Echo: probe geometry, acquisition & quantification

Scan, companion panels & tools

Echo opens in Scan with three persistent desktop panels: the draggable chest/TEE probe, anatomical heart with the actual ultrasound plane, and a larger acquired echo display. Section and 3D heart switch between the finite section and the independent spatial teaching model. Fine probe movements and physiology expand below the visible patient. Narrow screens stack the panels with jump controls; the probe and anatomy are not removed. Full chamber names, colour keys and selected-structure identification remain available. Changing the teaching camera does not change acquisition. Learn, Cases and Quantify open the corresponding teaching workspace; discrepancy challenges, assessment, teacher controls and references are available through More. Assessment hides anatomical answers in the middle panel.

Depth, sector, gain and focus remain available below the image. Additional image controls and modality-specific settings expand on demand. In Explore, drag the green CW, PW, tissue or M-mode line handle to steer the actual sampling ray in TTE or TEE. Drag the gold PW/tissue gate to change its depth and direction, or tap the sector to place it. Steering the green PW/tissue handle preserves gate depth; CW and M-mode have no range gate. Pointer capture keeps mouse, pen and touch drags active across the image boundary; sector and displayed-depth limits include zoom. The vector cursor responds immediately while the acquired image retains its true frame timing. Calipers and Trace remain separate point-placement tools. Explore identifies tissue; Calipers and Trace place calibrated points on the actual acquisition. Measurements & saved views opens contour methods, retained acquisitions and sharing. Freeze resolves the same state at higher detail. The light control surfaces and navy/mint palette follow the main laboratory.

Recognizable sections and acquisition timing

Landmark chips and acquisition scoring inspect actual labelled voxels inside the finite ultrasound sector. They no longer infer visibility from intersecting bounding shapes. Required structures can remain absent at a named starting pose: translate, rotate or fan and verify the actual borders. Geometric intersection is not proof of acoustic visibility or clinical adequacy. Leader lines terminate on in-plane voxels; off-plane structures do not appear in the anatomical section.

TG short-axis levels distinguish basal valve, mid-papillary bodies and the apical cavity. The shared geometry includes a small LA appendage continuous with the LA. The arch turns toward patient left; the descending thoracic aorta lies left/posterior to the oesophagus and is separated from the right-sided IVC, rather than inheriting a rotated heart-local vessel path. Corrected TEE aortic and transgastric starting positions act through the ordinary insertion, shaft, flexion and omniplane mechanics. Absolute positions and synthetic geometry remain original implementation parameters, not universal patient coordinates. The five-chamber and other challenging sections may still need targeted reacquisition.

Linked anatomy, ECG and spectral displays follow the state and cardiac time of the completed ultrasound frame. Freeze, calipers and saved acquisitions retain that displayed state, rather than a newer pose waiting in a worker queue. Frozen full-resolution images use the same acquisition and timing; image refinement does not advance the cardiac cycle.

Only the acquisition canvas follows requested animation-clock ticks, up to approximately 60 requests per second rather than the former 30-request cap. The workstation follows completed images; pausing stops the acquisition clock. Exact acquisition feedback is computed once in the sampling workers alongside the completed frame rather than independently resampling the plane on the main thread. Busy renderers coalesce pending requests. During a probe sweep, completed full-detail images remain visible even when a newer pose is pending; each image and its linked metadata retain the pose and cardiac time actually sampled. Patient, calibration, receiver, mode, freeze, restore and reverse/cine transitions invalidate older work with monotonic generations, including an A-to-B-to-A change. Frozen probe sweeps retain their exact cardiac time. Native sampling, frozen calibration and anatomy are retained; request rate is not a guarantee of displayed frame rate.

The stationary TTE chest diagram redraws when the probe or its display changes, rather than rebuilding anatomy for every cardiac tick. Offscreen patient, spatial-anatomy and signal-history panels retain their controls and resume drawing from the latest completed acquisition when visible again. The histories retain their selected native density and reconstruct the exact requested frozen endpoint on return. This saves work on stacked mobile layouts without substituting another phase or a preset image.

