In London, a question as old as medicine itself — how do we extend the reach of expert care to all who need it — has found a partial answer in an unexpected collaborator: an autonomous artificial intelligence. Over sixteen months and more than eight thousand patients, a system called AIaMD demonstrated that machines can responsibly triage suspected skin cancers, not by replacing the dermatologist's judgment, but by reserving it for the moments when it is most irreplaceable. The study, presented in Vienna, suggests that the future of specialist medicine may lie not in more specialists, but in w
AI System Expands Dermatology Capacity by 62%, Safely Screening Low-Risk Skin Lesions
Every hour saved reviewing low-risk lesions can be reinvested in patients with skin cancer
Mark
So the AI discharged a quarter to a third of patients without any doctor seeing them at all. How confident are we that it didn't miss anything dangerous?
Mimi
The sensitivity for melanoma and other serious skin cancers was 98 percent across the national dataset. That's very high. But there were six false negatives—cases the AI missed—that came to light later through surveillance.
Luke
Six cases out of how many total? The study says 8,391 patients went through the pathway, but I need to know what denominator we're using for that 98 percent figure. Is it 98 percent of the cases the AI actually reviewed, or 98 percent of all skin cancers in the broader population?
Mimi
That's a fair question. The 98 percent sensitivity comes from a national dataset that included both study sites, so it's a broader measure. The six false negatives were identified through post-market surveillance—meaning they were caught later, not during the study itself.
Mark
And none of those six patients had bad outcomes?
Mimi
Correct. No adverse outcomes were identified in the available follow-up period. But that's an important qualifier—the follow-up may not have been indefinite.
Luke
Right. We don't know how long they were monitored. A melanoma in situ that's missed could become invasive later. The study shows the system works in this specific UK context with these two hospitals. Replicating it elsewhere is a different question.
Mark
What about the patients themselves? Did they understand they were being screened by AI?
Mimi
Eighty-six percent consented to autonomous decision-making, so there was informed consent. Thomas also said patients need to know what warning signs to watch for—the AI isn't the end of the story.
Luke
That consent rate is interesting. Fourteen percent of patients declined. We don't know why, and we don't know if there were differences in outcomes between the two groups.
Mark
The time savings seem real though. Nearly 2,900 hours freed up.
Mimi
Yes, and that's based on actual clinical data from 16 months of operation. The math is straightforward: if you're not seeing those patients face-to-face, you're not spending that time. The question is what happens when you scale it up.
Luke
And whether other hospitals can replicate the same safety profile. This was a controlled study in two sites with presumably good infrastructure and training. Real-world deployment is messier.
Der Puls
- Dermatology services across the UK face a quiet crisis — demand for skin cancer screening far outpaces the number of specialists available to meet it.
- An autonomous AI system processed over 8,500 urgent referrals across two London hospitals, independently clearing roughly one in four patients without any clinician ever reviewing their case.
- Six patients were initially misclassified as low-risk, a sobering reminder that no system — human or machine — operates without error, and that continuous post-market surveillance is not optional but essential.
- The pathway cut routine follow-up appointments nearly in half and reduced biopsies from 43 to 27 percent, redirecting thousands of clinician hours toward complex cancers and severe inflammatory disease.
- With 86 percent of patients consenting to autonomous decision-making, the study signals that public trust in AI-driven care may be more attainable than many assumed — if transparency and safety monitoring remain central.
In London, a question as old as medicine itself — how do we extend the reach of expert care to all who need it — has found a partial answer in an unexpected collaborator: an autonomous artificial intelligence. Over sixteen months and more than eight thousand patients, a system called AIaMD demonstrated that machines can responsibly triage suspected skin cancers, not by replacing the dermatologist's judgment, but by reserving it for the moments when it is most irreplaceable. The study, presented in Vienna, suggests that the future of specialist medicine may lie not in more specialists, but in wiser stewardship of the ones we have.
A team of London dermatologists has shown that an autonomous AI system can safely shoulder a meaningful share of skin cancer screening, preserving specialist time for the cases that need it most. Over sixteen months, the system — known as AIaMD — evaluated more than 8,500 patients referred urgently to two UK hospitals, using clinical and dermoscopic smartphone images to distinguish benign lesions from those requiring human review.
