Cardiac left ventricular myocardial tissue density, evaluated by computed tomography and autopsy

The study population is part of a prospective nationwide autopsy-based Danish forensic study, SURVIVE-let the dead help the living, which focuses on deceased with mental illness [25]. The deceased were autopsied at the Section of Forensic Pathology at the Institute of Forensic Medicine, University of Copenhagen over an eleven-month period, from January 2014 to November 2014. The study group comprised 116 deceased that fulfilled the following inclusion criteria: individuals with a known or a suspected mental illness, and the intake of antipsychotic, antidepressive or anti-anxiety medication or a suspicion thereof. The exclusion criteria were putrefaction, suspicion of a criminal act, inadequate CT image quality for the analysis and failure to scan due to severe obesity. This led to the exclusion of seven individuals. After a medical inquest, the deceased underwent a whole-body PMCT. The scans occurred within 24 h prior to autopsy. At the autopsy (see details below), the sex, age, weight and height of the deceased were recorded, as were the size, shape and macroscopic changes of the heart. The project was approved by the Danish National Committee on Health Research Ethics (1377517).

Non-contrast PMCT imaging was performed using a 64-slice Multi Detector Computed Tomography scanner (MDCT) (Siemens Somatom definition syngo 2010A; Siemens medical solutions, Forchheim, Germany). The following scanning parameters were used: tube current time: 300 mAs, tube voltage: 120 kV, slice collimation: 32 × 0.6 mm and slice acquisition: 64 × 0.6 mm. The iterative reconstructed slice thickness was 1.5 mm using a soft tissue convolution kernel. All images were acquired with the deceased wrapped in an artefact-free body bag in a supine position with elevated arms (except for one due to a BMI of 50). The bodies were placed in the CT gantry and scanned from head to toe. After reconstruction of the raw data, the images were transferred to the local PACS server.

Assessment of the myocardial tissue was performed with commercial software (Vitrea 6.3, Vital Images, Inc., MN, USA). Imaging analysis was performed by individuals who were blinded to the autopsy results. The non-contrast PMCT data were reviewed in multiplanar reconstructions after adjusting the planes to a long-axis view of the heart. The LV myocardial tissue was segmented by manually tracing the endocardial and epicardial border for each long-axis slice and adding up the corresponding volumes from each slice (Fig. 1). We used normal cardiac window settings with a level = 200 Hounsfield Units (HUs) and width = 1000. Voxels corresponding to myocardial tissue were identified using mean threshold attenuation values of 50 ± 10 HUs to distinguish it from the LV blood pool (Fig. 1). The region of interest (ROI) for the segmentation of both LV myocardial tissue and the LV blood pool was defined at the height of the septum and the LV outlet by HUs. The ROI was set at a minimum of 1 cm². Depending on the heart size, the number of manually traced slices varied from 70 to 100 in each case. Finally, based on the segmentation, the LVShV was calculated in millilitres and converted to mass, by multiplying it with 1.055 g/ml. Inter-observer variability was assessed in 25 randomly chosen individuals. Fifteen randomly chosen LV myocardial tissues were re-analysed using a different cardiac setting with level = 45 HU and width = 50 HU.

The autopsies were performed in accordance with the extended autopsy protocol of the SURVIVE study and the departments guidelines and were accredited by the Danish accreditation fund (DANAK). The length of the corpses were measured in the supine position from the top of the head (vertex) to the heel using an inelastic measuring tape. The heart was removed during the autopsy by preserving an adequate extension of the base and pulmonary vessels and dissected according to international standards detailed by Basso et al. [26]. After the removal of clotted blood, the whole heart was weighed. Dissection of the left ventricle was performed according to the method described by Bove et al. [27] with modifications; an incision was made in the posterior part of the left atrium following the atrioventricular groove to remove the atria from the ventricles. The ventricles were cut into transverse slices (short-axis direction). The right ventricle was separated from the septum and the left ventricle. The trabeculae encrusted in the septum and papillary muscles were preserved, and the hearts valves were removed. The epicardial fat was resected. The LV slices were blotted dry and separately weighed using an electronic scale (Mettler Toledo, ICS425, Glostrup, Denmark) with a 1-g precision.

The heart was recorded as being hypertrophic according to international forensic pathology standards and charts [28]. Anterior, lateral, posterior and septal wall thicknesses were measured at the level of the midventricular transversal slice. Papillary muscles were excluded from this measurement. Asymmetric LV hypertrophy was defined as biggest heart wall (millimetres) > 1.3*thinnest heart wall (millimetres) [7].

Data analysis was performed using the SPSS statistical software package (IBM Corp. IBM SPSS Statistics for Windows, Version 24, USA). Data was presented as the mean ± standard deviation, and a p-value of < 0.05 was considered statistically significant. Data distribution was tested with the Ancova/ linear regression test. Bland-Altman analysis was used to assess the re-analysis of LVShV and the degree of inter-observer variability. Scatterplots were used to illustrate the level of agreement and linearity between the CT obtained LVShV and the anatomic LVM as well as asymmetric LV hypertrophy. The degree of correlation was determined by Pearson’s correlation coefficient. The association between the CT-obtained LVShV and the anatomic LVM, with and without adjustment for gender, was assessed by linear regression analysis. A density factor was calculated by linear regression analyses, forcing the slope through origo (as done in comparable studies) [7, 8].