Scrolling spectral and pressure histories reuse model evaluations on an absolute time lattice at the selected sampling density, with independently evaluated exact window endpoints. The time-to-screen mapping uses each recorded timestamp. Changing patient state, rhythm, probe or sample gate invalidates the relevant cache. Values are not interpolated from a representative heartbeat; AF and respiratory variation remain time dependent. The lattice changes interior sample timestamps from the older moving-window renderer while retaining temporal resolution.

Coordinates, marker & imaging plane

Patient axes use millimetres: +x toward patient left, +y posteriorly and +z superiorly. The heart has its own long-axis frame. The probe frame carries its origin, depth/beam direction, marker/display-right direction and plane normal. Cardiac display conventions put the physical index marker on display-right, with the probe at the top and increasing depth downward. An apical four-chamber image places left-sided chambers on display-right and right-sided chambers on display-left. Camera rotation never changes acquisition or these conventions.

Drag the physical TTE footprint on the torso to translate it. Drag the TEE tip along its drawn path to advance/withdraw it; the other TEE controls remain independent. TTE translation, axial rotation and two local angulation axes act on the pose. TEE insertion follows an idealised oesophageal–gastric path. Shaft rotation, tip flexion and electronic omniplane rotation have distinct mechanical transformations. The GE junction is an anatomical transition rather than a universal standardized view. Fixed angles do not guarantee a view in clinical practice.

Image formation & supported acquisition modes

The current geometry uses tapered and sheared chambers, a crescent anterior RV, separate moving AV leaflet sheets with saddle hinges and a tapered aortic sinus/STJ. LV radial and longitudinal deformation preserve the reference cavity-volume determinant and parent myocardial volume; local LV tissue velocity follows that deformation. Annular Doppler remains a reduced-order teaching waveform. These are original parametric surfaces, not patient-derived anatomy or a validated segmentation.

Procedural chamber, vessel, valve and lesion geometry is intersected at sampled sector locations throughout the cardiac cycle. A deterministic complex scatterer field creates correlated speckle in moving tissue coordinates. Grayscale myocardium and dark blood are combined with beam-relative interface reflections, frequency/depth attenuation, focus-dependent resolution and representative window limitations. Semantic labels do not receive a uniform white outline. These are original synthetic material coefficients and image-formation approximations, not measured acoustic properties or a wave-propagation solution.

Image quality changes sampling density, never anatomy or millimetre calibration. Detailed quality retains 1080×720 live sectors and genuinely resamples frozen sectors at 2160×1440. Ultra detail uses 1440×960 live and 2520×1680 frozen sampling. Load never silently lowers either selected resolution; a slower device completes fewer full-detail images. Existing saved Automatic settings restore into Detailed and saved High detail settings restore into Ultra detail. Labels, axes and calipers have a separate HiDPI backing, bounded between 1440 and 3240 pixels wide according to display size and pixel ratio. Two to eight shared browser workers, bounded by available CPU capacity, partition the same calibrated sector with point-spread halos and identical proximal shadowing. The labelled Section companion uses the same completed tissue buffer; selecting the spatial heart avoids generating an unused labelled raster. Worker-unavailable devices retain the same model with an explicitly labelled CPU fallback. Reported display fps counts completed images, not animation callbacks. Thin tissue, shadowing, specular dropout and reverberation remain simplified and can limit visibility.