At each site, between a quarter and a third of patients were discharged by the AI alone, with no clinician involvement. Teledermatologists then cleared a further portion of remaining cases. The combined effect was striking: routine follow-up appointments fell from 27 percent to 12 percent, and biopsy rates dropped from 43 percent to 27 percent compared with traditional face-to-face care. Across the study period, the pathway freed an estimated 2,851 clinician hours — the equivalent of more than 8,500 additional appointments — representing a 62 percent expansion in capacity without a single new hire.
Safety was treated as a continuous obligation rather than a baseline assumption. The system achieved 98 percent sensitivity for invasive melanoma, squamous cell carcinoma, and basal cell carcinoma. Post-market surveillance identified six false-negative cases — five basal cell carcinomas and one melanoma in situ — none of which led to adverse outcomes within the follow-up window. Lead researcher Lucy Thomas stressed that safe AI deployment demands ongoing monitoring and patient education, not a single moment of validation.
Eighty-six percent of enrolled patients consented to autonomous decision-making, suggesting the public may be readier for this shift than expected. The researchers are careful to note that these results emerge from a specific NHS context, and broader replication will be necessary. But the underlying ambition is clear: not to replace dermatologists, but to ensure their expertise reaches the patients who need it most urgently.
A team of dermatologists in London has demonstrated that an autonomous artificial intelligence system can safely screen thousands of patients with suspected skin cancer, freeing up specialist time for more complex cases. The system, called AIaMD, processed over 8,500 patients across two UK hospitals during a 16-month study period, identifying which lesions were benign and which warranted closer examination by a human clinician.
The work was presented at the 2026 European Academy of Dermatology and Venereology Congress in Vienna. The researchers found that when patients with urgent suspected skin cancer referrals arrived at the two hospitals, the AI could evaluate their lesions using clinical and dermoscopic smartphone images. For roughly one in four to one in three patients at each site, the system confidently identified benign lesions and discharged them without any clinician review. Teledermatologists then reviewed the remaining cases, with another quarter of patients cleared through that secondary assessment. The net effect was that routine follow-up appointments dropped from 27 percent to 12 percent, and biopsies fell from 43 percent in traditional face-to-face care to 27 percent in the AI-supported pathway.
The capacity gains were substantial. The autonomous pathway saved an estimated 2,851 hours of clinician time over the study period—equivalent to more than 8,500 additional face-to-face appointments, assuming 20-minute consultations. That represents a 62 percent expansion in clinical capacity without hiring additional staff. Lucy Thomas, a consultant dermatologist at Chelsea & Westminster Hospital NHS Foundation Trust and one of the study's lead researchers, noted that this freed-up time could be redirected toward patients with confirmed skin cancer who need timely treatment, and toward those with severe inflammatory skin conditions where early specialist access can meaningfully improve outcomes.
Safety was woven into the study design from the start. The AI system achieved 98 percent sensitivity for invasive melanoma, squamous cell carcinoma, and basal cell carcinoma across national datasets, meaning it correctly identified nearly all cases of these serious conditions. Specificity—the ability to correctly identify benign lesions—was 72.1 percent. Six false-negative cases were identified through post-market surveillance: five basal cell carcinomas and one melanoma in situ. None resulted in adverse patient outcomes within the available follow-up period. Thomas emphasized that deploying an AI system safely is not a one-time event but an ongoing process requiring continuous monitoring, learning from errors, and ensuring patients understand what warning signs to watch for.
The study enrolled 8,391 patients, representing 94 percent of urgent suspected skin cancer referrals across the two hospitals. Eighty-six percent of patients consented to autonomous decision-making, suggesting broad acceptance of the approach. The researchers acknowledged that these results come from a specific UK healthcare context and that replication across larger populations and different settings will be necessary before the system becomes standard practice. But if the findings hold, Thomas suggested, autonomous AI could reshape dermatology by allowing scarce specialist expertise to concentrate on the cases where it matters most—not by replacing dermatologists, but by redirecting their work toward patients who need it most urgently.
Bemerkenswerte Zitate
Deploying an AI system safely isn't a one-off exercise. You need to keep monitoring it, understand when things go wrong, learn from those cases and make sure patients themselves know what to look out for.— Lucy Thomas, Consultant Dermatologist, Chelsea & Westminster Hospital NHS Foundation Trust
If these findings are replicated across larger populations and different healthcare settings, autonomous AI could become an important part of creating a more sustainable dermatology service—not by replacing dermatologists, but by allowing scarce specialist expertise to be focused where it can make the greatest difference to patients' lives.— Lucy Thomas
Möchten Sie die ganze Geschichte?
Das Original lesen bei thedermdigest.com ↗