Myocardial infarct size quantification in mice by SPECT using a novel algorithm independent of a normal perfusion database

Animal experiments were performed in accordance with institutional guidelines and approved by the Animal Care Ethical Committee of the University of Liège. Myocardial infarction was induced in 2-month old female C57Bl/6 mice (18.5 ± 0.9 g). Anaesthesia was induced and maintained using ketamine (100 mg/kg) and xylazine (5 mg/kg). Mice were orally intubated with a 22-gauge vinyl catheter, and positive pressure ventilation (0.3 ml/min, 160 breaths/min) was maintained with a rodent ventilator (Harvard Apparatus, Les Ulis Cedex, France). The heart was exposed through a 1-cm left lateral thoracotomy. The left coronary artery (LCA) was ligatured by passing an 8-0 polypropylene silk suture around the artery. The chest wall, muscle layers and skin were then closed with interrupted 6-0 polypropylene silk sutures. Intubation was discontinued once spontaneous breathing resumed, and the mouse was allowed to recover on a heated platform.

A total of 11 infarcted animals and seven healthy controls were scanned. SPECT acquisitions were performed using the Linoview SPECT system (Linoview System, Amsterdam, The Netherlands) [12]. The slit width was set to 0.6 mm. Acquisitions were performed 7 days after the LCA ligature. Animals received an intravenous injection of 173 ± 27 MBq of ^99mTc-sestamibi through the tail vein. Two hours later, 0.1 μg/kg of cholecystokinin was injected intraperitoneally to induce emptying of the gallbladder. The animals were scanned 30 min later, under anaesthesia (50 mg/kg of ketamine and 2.5mg/kg of xylazine) for 30 min.

Data were reconstructed using the expectation maximization maximum likelihood algorithm without attenuation or resolution correction. The images were reoriented into conventional short-axis, vertical long-axis and horizontal long-axis views.

As previously reported by Constantinesco et al. [6], an appropriate scaling factor was applied on the reconstructed images in order to approach the human reference heart volume, and a 20-segment polar map was automatically derived with the QPS software (Cedar-Sinai Medical Center, Los Angeles, USA) with constraining limits being manually adjusted [13‐15]. Relative segment level was defined as the ratio of the segment value given by the QPS software with that of the segment having the highest tracer uptake.

The novel proposed method for AIF determination includes two steps: the determination of the myocardial wall volume corresponding to each segment of the polar map and the determination of the contribution to each segment to the total infracted volume.

Step 1. The volume of each segment was measured on the central vertical long axis slices as shown in Figure 1 where the lines are the cross section of the volume of interest drawn by QPS to derive the polar map. The volume V _s of the segments is as follows:

The numerator of Equation 1 is the volume of the ring made by the N _s segments at the same radials distance in the polar map. Equation 2 gives the circumference C _s of the ring assuming that the ring is circular with a mean diameter D _s measured between the centres of the two sections of the ring. σ_s is the area of the ring section estimated as the product of the length L _s and thickness T _s. These two distances were measured starting at the middle section sides in order to take into account the non-parallelism of the sides. The number of segment N _s of the polar map at each level are illustrated in Figure 1 (basal, n = 6; middle, n = 6; apical, n = 6; apex, n = 2).

Step 2. A numerical model of the mice’s left ventricle was developed in order to determine the infarcted fraction in a segment as a function of its relative level. The ventricle was modelled by a 1-cm-length cylinder of 5.4-mm-diameter (mean diameter observed in SPECT, 5.4 ± 0.7 mm) and with a uniform active wall thickness of 1 mm. A 100% activity defect was modelled on the whole wall length and with a polar extension ranging from 0° to 240° that, in this model, corresponded to a total infarcted fraction ranging from 0% to 66%. The beating was modelled by a 0.35-mm radial oscillation of the wall associated with a longitudinal twisting of 3° [16]. The breathing was modelled by a 2-mm transverse oscillation of the whole ventricle [17]. The breathing oscillation direction was modelled perpendicularly and tangentially to the defect as well. A surrounding background with specific activity 10%, 20% and 30% of that of the non-infarcted region was modelled (19.3% ± 4.9% observed in the mice SPECT). In real SPECT, this background activity resulted from the residual vascular and extravascular pool activity and from the cross contamination of high uptaking of Tc-99m in tissues such as liver and gallbladder as well.

After being blurred by the beating and breathing, the distribution activity was convolved with the mean tomographic 3D spatial resolution of the system which was about 1.3 mm at the heart location for the slit width set to 0.6 mm. Motion blurring and system spatial resolution spread the surrounding background activity inside the defect area. Lastly, the wall activity was summed inside a 60° arc segment circumferentially moving in order to compute its infracted fraction as a function of its relative level (Figure 2).

When a segment circumferentially moved through an ideal defect-to-normal region edge, the variation of its infarcted fraction is linear versus its relative level (A to D in Figure 2). When this step transition was smoothed by the altering resolution effects (beating, breathing and system resolution), the relative segment level already increased although the segment was still fully located in the defect area (E to F in Figure 2). The same behaviour occurred when the segment was completely inside the normal region and came close to the defect area. As a result, the curve of the infarcted fraction in segment as a function of the segment relative level underwent a pivoting (I in Figure 2).