The spatial teaching surface uses WebGL2 when available. Reusable mesh topology evaluates every vertex and union boundary from the requested cardiac and respiratory state. Finer 36×72 chamber grids, 72-sided vessels and 24×48 mobile lesions improve silhouettes; shared normals provide smooth lighting within each structure. The mitral leaflet sheets and representative LV papillary bodies, heads and chordal fans are also evaluated from the existing sampled model, including rupture, and become visible within the cutaway. These simplified coefficients are not measured normal anatomy or a fully segmented mitral apparatus; detailed tricuspid leaflet/chordal morphology is not represented. Instanced quads avoid duplicating triangle attributes during upload. The sector, axes, targets and text stay on the labelled overlay. The camera projection and picking remain in patient coordinates. WebGL/context failure retains the Canvas drawing path and identifies the fallback. GPU shading is a teaching-model rendering optimization, not a 3D echo acquisition or a change to sampled ultrasound anatomy.

A dedicated anatomy worker builds, sorts and packs the exact-phase surface away from the interface thread. Transferable GPU attributes and patient-coordinate pick geometry use bounded recycled buffers. One active and one latest pending request prevent a growing frame queue. Plane and label overlays use the completed surface acquisition; freeze and cine reject obsolete live results. Picking reconstructs retained geometry only when needed. Probe and camera motion may display an already completed packet while a newer one is pending; projection, labels, the sector and picking all use that packet’s captured camera, pose and time. Devices without the worker or GPU keep an explicitly identified Canvas compatibility path.

The sector renderer caches calibrated beam geometry and receiver coefficients. Per-frame chamber and flow constants and a bounded 12 mm conservative occupancy index avoid repeated work. The index only rejects absent candidates; each retained structure receives its original continuous geometric test. It does not voxelize anatomy at 12 mm. Prepared frame geometry is fixed for its state and instant; edits rebuild it. The current respiratory offset, tissue deformation, lesion state and flow are still evaluated for every new image. Eight-row tile halos are derived from the bounded gradient, axial/lateral interpolation and colour-bloom support, retaining the same full-image samples and shadows. Borrowed sampling and material scratch buffers are consumed immediately; they never become saved acquisition state. Borrowed flow vectors reuse their output storage while ordinary sampling calls retain independent vectors. Receiver channels accumulate locally in their original arithmetic order; a final grayscale calculation is omitted only when valid acquired flow colour completely replaces its RGB, while all RF samples and neighbouring contributions remain. These optimizations do not interpolate between images or change native resolution.

Anatomical borders use additional subpixel samples of the same imaging plane to reduce stair-stepping. Selection adds a thin contour and subtle tint rather than replacing tissue texture. Segment selection follows the intersected segment; an off-plane structure cannot become visible through a highlight. Structure contours are an optional teaching overlay. The spatial teaching heart uses smoother surface tessellation and interpolated lighting over its existing geometry; those display improvements do not add patient anatomy.

Colour uses the projection of a synthetic velocity vector onto the local beam; scale controls wrapping and gain controls visibility/blooming. Colour does not label oxygenation. Aliasing alone does not establish turbulence. PW samples a bounded gate; CW samples along the beam without range specificity and must not use PW-style aliasing. Tissue Doppler samples synthetic moving tissue. Detailed spectral displays sample 1440 time points and ECG 2400 points per strip; Ultra detail uses 1920 and 3000 respectively. M-mode samples the actual selected line at 800 time positions by 480 depths in Detailed, or 960×576 in Ultra detail, and both axes increase by 50% when frozen. Shared workers sample complete depth rays on a global time axis. Live M-mode reuses identical historical columns from an absolute time lattice and evaluates only entering columns, permitting refresh up to 24 Hz. Its recorded endpoint trails the completed image by less than one native time sample before refresh scheduling; ECG and the displayed time axis use the strip's actual timestamps. Changing anatomy, rhythm, beam, receiver or sampling density invalidates reuse. Freeze independently samples its exact unsnapped endpoint at higher resolution. Row-wise assembly is byte-identical to a full acquisition on the same lattice. Queued live 2D tiles receive priority over queued history tiles.

