The study was based on a retrospective analysis of autopsy reports and PMCT scans of individuals admitted by the Danish police for autopsy at the Department of Forensic Medicine at Aarhus University, over a four-year period from August 2019 to August 2023. Inclusion criteria were asphyxia caused by manual strangulation (MS), ligature strangulation (LS), smothering (SM), choking (CH) and hanging (HA). Cases with aspiration were excluded because fluid in the airways has the potential to affect radiodensity on PMCT. Furthermore, cases with position-dependent asphyxia and compression of thorax were also not included due to conceptual difficulties.

The following exclusion criteria were applied to both the case and control groups: Postmortem Interval (PMI) > 5 days, penetrating injury to the chest, prior or current lung disease, age above 80 and below 16 years, and severe putrefaction. The level of putrefaction was addressed in the autopsy report by the forensic pathologist who had graded putrefaction as “none”, “mild”, “moderate” or “severe”, based on visual inspection and findings.

Specific causes of death in the control group expected to severely affect lung density were excluded: Exsanguination, burn victims, cold exposure victims, opioid intoxication and death following hospitalization or intensive care treatment [9,10,11]. The control group was matched in terms of age, height, weight and PMI.

All PMCT-scans were acquired using a Canon Aquilion Prime SP system (Canon Medical Systems Europe B.V., The Netherlands). Scan parameters were 120 kV and 320 mAs, 1.0 mm slice thickness, and the projections were reconstructed with a “hard” reconstruction kernel (FC08), as recommended for lung densitometry [18]. All scans were conducted with the arms of the body placed by the side of the torso and performed as part of routine pre-autopsy examination at the department. Data analysis was conducted using the “Lung Density Analysis” plugin on a Vitrea Workstation (Canon Medical Systems Europe B.V., The Netherlands).

The endpoints for this study were chosen based on previous consensus from pulmonary emphysema studies in the living [19,20,21]. Since there is no current agreement on the best lung densitometry endpoint, three of the most commonly applied measures were chosen. Perc15 is defined as the HU value below which 15% of the lowest density voxels are distributed. The value can be presented either as an absolute HU (Perc15) or by adding 1000 to the HU to get positive voxel-values expressed as Pulmonary Density (PD15) values in in g/L [22] (Fig. 1). For readability and coherency Perc15 was used for this study.

Graph illustrating the calculation of LAA and Perc15 from cumulative frequency of HU. LAA values are the percentage on the y-axis that corresponds to values of -950 HU and − 910 HU on the x-axis. Perc15 is obtained as the HU value corresponding to the 15th percentile point on the y-axis

LAA is defined as the percentage of voxels in the lung with HU’s below cutoff values of either − 950 HU or -910 HU (Fig. 1). To extract the values from the PMCT-data, a semi-automatic segmentation of the lungs and airways was conducted (Fig. 2A). Each segmentation was validated and adjusted by a trained medical doctor with previous forensic and radiological experience (LS).

After segmentation both LAAs and Perc15 was obtained from voxel frequency distributions (histogram, see Fig. 2B).

(A) A volume reconstruction of the thorax from an asphyxia case (hanging), with the red area illustrating low attenuation area (LAA-910). (B) Histogram displaying the voxel distribution of the same case (in A) displayed using the Vitrea “Lung Density Analysis” plugin

Normality was checked using visual inspection of QQ-plots and histograms. Parametric data were presented with mean and standard deviation (SD) while nonparametric data were presented with median and interquartile ranges (IQR). Statistical comparisons were conducted using Student’s t-test for parametric data and Wilcoxon Rank Sum test for nonparametric data. All statistics were performed using RStudio (RStudio Team (2020). RStudio: Integrated Development for R. RStudio, PBC, Boston, MA).