Establishing when a person died based on their decomposing body is a subjective problem with dozens of variables, so any two pathologists can arrive at different estimates. Now, researchers are developing an artificial intelligence model in an attempt to bring some objective replicability into the process.
and at Clemson University in South Carolina are two of the scientists working on , a tool that is partly a database of information on previously discovered bodies and partly an AI tool that takes in details on new cases and provides an accurate time of death.
Smith, who works听on the technological side of the project, says he arrived with the expectation from films and television that pathologists had a rich set of scientific methods to pinpoint the time of death, but soon realised that wasn鈥檛 the case. 鈥淚 was just really shocked to see that that鈥檚 not the state of affairs,鈥� he says.
Advertisement
鈥淚 would say the existing methods are very poor,鈥� says Weisensee, a forensics anthropologist听who does consultancy work with police forces. She believes that using AI and the data in geoFOR can take predictions 鈥渂eyond gut feeling鈥� into something more replicable. But she stresses that it won鈥檛 be used in a vacuum and its results should always be checked by humans.
鈥淚t鈥檚 not the end-all, be-all answer and I don鈥檛 think we鈥檙e ever going to come up with anything like that,鈥� says Weisensee. 鈥淏ut these are all tools that can be used for feeling more confident in our results, improving our results and making them more accessible to people around the world.鈥�
So far, more than 3200 cases, mostly from the US, have been entered into geoFOR. The researchers say that each additional one makes the model more accurate and adaptable. More than 200 people have signed up to record previous case data and the researchers are in conversations with users in Africa, Asia and Europe.
To use geoFOR, people need to answer a series of questions, such as the body鈥檚 precise location, its estimated size, details on whether it was found indoors or outdoors and in what condition, and whether it was buried or left open to the air. There are a range of simple medical questions too, such as whether the cornea are clouded, whether the fingertips are dry and shrivelled, whether the abdomen is bloated, whether the skin has a green tinge and whether any bodily fluids have leaked. The presence of any maggots, flies or beetles are also noted.
That information is then analysed by geoFOR and used to determine what forensic experts call postmortem interval, or the amount of time since death occurred. The researchers say the model gives an R2 value 鈥� a measure of how well the results reflect reality, where 0 is worst and 1 is best 鈥� of 0.8.
at the University of Kent in the UK says that existing methods to determine the length of time since death are imprecise, and become even more so as that time period increases. 鈥淚 think [AI] will be able to crunch the variables more easily than we can at the minute,鈥� says Green. 鈥淚 think it鈥檚 a good thing for us to actually pursue this type of research because we can hopefully pin that post-mortem interval down more accurately.鈥�
arXiv
