CARAT Tutorials: Myocardial Jeopardy

Although coronary angiography is primarily a descriptive tool, its ability to define more quantitative myocardial-coronary disease relationships continues to be explored. These explorations have been only modestly successful as they arise from empiric anatomic assumptions that fail to capture the significant anatomic variability seen between patients. The CARAT software has attempted to capture this anatomic heterogeneity, hopefully leading to greater utility in small cohort and individual patient assessments.

Direct myocardial imaging techniques are reliable when anatomic or distinct biochemical or functional myocardial boundaries are present, but they are of little value in the absence of these delimiters and angiography may then play a role. Clinical situations in which angiography may have a quantitative advantage include:

17 Segment Model

Unlike the anterolateral, obtuse marginal and posterolateral myocardial region designations used in most published quantitative angiographic models, CARAT has adopted the 17 spatial designations proposed by the Cardiac Imaging Committee of the AHA. (See figure 1 below). In this model the LV is divided into three (basal, mid and apical) slices of equal thickness from base to apex. Each slice contains free wall and septal segments that are separated by the insertion points of the RV free wall. There are 11 free wall segments, 5 septal segments and a distinct apical (non-cavitary) segment.

We have completed precise CMR measurements of the size of each of the 16 free-wall and septal segments in a cohort of normal volunteers. Although the original assumption that each of the 16 segments constituted 6% of the LV myocardium was close, more precise measurements are now available and have been included in the CARAT Jeopardy calculation algorithm. These normal measurements are shown in the table below. Myocardial volume measurements excluding papillary muscles were used.

The most effective graphic presentation of this model is one using a circumferential polar plot as indicated to the right of the diagram with the rings (from outside to inside) representing the basal, mid and apical layers and the center (area 17) the non-cavitary apex.

Figure 1: 17-Segment Model.

LV Free Wall Branch Assumptions

A major goal of CARAT is to link each side branch location and size to specific segments in the above 17-segment model. The reporting conventions of the CASS registry with respect to branch configurations have stood the test of time and, with minor modifications, are adopted in CARAT (Circulation 1981;63:I-1). In particular the specification of a maximum of 3 branches for each of the anterolateral, marginal and posterolateral regions, plus a tenth ramus intermedius branch, each with three size categories, has been adopted by CARAT. The basic CARAT assumption is that a Size 1 branch supplies the majority of one myocardial segment, a Size 2 branch supplies two segments and a Size 3 branch supplies three. (A similar assumption was made in the APPROACH Jeopardy score assessment derived from prospectively used HeartviewÓ diagrams in which a strong correlation between the resulting score and one year mortality was found in a cohort of 20,000 patients. M Graham et al Am Heart J 2006;142:254.)

Diagonal, marginal and ramus side branches are assumed to follow an oblique course consistent with the usual direction of superficial myocardial fibers. Branches are designated as basal, mid or apical with respect to their points of origin along the LAD and anterolateral, inferolateral or inferior along the AV circumflex. We avoided ordinal numbering of branches to emphasize their spatial association with specific myocardial segments. These associations are shown on the charts below. In addition to the usual oblique configuration of free wall branches, distribution variants are occasionally seen in which the oblique direction is not followed. The most common variants follow either axial or transverse directions for diagonal branches and a transverse direction for ALM branches. These options are highlighted in yellow along with their corresponding myocardial region assignments.

Figure 2: Quantitative LV Assignments - Diagonal and Marginal Branches.

Posterolateral Branch Assumptions

The table in figure 3 below specifies segmental myocardial assignments for the PDA as well as right, mid and left posterolateral branches. The PDA assignment procedure is unique. Unlike the anterior interventricular artery (LAD), which makes its contribution to the anterior LV free wall through prominent diagonal branches, the PDA makes its important contributions to the adjacent inferior segments through channels that are very much less distinct. For this reason we have had to rely on the classic pathologic observations of Kalbfleish and Hort (Am Heart J 1977;94:183) to specify the quantitative and spatial contributions of the PDA to in the inferior and septal segments.

Figure 3: Quantitative LV Assignments - PDA and Posterolateral Branches from RCA.

Septal Branch Assumptions

The picture in figure 4 below, from Bertho and Gagnon (Chest 1964;46:251), illustrates the usual balance between perforating arteries to the septum from the posterior (white) and anterior (red) interventricular arteries. Although there is variability, the anterior (LAD) septal branches usually supply 2/3 of the septum. As shown in the picture, there tends to be a gradient in relative inferior and anterior septal contribution between the base and apex with the LAD assuming proportionately more responsibility for the septum as one proceeds toward the apex. CARAT assumes that the relative contribution remains the same from base to apex.

The 17-segment model proposes 5 septal segments, 2 basal, 2 mid and 1 apical. The basal and mid anteroseptal segments and the anterior 25% of the basal and mid inferoseptal segments are supplied from the LAD. The remaining inferior 75% of the basal and mid inferoseptal segments are supplied by from the PDA perforators. The apical septum blood supply is more difficult to specify as this area is determined to a large extent by the variable size balance between the LAD and PDA. We have specified that the distal LAD and PDA together supply the apical septum, 1/3 by distal PDA perforators and 2/3 by the distal LAD in a system with a Type 2 LAD, 100% from the PDA with a Type 1 LAD and 100% from the LAD with a Type 3 LAD.

Figure 4: PDA and LAD contribution to septum.