The results for anatomic LVM, LV wall thickness and LVShV are given in Table 2. There were 36 hypertrophic hearts, (25 men and 11 women) based on total heart weight. Thirty cases had asymmetric LV hypertrophy (19 men and 11 women) of which thirteen had hypertrophic hearts (total weight).

Overall, the CT-determined LVShV was highly correlated with the anatomic LVM (Pearson r = 0.857, p < 0.0001), also allowing us to assume linearity. Significant correlation was also found when the cohort was stratified by the presence of asymmetric LV hypertrophy (Fig. 2, Table 3) and by gender (Table 3). The inter-observer mean difference in LVShV assessment by CT was - 4.4 ml (95% CI: -26.4; 17.6). The mean HUs value of the ROI was 47 ± 7 in the segmentation of myocardial tissue and 68 ± 13 in the LV blood pool. A linear regression equation was calculated for each gender (Table 4) and for the genders combined (Fig. 3, Table 4). Linear regression analyses were also performed by forcing the line of regression through the origo for each gender (Table 4) and for the genders combined (Fig. 3, Table 4). The differences between the models based on gender were negligible. Forcing the line of regression through the origo resulted in the following estimated model: LVM = LVShV*1.265 g/ml, with 1.265 g/ml as the myocardial tissue density. The R² values and residual standard error are presented in Table 4. Hypertrophic hearts based on total heart weight did not alter the myocardial tissue density compared to non-hypertrophic hearts (data not shown). Using the hitherto clinically accepted myocardial tissue density of 1.055 g/ml, the anatomic LVM was underestimated by 18.1% (95% CI = 29.9–40.0 g), p < 0.0001, (Fig. 3, Table 5). The mean difference in LVShV assessment by CT with the different settings in 15 randomly chosen LV myocardial tissues was - 19.4 ml (95% CI: -65.3; 25.5), p > 0.5.

The following study limitations need to be addressed. This study included 73 non-hypertrophic hearts, but in order to develop LVM reference values based on the density, a bigger study population would be preferable. We did not take into consideration the status of ischaemic heart disease, fat infiltration or LV fibrosis and how this potentially could affect the size and shape of the left ventricle. The papillary muscles were included in the LVShV measurements. However, the impact of the inclusion or exclusion of papillary muscles on the assessment of LV function is negligible [15]. The scans were non-contrast scans, which in a clinical setting makes the differentiation between the LV blood pool and the LV myocardium difficult. However, non-contrast scans are possible in deceased individuals (cf. Fig. 1). We chose to perform the present study without contrast to avoid extravasation of contrast media in the examined deceased individuals and thereby omit a possible weight effect on the myocardium.

A study by Bai R et al. [20] has suggested a tendency of lower myocardial tissue density associated with pathological changes as oedema, none of our included cases had underwent putrefaction, as these were excluded. Therefore, we do assume that post mortal changes did not have any impact on the myocardial tissue density calculation. In vivo, the myocardium consist of intra-myocardial blood-volume. Postmortem, this volume may change, e.g., postmortem extravasation or, conversely, fluid accumulation from leaking and decomposing endocardial structures. Such changes are small and usually only become pronounced with extended postmortem intervals [35]. The deceased individuals in this study were kept at reduced temperature and autopsied rather quickly after declaration of death. Morphological observations as well as quantitative results suggest that elements of the blood are resistant to autolytic effects [36]. Overall, this leads us to assume that no significant organ volumetric changes took place. Water displacement was not performed in this study. There are several factors to take into consideration when performing water displacement measurements/techniques on cadaver hearts. Although it works accurately with solid objects, biological tissue is by its nature permeable and may be compressed or distended, and it may be fixed or unfixed. Given that we wanted to investigate CT-derived volume measures, as this is what is used clinically, we hence chose to base our volumes on this method. Boundaries are sensitive to threshold and windowing. We used a HU threshold for myocardial tissue of 50 ± 10, and 1 mm slice thickness for CT evaluations. The CT settings used for this study are optimized for contrast-based studies, and a different cardiac window settings could result in different myocardial volume. Also, different thresholds may apply for epicardial and endocardial borders and partial volume effects may be important. Due to the relatively close HU values of the blood pool and the myocardium, the myocardium may have been overestimated in some cases and underestimated in other cases.

Finally, our calculation of a new value for myocardial density is made performing a linear regression on the LVM and LVShV values of our study population. Thus, theoretically, we cannot be sure that very small hearts or LV masses below 96 g will conform to the linear model. This can probably only be investigated by also applying our method to a subadult study population. However, we note that all other analyses on myocardial tissue density were also constrained or even more limited. We would also note that when comparing with the hitherto-used value for myocardial density the differences are most pronounced for bigger hearts, so even if our new value has not been tested on very small hearts, any differences might be assumed minor.