The 3D acquisition has finite pyramidal coverage and cardiac timing. Detailed acquires 144 lateral angles × 108 elevational angles × 168 depths (2,612,736 voxels); Ultra detail acquires 160×120×192 (3,686,400 voxels). Cropping, volume views and enlarged MPR sections operate on these recorded samples, separately from the unrestricted teaching anatomy. MPR uses trilinear intensity interpolation and acquired voxel labels, with explicit orientation and finite crop limits. Coverage, sampling density and synthetic multibeat seams limit what can be assessed. Smoother display interpolation does not recover unobserved tissue or turn the synthetic model into a clinical image.

The current side of a saved-view comparison reuses the primary completed native image and captured acquisition metadata, avoiding a duplicate live scan. The saved side still acquires its retained pose/time. Derived contour and Doppler calculations are recomputed when their actual learner inputs change rather than on each animation frame. A bounded three-plane MPR raster cache reuses unchanged acquired sections during camera rotation. Its key includes the captured volume, native section dimensions, crop boundary, plane depth and selected recorded structure. Live gain or Doppler cursor changes do not resample an already acquired volume; changing an acquisition or crop invalidates incompatible sections. Camera-independent MPR picking still uses the actual current acquired coordinates.

Learner measurements, units & numerical assumptions

Caliper distance is computed in the acquired plane in millimetres and displayed in centimetres. Polygon area is computed from actual learner points and converted from mm² to cm². Biplane Simpson volume uses 20 elliptical discs from paired traced chamber silhouettes with the shorter acquired long axis; mm³ is divided by 1000 for mL. Start at one annular hinge, follow the endocardium around the apex, and finish at the other hinge. Long-axis mismatch is flagged for review rather than corrected automatically.

ED/ES, A4C/A2C and maximal LA labels are learner declarations. Saved contours retain the view and time. Verify end-diastolic/end-systolic timing, maximal LA timing, complete borders, matching beats and exclusion of pulmonary veins/LAA. Foreshortening, wrong contour or phase remains a measurement error. LVEF=(EDV−ESV)/EDV; RV FAC=(EDA−ESA)/EDA. LA volume indexing divides acquired volume by entered BSA in m².

Spectral tracing uses learner points calibrated to the displayed time/velocity axes. VTI integrates the piecewise-linear velocity envelope. Mean gradient integrates instantaneous 4v² analytically along each traced line segment, rather than squaring mean velocity. Trace one complete valid envelope with zero-velocity endpoints and verify beat selection. The workbench accepts these acquired results or explicitly entered measurements. LVOT stroke volume=πD²/4×LVOT VTI, continuity AVA=LVOT SV/AV VTI, and CO=SV×heart rate/1000 L/min. Diameter is in cm and VTI in cm. A 10% diameter error produces a 21% area/SV error. These equations assume appropriate sampling and matched representative beats; substantial shunting, flow contamination and regurgitant flow require context.

Simplified Bernoulli uses 4v² with v in m/s. Mean gradient must integrate instantaneous gradients over the envelope, not square mean velocity. RVSP=4(TR velocity)²+estimated RAP; equating it with PASP requires no intervening RVOT/pulmonary obstruction. A truncated TR envelope may underestimate velocity. E/e′ is an estimate within an integrated context-specific diastolic assessment, never a direct wedge-pressure reading.

PISA, vena contracta, valve areas & equations

Select Quantify within Echo for five acquisition guides, conceptual diagrams, editable calculators, worked examples and method-specific limitations. Inputs begin as illustrative values; deliberate caliper/trace copies or manual entries feed the calculations. No underlying patient truth is silently inserted. Calculator fields are working notes; saved source acquisitions retain calipers, traces, phase and image settings. The coarse synthetic convergence/orifice is not a validated clinical PISA or vena-contracta reference image.