Myocardial Assignment Process

As each coronary segment or branch is constructed in CARAT, specific properties are assigned to each component (line) that traverses a new LV segment. If a branch traverses more than one myocardial segment, myocardial values are assigned to each segment. This is a transparent process that can be seen by right-clicking on a branch of interest > right-clicking on ‘Edit Vessel’ > right-clicking on the LV segment of interest > selecting ‘Line properties’. You will then see a view (similar to the one shown in figure 5 below) with segment number and the weighting value (RMA) assigned – from 0 to 1.

Figure 5: Process of LV assignment to coronary segments.

Initial Tally of Myocardial Assignments

In constructing the coronary tree, we ask operators to pick template combinations that best depict the location and distribution of branches and segments. We also ask that they designate branch sizes that indicate the desired size/distribution area relationship. It is not necessary to make template selections that yield a summed value for each segment of “1.0” in all cases, for we have included a reasonable equilibration process that adjusts for under and over-assignment of myocardial values to individual segments.

The first step in this equilibration process is to make a myocardial assignment tally based upon the tree diagram generated by the angiographer. Selecting File/Unadjusted RMA Form from the tool bar will reveal this tally. An example is shown in figure 6 below. Some segments have the optimal value of 1.0 but other regions have summed values that are less than or greater than unity. For these areas adjustments will be made.

Figure 6: Tally of assigned myocardial values.

Myocardial Assignments Talley Adjustments

For segments such as 1, 6, 12, 4 and 14 in the above tally illustration, the total myocardial assignment from contributing branches has exceeded 1.0. When this occurs, the contribution from each vessel is proportionately reduced to end up with a final summed assignment of 1.0. When the contribution tally for one or more regions is less than 1.0, a different strategy is needed. Specifically a default arterial segment has been assigned to each myocardial segment.

When a tally is less than 1.0, the unassigned value is given by default to pre-specified regions, as shown on the table in figure 7 below. (These defaults were derived from recent CMR-angiographic correlation work in a series of single-vessel STEMI patients. The report correlates infarct-related arteries to injured ventricular segments from the 17-aegment model (JT Ortiz-Perez et al JACCImg 2008;1:282). Importantly, apical segments and the mid- anterolateral segment default to the LAD in all LAD-related STEMIs.

Figure 7: Default artery assignments for LV segments.

Final Quantitative/Qualitative Myocardial Picture

The finished CARAT diagram provides an adjusted index of the amount of myocardium supplied by arteries that are narrowed by > 75% (or >50% in the case of the LM). The total jeopardy score is indicated in the upper right portion of the final CARAT diagram. In the example in figure 8 below, an acute inferior STEMI is described. The area supplied by the occluded RCA was calculated to be 31%, a designation that also can be displayed by right-clicking on the lesion of interest while in the ‘lesion’ editing mode and then select ‘Area at risk’. In this example there were other lesions greater than 70% in the LCA yielding the total jeopardy calculation of 48%. The myocardial segments affected by lesions greater than 70% are highlighted on the circumferential polar plot to the lower right.

Figure 8: Example report with jeopardy and area-at-risk scores.

Sources of Error in Model

1. Branch length – myocardial volume relationship. The strongest theoretical association between branch size and the amount of myocardium subtended by a branch is based upon accurate assessment of vascular luminal area, a measurement that is not possible in the presence of coronary disease. The branch length-to-myocardial volume relationship has been carefully explored by C. Seller et al (Circ. 1992;85:1987) who demonstrated a strong relationship between the summed length of secondary plus tertiary branches (easily visible on angiography) and myocardium supplied. This summed value is not practical for clinical use either. The assumption made in CARAT, however, is that a dominant vessel crossing most of a myocardial segment will supply most of that segment, a significant oversimplification of the relationship described by Seller and introduces a potentially important source of error. The process used in CARAT of adjusting vessel contributions such that the total value for a segment does not exceed or fall short of unity should reduce, but not eliminate, the size of this error.

2. Attention to detail in vessel template selection. As a reporting tool increases in complexity, there is increasing dependence upon due care and attention of the operator as vessel template options and lesion features are selected. A reporting tool must ‘walk the balance’ between achieving inter-observer reproducibility yet capturing important complexities in individual coronary trees. Reporting consistency in CARAT is encouraged by the use of intuitive template options of basic coronary structure and by providing immediate diagrammatic feedback of selections made. To minimize errors, sub-branching complexity has been removed from initial template options with the freedom to add this complexity at a later stage by vessel editing functions as described in other tutorials.

3. Is CARAT too complex for widespread use? Most published quantitative coronary models are based upon common major epicardial coronary artery configurations that rarely capture the anatomic variations seen in individual patients. For this reason the utility of published models is limited to global observations in larger populations with little precision when applying the model to individuals or small patient cohorts. The CARAT methodology should allow greater precision, but the tool is somewhat complex and is not widely available. There will be an ongoing effort to clarify the methodology by making an interactive electronic spreadsheet available to all interested investigators. The document will allow ‘point-and-click’ selection of branch configurations and lesion locations with automatic myocardial volume calculations.


Although the regional myocardial assignment and jeopardy values in CARAT are based upon reasonable assumptions, they are not derived from direct area or mass measurements. As such their precision, in relative and in absolute terms can be questioned. Caution must be exercised in basing clinical decisions upon the specific numbers generated.

Despite this important disclaimer, the CARAT calculations and renderings can give useful insights into the myocardial regions affected by significant coronary lesions thereby allowing more informed integration of data from multiple imaging techniques. The CARAT data may also present useful insights into area at risk in acute coronary syndromes and into the completeness of revascularization after intervention, particularly when used in qualitative and relative terms.