What If AI Could See Cancer Before We Can?
What if a CT scan that appeared completely normal was already holding a warning about pancreatic cancer?
Today’s tech. Five human questions. One healthier future.
By T. Robinson | TT’s Tech Care
When a Normal Scan Might Have More to Say ❤️
Now, THIS is one of those healthcare technology stories that immediately caught my attention!
And if you've been following TT's Tech Care for a while, you already know how excited I get when I discover research that has the potential to make a meaningful difference in people's lives!
But this particular discovery made me think about something beyond artificial intelligence and sophisticated technology.
It made me think about time, and how much of a difference having more of it could make when someone is facing a serious medical diagnosis.
Time to ask questions, explore treatment options, understand what's happening, and make important decisions with the people they love.
After spending more than a decade working in healthcare, across hospitals, clinics, and other clinical settings, I've developed a deep appreciation for everything that happens between a patient's first concern and the moment they finally receive answers.
And believe me, that journey isn't always straightforward!
There are appointments to schedule, diagnostic examinations to coordinate, results to communicate, referrals to process, and patients trying to navigate a healthcare system that can feel overwhelming even on its best days.
Now imagine discovering that an abdominal CT scan taken months or even years earlier might have contained subtle warning signs of pancreatic cancer.
Signs that weren't visible to the radiologist at the time, but that an artificial intelligence model could potentially recognize.
That's exactly the possibility Mayo Clinic researchers are investigating, and the implications are fascinating!
When I read about their findings, my first thought wasn't about how sophisticated the algorithm must be.
Although, let's be clear, the technology is pretty remarkable! 😂
I immediately thought about the patients and families whose lives could eventually be affected by an earlier opportunity to investigate such a devastating disease.
Because behind every medical image is a person waiting for answers, and behind that person is a life that matters. ❤️
🔬 Mayo Clinic's Research: Finding What We Cannot Yet See
In April 2026, researchers published findings in the medical journal Gut describing an artificial intelligence system called the Radiomics-based Early Detection Model, or REDMOD.
The model was developed to analyze abdominal CT scans and identify subtle pancreatic tissue patterns associated with pancreatic cancer before a tumor becomes visibly apparent.
And here's where the research gets especially interesting!
In an independent test involving 493 CT scans, including 63 scans from patients who were later diagnosed with pancreatic cancer, REDMOD correctly identified approximately 73% of those prediagnostic cases.
Expert radiologists reviewing the scans without AI assistance identified approximately 38.9%.
Even more compelling, the model recognized cancer-associated patterns at a median of 475 days before diagnosis, which is approximately 16 months!
Some of the scans had been obtained nearly three years before the eventual diagnosis.
Now, I don't know about you, but those findings certainly got my attention!
Researchers also evaluated the model across different institutions and imaging datasets, an important consideration because healthcare technology needs to perform reliably beyond the environment where it was originally developed.
However, before we start imagining AI becoming the world's newest superhero in a white coat, there's something important we need to discuss! 😂
REDMOD is still an investigational research tool.
It is not FDA approved for routine clinical use or population screening, and these retrospective findings do not establish that earlier identification improves survival.
They also don't mean that every suspicious AI result represents cancer.
Mayo Clinic researchers are continuing their work through a prospective study called AI-PACED, which is designed to evaluate how this technology might function in clinical practice among patients at elevated risk.
And that's a critical next step!
Because identifying a potential warning sign is one thing, but understanding how to act on that information safely, responsibly, and effectively is another.
Which brings me to five questions I believe deserve our attention!
💻❤️ TT'S TOP 5: Five Human Questions
1️⃣ What makes this discovery so important?
For me, the most exciting part of this research is the possibility that valuable information we previously couldn't recognize may already exist within medical images!
Think about that for a moment.
A patient undergoes an abdominal CT scan for a clinical concern, and a radiologist carefully reviews the images, looking for abnormalities that can be identified and interpreted.
