AI determines time of death more accurately than standard methods used today.
When a body is found, one of the first questions for investigators is when exactly death occurred. Verifying alibis and reconstructing the chain of events depends on this. The problem is that classical forensic methods only work well within the first few days. Now, research by scientists from Linköping University shows that artificial intelligence can determine the time of death significantly more accurately by analyzing chemical changes in the blood.
What Methods Are Used to Determine Time of Death and Why They Fail
Today, forensic experts determine the time since death using several indicators. The most common methods include:
- body temperature
- rigor mortis
- potassium levels in the vitreous humor of the eye
The last method is considered the “gold standard” of forensic medicine. But it has a serious drawback: accuracy drops sharply after just 48 hours. After that, the margin of error can reach several days.
For investigations, this is critical. Sometimes a difference of just one or two days can confirm or disprove an alibi.
Modern standard methods provide reliable results only within one to three days after death.
How AI Determines Time of Death Through Blood Analysis
A team led by Rasmus Magnusson trained an artificial intelligence model to analyze chemical changes in blood after death.
For training, they used 4,876 blood samples collected during routine forensic toxicological examinations.
After death, predictable processes begin in the body:
- cells stop receiving oxygen
- proteins break down into amino acids
- fats decompose and alter the structure of cell membranes
These processes leave measurable chemical traces — metabolites. The neural network analyzes thousands of such signals and correlates them with the known time of death.
As a result, the model was able to determine the time since death with an average error of about 1.45 days, even at intervals of up to 67 days after death.
Although postmortem intervals in the dataset ranged from 1 to 67 days, 97% of cases fell within the first 13 days, and the model’s effectiveness was primarily evaluated on this data.
Why AI Could Become a New Method for Determining Time of Death
Models that work well in one laboratory often fail in another. But this model held up.
Scientists validated the model on an independent set of 512 cases collected in a different year and on different laboratory equipment. Even under these conditions, the margin of error was about 1.78 days.
The method has another important advantage: the technology can work in regular forensic laboratories.
The study was conducted using material from Sweden, where bodies are typically refrigerated within 48 hours; the authors specifically note that climate, storage conditions, and temperature may affect the method’s transferability to other countries.
But it is already clear: artificial intelligence can become one of the most accurate tools for determining time of death in modern forensics.
