Segmentation of the lungs using post-mortem computed tomography (PMCT) data was so far not feasible due to post-mortem changes such as internal livores. Recently, an Osirix plug-in has been developed allowing automatically segmenting lungs also in PMCT data. The aim of this study was to investigate if the Hounsfield unit (HU) profiles obtained in PMCT data of the segmented lung tissue present with specific behaviour in relation to the cause of death. In 105 PMCT data sets of forensic cases, the entire lung volumes were segmented using the Mia Lite plug-in on Osirix. HU profiles of the lungs were generated and correlated to cause of death groups as assessed after forensic autopsy (cardiac death, fatal haemorrhage, craniocerebral injury, intoxication, drowning, hypothermia, hanging and suffocation). Especially cardiac death cases, intoxication cases, fatal haemorrhage cases and hypothermia cases showed very specific HU profiles. In drowning, the profiles showed two different behaviours representing wet and dry drowning. HU profiles rather varied in craniocerebral injury cases, hanging cases as well as in suffocation cases. HU profiles of the lungs segmented from PMCT data may support the cause of death diagnosis as they represent specific morphological changes in the lungs such as oedema, congestion or blood loss. Especially in cardiac death, intoxication, fatal haemorrhage, hypothermia and drowning cases, HU profiles may be very supportive for the forensic pathologist.
Keywords: Automatic segmentation; Cause of death; HU profile; Lungs; Postmortem CT.