The findings are documented and communicated to the referring provider, helping guide the next steps in the patient's care.
That's an essential part of modern healthcare, but what happens when a disease has begun causing subtle changes that are too small or complex to recognize through conventional visual interpretation?
That's where technology like REDMOD could eventually provide additional insight!
Using a technique called radiomics, the model examines quantitative patterns within pancreatic tissue rather than simply searching for a recognizable tumor.
I find that fascinating, especially because it raises the possibility of learning more from imaging examinations that patients have already undergone.
Having worked in healthcare environments where diagnostic information plays such an important role, I appreciate how valuable that additional insight could become.
But here's where I want us to be thoughtful.
An AI-generated risk signal is not a cancer diagnosis, and I would never want patients or healthcare professionals treating it as one.
I want the radiologist and treating clinician to carefully evaluate any concerning result alongside the patient's medical history, other clinical findings, and established standards of care.
Because the goal shouldn't be to generate more alerts simply because we have the technology to do so!
The goal should be to provide meaningful information that supports clinical judgment and helps healthcare professionals make better-informed decisions.
And if we can eventually give clinicians an earlier opportunity to recognize a serious concern?
Now THAT is the kind of advancement I want to see! ❤️
2️⃣ Could detecting pancreatic cancer earlier actually save lives?
This question is the one that really pulls at my heart.
Pancreatic cancer is particularly challenging because symptoms often don't appear until the disease has progressed, and for many patients, diagnosis happens when treatment options may already be limited.
So naturally, the possibility of identifying warning signs earlier is incredibly encouraging!
But I want us to be careful about what the research actually tells us.
REDMOD identified cancer-associated imaging patterns a median of approximately 16 months before diagnosis.
That doesn't mean researchers demonstrated that patients lived 16 months longer.
Those are two completely different findings, and in healthcare, that distinction matters tremendously!
We still need evidence showing that earlier alerts lead to timely evaluation, earlier-stage diagnoses, appropriate treatment, and ultimately better patient outcomes.
And I genuinely hope that future research brings us closer to those answers!
Imagine what an earlier opportunity for evaluation might eventually mean.
A patient could potentially have more treatment options to discuss, a family might have additional time to understand a diagnosis and plan for what comes next, and a clinical team could have an earlier opportunity to coordinate specialized care.
Those possibilities are meaningful, especially when we're talking about a disease that can be so difficult to identify early.
But until clinical studies demonstrate improved outcomes, they remain possibilities rather than established benefits.
I'm hopeful about where this research could lead, but I'm equally committed to keeping that hope grounded in evidence.
Because patients deserve both!
They deserve innovation that pushes healthcare forward, and they deserve honest information about what that innovation has actually accomplished.
3️⃣ Does this mean AI is becoming better than radiologists?
Now, you already know my answer to this one!
Absolutely not! And we're not handing over anybody's white coat just yet! 😂
But let's talk about why.
REDMOD performed better than the expert radiologists on the specific research task of identifying subtle prediagnostic pancreatic cancer patterns.
That's a significant result, and I don't want to minimize it!
However, it doesn't mean the model can perform everything a radiologist does.
Radiologists evaluate medical images in the context of anatomy, clinical history, prior examinations, and a wide range of possible conditions.
They communicate important findings, consider uncertainty, and collaborate with other healthcare professionals to help guide patient care.
A model designed to recognize one particular disease-associated pattern simply isn't a substitute for all that expertise.
And having worked in clinical environments, I know healthcare depends on much more than a single result.
It depends on communication, coordination, professional judgment, and teams of people working together to make sure important information reaches the right place.
I don't want healthcare technology creating an environment where clinicians feel replaced or disconnected from their patients.
I want it helping them do what they do best!
Imagine a future where a validated AI tool flags a subtle pattern within a CT scan, prompting a radiologist to take a closer look.