For MR PISA, hemispheric Q=2πr²Va in mL/s uses radius in cm and the actual directional alias velocity in cm/s. EROA=Q/VMR in cm² requires MR velocity converted from m/s to cm/s and phase-matched to the radius. EROA×MR VTI in cm gives mL only if that EROA represents the regurgitant interval. Dynamic, nonhemispheric, constrained or multiple-jet flow can invalidate the simple method. Radius error is squared. Vena contracta width crosses the narrowest downstream neck perpendicular to local flow; it is neither convergence radius nor distal jet area. An irregular orifice cannot be assigned a 3D area from one 2D width. No universal width-to-severity rule is applied.

Continuity AVA uses matched forward LVOT SV divided by AV VTI; LVOT area is πD²/4 and SV is area×VTI. Diameter in cm, VTI in cm and mL/cm produce cm². Dimensionless VTI ratio avoids the diameter term while retaining sampling/alignment errors. AR alone does not break forward LVOT–AV volume equality. Mitral continuity is a selected complementary method: significant MR/AR and unmatched filling can break LVOT–MV volume equality. AF requires careful comparable-cycle selection; this is not a universal AF workaround.

PHT-derived MVA=220/PHT(ms) is an empirical rheumatic-MS calculation, subject to chamber compliance, MR/AR, fused or truncated slopes, calcific disease and recent valvotomy. Pressure half-time ends when E velocity reaches E/√2, not E/2. Anatomical planimetry traces the smallest true leaflet-tip opening orthogonally in the appropriate maximal-opening phase; it describes a different quantity from Doppler effective area. Historical 2009 technical methods are contextualized by 2023 rheumatic guidance; superseded severity tables are not reused.

Bernoulli peak ΔP=4V2² or, when appropriate, 4(V2²−V1²), with velocities in m/s and gradient in mmHg. The traced mean remains the simplified mean of instantaneous 4v(t)², not a modified-Bernoulli mean or the square of mean velocity. Misalignment, pressure recovery and non-negligible proximal velocity affect interpretation. Doppler maximum instantaneous and catheter peak-to-peak gradients are different quantities. Interpret areas, gradients and flow together.

Consistent pathology & conflicting findings

Cases define persistent structural phenotypes separately from acute loading, rhythm and contractility. Vegetation, disrupted papillary apparatus, effusion and other synthetic lesions belong to the same spatial model and should appear only when intersected. The mandatory LA discrepancy uses asymmetric atrial enlargement. RV discrepancies distinguish a plane error from true regional RVOT enlargement.

The learner identifies a conflict, checks quality and plane, verifies phase/beat/contours/units/indexing, identifies the compartment and variable, reacquires targeted views, integrates independent evidence, and writes a conclusion with remaining uncertainty. Automated formative acquisition feedback scores actual finite-sector landmark intersections, plane and apical completeness; it is not a clinical reasoning grade. Teacher mode supplies a separate faculty-entered rubric for acquisition, alignment, measurement and reasoning, retained with the acquisition. Faculty must review interpretation, measurement technique and validity of alternative views.

Saved acquisition, assessment & teaching boundaries

Browser-local synthetic acquisition records retain probe and patient state, acquisition settings, cardiac time, calipers, contours and report. Export/import and share-state links support reproducible teaching. These are not an authenticated clinical archive. Teacher model assumptions are separate from acquired learner measurements and are hidden in the assessment presentation.

Links connect to the valve editor, PH education, haemodynamic laboratory and classroom. Echo state sharing is presently by file/link; the existing live classroom haemodynamic broadcast does not automatically carry Echo acquisitions. No generative AI service is required to produce frames.

Guideline versions & review priorities

The editable atlas maps the 2013 comprehensive ASE/SCA 28-view TEE framework, with additional transition teaching distinguished from numbered standard views. Adult chamber quantification, right-heart, diastolic, valve, pericardial, 3D and artifact references are versioned in the Echo content. Guidance is educational context rather than validation of original synthetic coefficients.

Faculty review is still required for normal view anatomy, probe marker conventions, lesion fidelity, numerical calibration, quantitative-task design and interpretation. Do not infer diagnostic sensitivity, validated stenosis/regurgitation severity or procedural training adequacy from a passing software test.