people here know, has been dealing with a mental health crisis for years — not enough providers, long wait times, people falling through the cracks. Some health systems are experimenting with AI-powered screening tools that can analyze patient responses during intake to flag depression, anxiety, or suicide risk more consistently than a rushed 10-minute check-in allows.
Honestly, this is where I think the ethical questions get really interesting. An algorithm flagging someone as high-risk for self-harm is a very different kind of decision than one flagging an abnormal chest X-ray. The stakes around bias, false positives, and what happens after the flag are enormous. Who follows up? What if the model is systematically less accurate for certain demographic groups? These aren't hypothetical concerns — they're active conversations happening in hospitals right now.
I'm not saying don't do it. I'm saying the implementation needs to be done with a lot of humility and a lot of human oversight baked in from the start.
Natural Language Processing Is Doing Heavy Lifting
A huge amount of clinical information lives in unstructured text — doctor's notes, discharge summaries, phone call transcripts. NLP tools are getting genuinely good at pulling structured meaning out of that mess. Mass General Brigham has invested heavily in NLP for clinical documentation and research, and it's starting to show up in practical ways.
Ambient clinical documentation is one application that clinicians seem to actually like, which is rare. Tools like Nuance's DAX use AI to listen to a patient-provider conversation and automatically generate a draft clinical note. Doctors spend an absurd percentage of their time on documentation — some studies put it at over 40% of their working hours. If AI can claw some of that back, that's more time with patients. That's a genuine win.
The Data Privacy Reality Check
Here's something worth being honest about: the reason New England health systems can build and train these models is because they have access to enormous amounts of patient data. That data is protected by HIPAA, de-identified for research purposes, and governed by institutional review boards. But the public trust piece is fragile. Patients aren't always aware of how their data is being used, even in aggregate.
This is an area where I think the healthcare AI community needs to be more proactive about transparency. Not because the current practices are necessarily wrong, but because the moment something goes sideways — a breach, a biased model causing harm, a vendor misusing data — the backlash could set the whole field back years. Building public trust now is worth the investment.
What's Coming in the Next Few Years
A few things seem pretty likely based on where the research and investment is flowing. Multimodal AI — models that can reason across imaging, lab values, clinical notes, and genomics simultaneously — is going to become more clinically relevant. Right now most tools are narrow. They do one thing well. The next generation will be able to synthesize across data types in ways that start to look a lot more like how an experienced clinician thinks.
Federated learning is also getting more traction. The idea is that health systems can collaboratively train AI models without actually sharing raw patient data — the model learns from data at each institution without that data ever leaving the building. For a region like New England where you've got multiple competing health systems who'd never share data directly, federated approaches could unlock training datasets that just weren't possible before.
And honestly? AI scribes and ambient documentation are going to become standard. It's just too useful and clinicians are too burned out for this not to spread quickly.
What This Means for Our Community
For those of us in the NH AI Meetup community, healthcare is one of the most consequential domains where our skills and interests intersect with real-world impact. Whether you're a developer, a data scientist, or just someone who thinks hard about AI ethics, there's meaningful work to be done here.
If you're curious about getting involved, Dartmouth Health has research collaborations, and there are a handful of health-tech startups in Manchester and Portsmouth working on exactly these problems. The problems are hard, the stakes are high, and the people doing this work are genuinely trying to get it right. That's a pretty good combination.