Turnaround time — the interval from study acquisition to signed report — is one of the most watched metrics in radiology department operations. It affects referring physician satisfaction, emergency department throughput, and increasingly, payor contract performance. When a radiology administrator sees TAT trending upward, the diagnosis is usually the same: "we need more radiologist capacity." More often than not, that diagnosis is incomplete.
Hiring more radiologists addresses one specific bottleneck in the TAT chain. But the TAT chain has three segments, and only one of them is directly about radiologist reading speed. The others sit on either side of the read itself and are invisible in most departmental dashboards because they get measured as total TAT rather than as component intervals. You cannot fix what you cannot see.
Bottleneck One: Pre-Read Queue Management
The first segment is the interval between study acquisition and study assignment — the time a completed scan sits in the PACS worklist before any radiologist opens it. In departments with manual worklist management or priority assignment, this interval is longer and more variable than anyone wants to acknowledge.
Studies in most PACS systems arrive in a single queue ordered by acquisition time. Unless the department has configured worklist rules to segregate by priority, modality, or body part, a radiologist pulling from the top of the queue will pull the oldest study regardless of clinical urgency. A CT chest ordered STAT for a pulmonary embolism workup may be sitting behind 15 outpatient studies ordered routine two hours earlier. The TAT clock on the STAT study is running; the study is not being touched.
Priority-based worklist configuration is the obvious fix, and most radiology IT teams have done it. But the configuration is only as good as the priority data feeding it — and referring physician priority assignment is notoriously inconsistent. In a busy ED or urgent care setting, "STAT" becomes a reflexive label applied to anything that the ordering clinician wants quickly, which is often everything. When 40% of studies are marked STAT, the effective priority queue collapses back into a time-ordered queue.
The more durable fix is modality-specific worklist segmentation with read time targets by category. A chest CT ordered for a known lung nodule follow-up in a stable outpatient does not need the same TAT as a chest CT ordered for a trauma patient. Separating these into distinct worklists and assigning radiologist coverage accordingly — rather than treating all chest CTs as interchangeable units of read work — is where significant TAT reduction in the pre-read segment comes from.
Bottleneck Two: Within-Read Mechanical Work
The second segment is what happens inside the read itself — and this is the segment most relevant to Neurmorph's product. Even after a study has been correctly triaged and assigned to the right radiologist, the clock continues running while the radiologist completes the mechanical scaffolding of the read before any diagnostic work begins.
In chest CT, the sequence we see most often looks like this: the radiologist opens the study, scrolls through the axial stack to locate candidates, places measurement calipers, manually pulls the prior series from the PACS archive (which may require a separate lookup step if priors are stored in a legacy archive), registers the prior visually against the current series, calculates growth rate manually or by eye, assigns a Lung-RADS category, and then begins dictating or typing the report. All of that work — localization through category assignment — precedes the point where any clinical judgment is exercised about what the finding means and what follow-up is indicated.
For a routine nodule follow-up CT with one or two stable nodules and an available prior, this mechanical work takes 3–5 minutes. For a study with three or four nodules at different sizes requiring individual Fleischner assessments, no prior available, and ground-glass opacity that complicates density measurement, it can take 8–12 minutes before the report dictation begins. Multiplied across 60 chest CTs in a morning session, the range between a mechanically-efficient read environment and an inefficient one can be 2–4 hours of total session time.
The intervention here is pre-annotation — having the finding already located, measured, categorized, and compared to prior when the radiologist opens the study. We are not describing Neurmorph's role as "the radiologist doesn't need to check the annotation." The radiologist checks every annotation. The difference is that verification is faster than initial annotation, and the cognitive entry point into each case is confirmatory rather than investigative.
What pre-annotation does not address within TAT: complex cases that require extended diagnostic deliberation, multi-system reads, or cases where the annotation raises a finding the radiologist needs to evaluate more carefully. For those cases, TAT should be longer — the goal is not to compress every read to the same duration, but to eliminate the time that isn't about clinical complexity.
Bottleneck Three: Post-Read Report Finalization
The third segment is the interval between read completion and report signature. This one is underappreciated as a TAT driver and tends to be invisible in TAT tracking because most TAT metrics measure time-to-signature, which means everything between acquisition and signature is aggregated. The pre-signature delay is embedded in the overall number.
In practice, multiple things delay report signature after the radiologist has made their clinical determination:
Structured report template completion. Many hospitals use structured reporting templates for specific study types — chest CT, brain MRI, and so on. If the template requires the radiologist to fill in discrete fields for each finding category before the report can be finalized, and if those fields weren't pre-populated during the read, completing the template is a post-read task that holds up signature. A radiologist who has dictated a narrative report and then needs to go back and fill in 12 structured data fields before clicking finalize will delay signature by 2–4 minutes per study. Across 40 studies, that adds up to 80–160 minutes of session time that is not clinical work.
Transcription and addendum cycles. In departments using voice recognition with post-processing transcription review, a report may require a review and correction pass before it can be signed. If the transcription review is handled by the radiologist rather than a transcriptionist, and the review queue grows during busy sessions, reports can sit unsigned for 20–45 minutes after the radiologist has completed their verbal dictation.
Peer review and quality check queues. Some departments require a second-read or quality check on a percentage of studies as part of their quality assurance program. If the quality check queue is managed manually and the reviewing radiologist's capacity is constrained, studies can sit pending review for hours. Automated peer sampling tools that queue randomly selected studies for asynchronous review reduce this latency — but only if the second-read reviewer has capacity, which brings the problem back to the overall staffing question.
Why All Three Bottlenecks Need Individual Attention
The reason TAT improvement initiatives often underperform their projections is that they target one segment of the chain in isolation. Hiring a teleradiology group to cover overnight volume addresses segment one (queue assignment) but does nothing about segments two and three. Adding a structured reporting tool addresses part of segment three but may actually slow segment two if the template is cumbersome to use mid-read.
Mapping your department's TAT by segment — pre-read queue hold time, read session time per study type, and post-read signature delay — is the prerequisite to an effective intervention. Most PACS systems can generate these intervals if event timestamps are captured correctly: study arrive time, first open time, dictation start time, dictation end time, signature time. If those timestamps are not being captured or not being reported separately, the first step is a data audit, not a staffing hire.
We are not saying staffing capacity is irrelevant to TAT — it clearly is not. We are saying that in our experience working with imaging centers on pre-annotation deployment, the departments that see the largest TAT improvement are the ones that treat segment two as a distinct problem with a distinct solution, rather than assuming that all TAT problems reduce to "not enough radiologists."