The radiologist considers that information alongside the complete examination, while the treating provider reviews the clinical context and determines whether further evaluation is appropriate.
THAT is the kind of collaboration I can get behind!
If AI helps clinicians identify something they might otherwise miss, that could be incredibly valuable.
But medical decisions still require qualified professionals, appropriate evidence, and accountability.
I want technology strengthening clinical expertise, not pretending to replace the people who provide it.
After all, an algorithm might recognize a pattern, but it cannot replace the compassion of a healthcare professional sitting with a patient and carefully explaining a difficult finding.
And that human connection deserves to be protected! ❤️
4️⃣ What concerns should we consider before introducing this technology into healthcare?
Okay, now we're getting to the part of the conversation that might not sound quite as exciting as the headline, but could be just as important!
Because whenever I see an impressive healthcare AI result, I have questions.
And I mean QUESTIONS! 😂
First, what happens when the model gets something wrong?
In the independent test set, REDMOD demonstrated approximately 81.1% specificity.
That means some scans from patients without pancreatic cancer were incorrectly flagged as concerning.
Now imagine being told that an AI system identified a possible cancer-related warning in your imaging.
Even if the result isn't a diagnosis, that information could create tremendous anxiety.
A patient might undergo additional imaging, consultations, or procedures that ultimately prove unnecessary, and those appointments can come with financial costs, transportation challenges, missed work, and emotional stress.
There's also the opposite concern.
What happens when the model fails to identify a warning sign?
Could patients or clinicians develop too much confidence in a negative result?
Could the technology perform differently across patient populations, healthcare institutions, or imaging protocols?
These are questions I want researchers and healthcare organizations to take seriously.
As someone studying cybersecurity and information assurance, I also think about patient privacy, secure handling of medical images, system integrity, and responsible access to sensitive information.
Clinical accuracy is essential, but so is protecting the trust patients place in healthcare organizations!
Before any tool like REDMOD becomes part of routine care, I want rigorous clinical validation, clear protocols for managing alerts, strong security safeguards, and meaningful physician oversight.
And perhaps most importantly, I want patients to understand what the technology can and cannot tell them.
A concerning AI result should begin a thoughtful clinical conversation, not leave a patient frightened and searching the internet for answers at two in the morning!
Healthcare technology should help reduce uncertainty when possible, not create confusion without support.
Because behind every false positive is a real person who may spend days or weeks worrying, and behind every missed warning is someone who deserved the strongest possible standard of care.
That's why responsible implementation matters just as much as innovation itself! ❤️
5️⃣ Who should benefit from this innovation?
Everyone!
And sis, when I say everyone, I mean EVERYONE! ❤️
This is where my passion for rural healthcare and vulnerable populations comes right back into the conversation.
One of my biggest concerns about emerging healthcare technology is that advancements often become available in well-funded medical institutions long before reaching communities that already struggle with access.
Consider someone living in a rural area whose closest major medical center may be hours away.
Specialty appointments might involve significant travel, transportation could be unreliable, insurance coverage might be complicated, and missing work for medical appointments could create financial hardship.
Now imagine that a future, clinically validated AI system identifies a concerning pancreatic cancer risk signal in an existing CT scan.
That might sound like progress, and potentially it could be!
But what happens next?
Can the patient reach a specialist, afford the recommended follow-up, and receive timely evaluation?
Will someone help coordinate their care and explain the findings in a way that is understandable and compassionate?
Because here's something I feel strongly about.
An earlier warning is only meaningful if patients have a realistic opportunity to receive the care that follows!
We cannot simply introduce sophisticated AI tools and assume access problems will magically disappear.
Wouldn't that be convenient, though? 😂
Unfortunately, healthcare doesn't work that way!
I want healthcare organizations to consider the entire patient journey, including patient navigation, specialty referrals, affordable services, communication, accessibility, and timely follow-up.
Not just the scan, the AI analysis, or the impressive research statistics.
