Desktop version

Home arrow Health

  • Increase font
  • Decrease font

<<   CONTENTS   >>

Present Status

Whether Ai can be an alternative to autopsy and determine causes of death without dissection is controversial. This has led to comparisons of causes of death determined by Ai and autopsy in several countries.

Fukayama et al. [166] described the value of Ai in an investigation of deaths associated with medical treatment. That study examined the value of Ai in 165 assessments of causes of death. They judged that autopsy was not necessary in only 3% of these assessments and that the highest correlation of Ai and autopsy findings was 20%. They also proposed classifications for pathological findings determined by Ai alone, compared with the results of both Ai and autopsy findings. They defined obvious pathological findings such as aortic dissection and vague conditions such as systemic infectious disease and embolism (Table 4.3).

Roberts et al. [167] assessed 182 random adult deaths in the UK to compare the accuracy of CT Ai, MRI Ai, and gross autopsy. The major discrepancy rates between causes of death identified by imaging and autopsy were 32% for CT, 43% for MRI, and 30% for the consensus radiology report, which was created by four general radiologists based on CT and MRI images. The most common imaging errors in identification of cause of death occurred in cases of ischemic heart disease, pulmonary embolism, pneumonia, and intra-abdominal lesions, findings

Table 4.3 The classification of disease/pathological findings by the certainty of diagnostic imaging


Disease/pathological findings

Confident Ai diagnosis

Aortic dissection, aortic aneurysm, end-stage kidney disease, pleural effusion and/or ascites, interstitial pneumonia, pneumothorax

Possible Ai diagnosis

Pericardial effusion, cardiac tamponade, pneumonia/bronchitis (when there is no complication of pulmonary edema), subdural hematoma, advanced liver cirrhosis/liver fibrosis

Uncertain Ai diagnosis

Systemic infection (e.g., miliary tuberculosis), thrombosis, embolism, slight liver cirrhosis/liver fibrosis, meningitis, neurodegenerative disease, acute and old myocardial infarction, carcinoma of unknown origin, diffuse infiltrative lesion that concurred with those described by Fukayama et al. Moreover, the major discrepancy rate compared with autopsy was 16% among 88 deaths for which radiologists ruled that autopsy was unnecessary.

The 10% lower discrepancy rate for CT compared with MRI suggests that CT is more accurate than MRI in detecting causes of death. Roberts et al. also discussed an adaptive difference between CT and MRI because CT images scan bone better than MRI because it is density based, not simply because it has slightly higher spatial resolution in most cases. CT is thus effective for visualizing fractures and intracranial hemorrhages. In contrast, MRI provides greater detail of soft tissues. They suggested that forensic situations would be better served by CT, whereas nonforensic and pediatric situations should use MRI [168].

They concluded that the error rate was similar to that for clinical death certificates when radiologists provided a confident cause of death and was therefore acceptable for medicolegal purposes. However, CT and MRI frequently miss common causes of sudden death, and unless these weaknesses are addressed, systemic errors in mortality statistics will result if imaging replaces conventional autopsy.

One German study compared causes of death determined by CT and gross autopsy among 162 (57%) of 285 patients who died in nine intensive care units (ICUs) in Hamburg [159]. Among 47 (16%) autopsies that were also assessed with Ai, the main causes of death overlooked by Ai were cardiovascular disease (12.5%) and cancer (40%). In contrast, gross autopsy overlooked 13 traumatic fractures and two cases of pneumothorax. Ai alone diagnosed new findings in 11 of the remaining 115 deaths. The authors concluded that Ai might be useful for establishing diagnoses that have traditionally been identified by medical autopsy. Ai can also at least in part prove equally instructive as gross autopsy in confirming antemortem clinical diagnoses.

A prospective study of the ability of PMCT to determine causes of death at the Department of Pathology at Aachen University Hospital in Germany analyzed 29 CT studies acquired before autopsy [169]. The accuracy of PMCT for determining causes of death was 68%, with a positive predictive value (PPV) of 75%. However, the accuracy and PPV of CT were 21% and 29%, respectively, for defining pathogenic mechanisms. The authors considered that the combined diagnostic yield of autopsy and CT was excellent compared with autopsy alone. These results were similar to those of Ezawa et al. [155].

A French study compared the abilities of PMCT and autopsy to determine the causes of unexpected death among 47 infants and children [170]. Among 18 (38.3%) causes of death determined by autopsy alone, those also determined by CT agreed with 15 of them. The major discrepancies between CT and autopsy findings were associated with pulmonary analysis, and the authors concluded that the autopsy and CT findings essentially agreed. CT is noninvasive and thus more acceptable to the relatives of deceased children.

<<   CONTENTS   >>

Related topics