Part 2 of 6 in Circle CVI's Structural Heart series. Every Scan. One Plan. One Platform.

A diagnostic cardiac CT report ends in an impression. A structural planning report ends in a set of numbers that a heart team will use to select a device.

That is a different product. The impression tolerates hedging, because the clinician reading it brings their own judgment to the question. The planning report does not, because a millimeter of annulus perimeter is not an opinion to be weighed. It is an input to a sizing chart.

Imaging physicians have absorbed this shift largely without anyone naming it.

What the heart team is actually asking for

The societies have been specific about the deliverable. The SCCT expert consensus on CT imaging for TAVR sets out five domains: acquisition protocol, aortic valve sizing and reporting including annulus and LVOT measurement, fluoroscopic angulation reporting, and assessment of vascular access, coronary anatomy and incidental findings. The European Society of Cardiovascular Radiology consensus goes further and proposes an actual CT and MR reporting template, on the explicit reasoning that procedural planning sites need consistent communication rather than consistent imaging alone.

The EACVI Task Force consensus states the clinical requirement plainly: pre-procedural assessment must establish annular size and shape, and the extent and distribution of valve and vascular calcification. Annulus and aortic root measurement are assigned to CT alongside 3D transesophageal echo.

None of that is exotic. What makes it demanding is that the report has to be the same report whoever produces it.

Reproducibility is the whole job

The evidence on how much annulus measurement moves is uncomfortable reading, and worth knowing precisely.

Across 82 TAVR screening CTAs read twice by each of three readers, a single reader produced a different device-size category between their own two reads in 22 to 26 percent of cases. A multi-reader strategy reduced that to 5 to 10 percent. The readers ranged from six years of experience to none, and intra-observer variability did not differ significantly across that range, which is the finding that should give any department pause. Experience did not protect against a reader disagreeing with themselves.

Software-assisted measurement narrows it. In 80 pre-TAVI patients read by three operators of differing experience, intraclass correlation between manual and semi-automated measurement ran 0.81 to 0.98 for annulus area and perimeter, and the experienced operator arrived at an identical area-derived prosthesis size in 96 percent of cases. Note where agreement weakened: coronary ostia distances came in at 0.75 to 0.79. The measurement that matters for occlusion risk is the one that holds up least well, and it deserves a second look regardless of how the rest of the study reads.

Reading time in the same study fell from 3.14 minutes manual to 1.50 and 1.72 minutes semi-automated. That is not the headline. The headline is that the tighter agreement and the shorter read arrive together, because they come from the same cause: less time spent placing points by hand is less opportunity for the placement to drift.

The reporting half of the problem

Standardizing the measurement solves nothing if the report that carries it is unstructured prose.

Cardiac imaging has direct evidence here. Comparing 50 coronary CTA reports read by six clinicians, structured impressions raised agreement on the number of vessels with significant stenosis from 53 percent to 68 percent, with kappa improving from fair to moderate. A larger study across 330 body CT reports found structured reports rated significantly higher on both content and clarity, though it is worth being straight about the limit of that finding: the same study saw no significant difference in the reports' graded clinical effectiveness. Referrers prefer structured reports and understand them better. That is a real benefit and it is not the same as a proven change in outcomes.

The European Society of Radiology's 2023 position on structured reporting takes the honest line: substantial potential to improve the service delivered to patients and referring physicians, an overwhelming majority of clinicians preferring itemized reports, and implementation in clinical routine still lacking. Cardiac CT has gone furthest of any subspecialty, through CAD-RADS and the SCCT interpretation and reporting guidelines. Structural planning has the templates. What it often lacks is a reporting path that carries the numbers out of the post-processing application without somebody retyping them.

In cvi42, measurements populate a structured report directly, and that report reaches the reporting environment through PowerScribe One, DICOM SR or HL7 rather than through a clipboard. The heart team receives the same document every time, in the same order, whoever read the study.

The generalist reader is not the problem

There is a persistent assumption that structural planning belongs to a subspecialist, and that a site without one is doing something provisional until it gets one.

The data does not support the premise. In the reproducibility study above, the reader with six years of subspecialty experience was not meaningfully more consistent with himself than the readers with three years and none. In the semi-automated comparison, the two less experienced operators reached intraclass correlations of 0.89 and 0.88 against the experienced reader's 0.96. What separated them was smaller than most departments assume, and it narrowed further with software assistance rather than widening.

Bicuspid anatomy is where this is usually tested, and even there the evidence pushes against the anxiety. In 217 bicuspid TAVI patients, annular sizing was appropriate in 96.3 percent of cases, and a supra-annular sizing strategy would have selected a different valve in 38.7 percent, worsening outcomes in more patients than it rescued. The registry evidence across 81,822 patients likewise shows 30-day mortality after TAVR in bicuspid and tricuspid anatomy that does not differ significantly, though 30-day stroke was higher in the bicuspid group at 2.5 percent against 1.6 percent.

Bicuspid disease demands more careful reading. It does not demand a different reader.

The workload nobody costed

Structural planning arrived in departments that were already absorbing more than they could comfortably carry.

A systematic appraisal of 440 published imaging studies found that 48.3 percent would increase radiologist workload in an academic setting against 4.5 percent that would reduce it. The dominant mechanism, in the authors' own words, was an increase in interpretation time for an existing application, often combined with an increase in post-processing time. Advanced post-processing is not an incidental cost of doing more sophisticated imaging. It is the cost.

Which is the argument for doing structural planning inside the environment where the rest of the cardiac work already happens, rather than in a dedicated application reached through a separate login. cvi42 reads cardiac MR, cardiac CT, vascular CT and structural planning in one place, against one set of priors, on one training curve. A reader who already works there does not learn a new tool to take on structural cases. They take on structural cases.

Structural planning arrived in departments that were already absorbing more than they could comfortably carry.

A systematic appraisal of 440 published imaging studies found that 48.3 percent would increase radiologist workload in an academic setting against 4.5 percent that would reduce it. The dominant mechanism, in the authors' own words, was an increase in interpretation time for an existing application, often combined with an increase in post-processing time. Advanced post-processing is not an incidental cost of doing more sophisticated imaging. It is the cost.

Which is the argument for doing structural planning inside the environment where the rest of the cardiac work already happens, rather than in a dedicated application reached through a separate login. cvi42 reads cardiac MR, cardiac CT, vascular CT and structural planning in one place, against one set of priors, on one training curve. A reader who already works there does not learn a new tool to take on structural cases. They take on structural cases.

The standard for a structural planning read is not that it is expert. It is that it is the same read twice, and the same read from the colleague covering next Thursday.

That standard is achievable now, with the templates the societies have already published and software that holds the measurement steady while the physician keeps the judgment.

See a structural heart read in cvi42 →

Stay tuned for more on Structural Heart from the perspective of 3D labs, imaging lab directors and department heads, IT and PACS, and procurement and administration.