To account for these features, a weighting factor P(U _s) giving the infarcted fraction in a segment as a function of its relative level U _s was modelled by a s-shaped curve:

U _c is the level corresponding to the inflection point of the curve and n determines its slope at the inflection point. The AIF of the myocardial wall was

The denominator of Equation 5 is the total myocardial wall volume. U _c and n were fitted in order to obtain the best agreement between the infarcted percentage assessed by TTC and by SPECT.

Because 2,3,5-triphenyl tetrazolium chloride or TTC staining is coupled with mitochondrial metabolism and Tc-99m-sestamibi is sequestered in mitochondria of viable myocardium, infarct size estimated by Linoview SPECT was compared to TTC staining [18, 19] which was performed as previously described [20]. Briefly, after SPECT imaging, the animals were euthanized, their chest was opened, and their heart was excised and sliced into serial 1-mm-thick short-axis slices, parallel to the atrioventricular groove. The myocardial slices were weighted and further incubated at 37°C in 1% TTC for 20 min. To avoid any issue of contrast lost or chromatographic change that can occur during digitalization, the boundaries between the infarcted and viable myocardium regions were manually drawn using a binocular magnifier on a transparent sheet directly set in contact with the heart slices. Afterwards, the transparent sheets were scanned and the area of the regions was assessed using the Image J software (Bethesda, MD, USA). With this method, the epicardial and endocardial borders of each slice as well as the contours of the infarcted area were hand-traced and expressed in percentage of the total slice area. The weight of the infarcted and non-infarcted myocardium was then calculated by multiplying their relative proportion by the weight of the corresponding myocardial slice. The infarct size was then calculated as the sum of the infarcted segments in each individual slices and expressed in percentage of total myocardial mass. SPECT and histological analysis were performed by the Université Catholique de Louvain and the Université de Liège, respectively, with prior knowledge of the analysis of the other team.

TPD [13] was computed using QGS (AutoQUANT release 7.2, Philips Medical Systems, San Jose, CA, USA). A normal perfusion database and limits were built using the seven normal mice SPECT data.

Results are expressed as mean ± SD. Data were compared by the paired two-tailed Student’s t test, by linear regression and by the Bland-Altman test. P < 0.05 was considered statistically significant.

Short axis, vertical long axis, horizontal long axis slices and the corresponding polar map of a normal mouse are shown in Figure 3. The left ventricular myocardium is clearly delineated, and the thinner right ventricular myocardium is also visible on some slices. Images from a mouse with a myocardial infarction are shown in Figure 4. The perfusion defect is clearly apparent and was delineated by QPS.

Figure 5 shows the impact of the different effects on the infarcted fraction in a segment as a function of its relative level for the 60° defect polar extension. The beating had a marginal impact. The system resolution and the breathing induced a significant clockwise pivoting of the curve, while the surrounding background induced a significant shift of the curve to the right. When the resolution is too altered, segment relative level below a certain threshold cannot longer be observed. The infarcted fraction in a segment with 0.55-relative level can range from 0.70 to 1.00, but there is no possibility to assess this fraction. Note that the breathing also induced an artifactual hypo-perfusion in the normal ventricle model (Figure 5 bottom row).

Figure 6 shows the infarcted fraction in a segment as a function of its relative level for a defect polar extension of 40° and 240° and for a breathing motion perpendicular or tangential to the defect. When the defect polar extension decreases below 60°, the segment always intercepts the normal wall. For 40° polar extension of the defect, the infarcted fraction of the segment level of 0.7 is indeterminate from 0.4 to 1. Figure 7 shows the impact of the background level on the curve of the infarcted fraction in a segment as a function of its relative level for the 60° polar extension defect.

The optimal s-curve parameters found in order to have the best correlation between AIF and TTC were U _c = 0.55 and n = 9. This experimental curve was very close to the theoretical predictions derived from the numerical ventricle model (Figures 6 and 7).

Figure 8 shows the correlation between AIF and the total infarcted fraction for different surrounding background levels and for different defect polar positions versus the polar map segmentation and versus the breathing directions. AIF was obtained from the numerical ventricle model using Equations 4 and 5 with the experimental parameters U _c = 0.55 and n = 9. The linear regression on all the simulated infarctions was y = 0.97x – 0.01; R² = 0.94.

All the normal mice used to build the normal limits in QGS got 0 for their TPD assessment, while their AIF value was 4.4% ± 1.6%. TTC ranged from 2% to 66%. Mean ± std infarcted fractions were 37.4% ± 13.8%, 37.9% ± 17.5% and 35.6% ± 17.2% for TPD, AIF and TTC, respectively. As shown in Figure 9, AIF provided an excellent correlation with TTC (R² = 0.94) which is higher than that found with TPD (R² = 0.76). Furthermore, the AIF-TTC linear fit passed through the axis origin. There was no significant bias between AIF and TTC, as demonstrated by the Bland-Altman plot (Figure 10).