About Us
Pacslens Health builds AI triage software for community radiology departments. Our focus is narrow: get the right study in front of the right radiologist before the time window for effective intervention closes.
Overnight and weekend radiology coverage gaps mean critical findings on chest CTs and head CTs sit unread for 4-10 hours before a radiologist flags them for the ordering physician. Community hospitals cannot staff subspecialty radiology coverage around the clock, and the failure mode is not staffing.
The worklist has no intelligence. That was Nikhil Kulkarni’s conclusion after years of overnight radiology call at a 280-bed community hospital in central Massachusetts. Studies arrived chronologically. Critical findings did not announce themselves. A head CT flagged for possible subdural hematoma sat in queue for seven hours one night because the teleradiology service was overloaded — by the time the read came back, the patient was in the ICU. The staffing was not the problem. The prioritization layer between the scanner and the radiologist’s eyes was missing entirely.
Make sure no critical radiology finding waits unread at a community hospital.
Nikhil Kulkarni was on call as a radiology resident at a 280-bed community hospital in central Massachusetts when a head CT flagged for possible subdural hematoma sat unread for seven hours because the overnight teleradiology queue was overloaded. By the time the read came back, the patient had been transferred to the ICU.
Nikhil and co-founder Ananya Krishnan built a single-model triage classifier integrated with the hospital PACS via HL7 feed. They deployed it as a pilot at two community hospitals in greater Boston. Critical-finding notification delay dropped from an average of 8 hours to under 2 hours. Both radiology chiefs asked for permanent deployment. That result — reproducible, measurable, requested by the people doing overnight call — became the foundation of Pacslens.
Pacslens Health focuses entirely on the triage-plus-worklist-intelligence layer for community and regional hospital radiology departments with 2-8 radiologist FTEs handling 150-600 daily imaging studies. We deliberately avoid the subspecialty AI diagnostic market in favor of the operational workflow problem community radiologists face every shift.
Every feature starts with a question: does this make the radiologist’s reading session faster or slower? AI annotations are delivered inside the existing PACS viewer. Pre-read summaries are collapsible and dismissible. Nothing requires a new login, a secondary monitor, or a workflow detour.
On critical findings, a missed hemorrhage or pneumothorax is worse than a false positive that gets resolved in 30 seconds. Pacslens is calibrated to minimize false negatives on the three primary critical-finding categories. We document the tradeoffs and publish the sensitivity and specificity numbers rather than hiding them.
Every triage classification comes with a probability score and the imaging region that triggered it. Radiologists can see what the model flagged and why. They can accept, override, or dismiss any annotation. No finding is acted on without a radiologist in the loop.
Large academic medical centers have 24/7 subspecialty coverage, fellows reading overnight, and teleradiology contracts for backup. Community hospitals do not. Pacslens is scoped to the 2–8 radiologist department handling 150–600 daily studies — not the 40-radiologist quaternary center that does not need us.
Every triage event, worklist promotion, and critical-finding alert is logged with a UTC timestamp, model version, and operator action. The compliance dashboard generates ACR Critical Results Communication reports on demand. An audit trail is not an add-on — it ships with every deployment.
If your radiology department handles after-hours coverage without a triage layer between the scanner and the worklist, we’d like to show you what changes when one is in place.