I also want rural and underserved healthcare organizations represented as these technologies are evaluated and implemented.
A patient's ZIP code should not determine whether they benefit from meaningful medical innovation!
For me, healthcare equity isn't simply about making the newest technology available.
It's about ensuring people have a fair opportunity to benefit from it, regardless of where they live or what resources they have.
And that's the kind of healthcare future I'm advocating for! ❤️
💬 TT'S TAKE: The Real Value of Seeing Earlier
Here's what I keep coming back to.
Medical technology has advanced tremendously since I first began exploring the healthcare profession, and I genuinely love seeing where innovation is taking us!
But even with all the changes in healthcare, one thing remains constant.
Patients are still people!
They're people with jobs, families, responsibilities, questions, worries, and hopes for their futures.
Some are waiting for answers, others are trying to understand unexpected results, and many are navigating a complicated healthcare system while carrying concerns they may not even know how to express.
Having spent years working in patient-facing clinical environments, I know how important it is for healthcare professionals to recognize the human being behind every appointment, examination, and result.
That's why this Mayo Clinic research resonates with me so deeply.
Not simply because an AI model recognized subtle patterns that radiologists couldn't visually identify, but because of what that capability could eventually mean if future studies establish that it helps patients.
Imagine a healthcare system where validated technology helps clinicians recognize meaningful risks sooner, patients receive clear explanations rather than confusing alerts, and follow-up care is coordinated thoughtfully.
Imagine those same advancements reaching people living outside major metropolitan areas, including rural communities that have historically faced barriers to specialized care.
That's the future I want to see!
Now, I'm certainly not suggesting that AI will eliminate cancer, solve every diagnostic challenge, or suddenly make our healthcare system perfect.
If only healthcare problems came with a convenient software update! 😂
But I do believe responsible research like this deserves our attention, because every meaningful advancement begins with asking whether we can do something better.
The next challenge is determining whether we can do it safely, equitably, and in a way that genuinely improves people's lives.
For REDMOD, that means continuing prospective research, evaluating patient outcomes, understanding false positives, and determining how the technology could responsibly fit into clinical workflows.
The results so far are promising, and I'm excited to see where the research goes!
But the work isn't finished, and that's perfectly okay.
We don't need to exaggerate a discovery to appreciate its importance.
The real achievement won't simply be teaching AI to see earlier. It will be demonstrating that earlier insight helps clinicians care better!
Because innovation becomes meaningful when technology creates more room for humanity.
And if research like this eventually gives patients and healthcare professionals more opportunities to act before a serious disease progresses, that would be an extraordinary step forward!
Not because technology replaced the human element, but because it helped strengthen it.
What good is advancing healthcare if we forget the people we're advancing it for? ❤️
🎤 Now I Want to Hear From You!
If a clinically validated AI system could help your healthcare team recognize possible cancer warning signs before symptoms appeared, would you want that information?
And what would matter most to you?
Accuracy and reliability, a healthcare professional helping you understand the results, access to affordable and timely follow-up care, or all three?
I would love to hear from patients, healthcare professionals, caregivers, researchers, and technology developers!
Because these conversations shouldn't happen only in research laboratories or technology conferences.
The people whose lives could be affected deserve to be part of them!
The future of healthcare belongs to all of us. Let's make sure we're building it with humanity at the center! ❤️
📚 Research and Sources
Mayo Clinic News Network. (2026, April 29). Mayo Clinic AI helps specialists detect pancreatic cancer up to 3 years before diagnosis in landmark validation study.
Mukherjee, S., et al. (2026). Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability. Gut.
https://doi.org/10.1136/gutjnl-2025-337266
Mayo Clinic. (2026, June 25). AI identifies pancreatic cancer years before diagnosis, offering potential for earlier treatment.
Editorial note: REDMOD remains investigational. Its reported performance comes from retrospective testing, and earlier detection has not yet been demonstrated to improve patient survival or other clinical outcomes.