In the event of war, natural disaster or murder, it is of utmost importance to find and rescue survivors and identify victims. Today, specially trained police dogs are used to pick up the scent of human remains at distance, and then to locate the body. The dogs can also identify traces at crime scenes, and mark finds which have been in contact with the missing person. These dogs are often called cadaver-detection dogs or search-and-rescue dogs.
“Dogs have an extraordinary olfactory capability, which is reportedly 10 000 to 100 000 times better than that of people. If trained properly, they can be exceptionally helpful, for instance in a search. They are considered the most rapid and efficient tool for odor detection among the police community. However, there are some legal and ethical dilemmas regarding the use of the dog response as an evidence in court”, says Donatella Puglisi, The information from the sensors are processed in a miniature computer. The small size is a prerequisite for future applications. Photo credit Olov Planthaber associate professor and docent at Linköping University’s Department of Physics, Chemistry and Biology.
Machine learning
Against this backdrop, her research focuses on developing a technical solution that can complement the search dogs for rapid location and identification of human remains in disaster and crime scenes. The solution comes in the form of a portable gas sensor system, i.e. an electronic nose, which can be trained through machine learning to identify human remains. The reasons for wanting a technical solution are many. But there are three main themes:
First, finds made by cadaver dogs are considered a dubious evidence in court. So if the dog marks a find by a suspected perpetrator, it is still not possible to link the find to the suspect without additional evidence. Second, the dog’s well-being must be considered. It is ethically questionable to send a dog into a collapsed mine, where there is a risk of hazardous gases, thus putting the dog at risk. Donatella Puglisi, associate professor and docent at IFM. Photo credit Olov Planthaber And third, it is very difficult to train dogs to find human remains. There is a lack of bodies donated for this purpose. Also, the training takes a very long time, which means that many bodies are needed. This has led to a shortage of cadaver dogs in Sweden.
“Society has a genuine need for our research. We received a high interest from the Swedish police to develop an innovative technical solution, scientifically robust, that could serve as a powerful complementary tool to cadaver-detection dogs”, says Donatella Puglisi.
Forensic medicine
The research project is interdisciplinary, and is being conducted in collaboration with the Swedish National Board of Forensic Medicine. Kerstin Montelius is a forensic geneticist at the National Board of Forensic Medicine in Linköping, where much of the work is devoted to identifying human remains, for instance after accidents. She sees potential for this research:
“In the event of a major accident, you need to find the human remains quickly, and identify the deceased. This is a very important job. The gas sensor that the research project is developing will make the process easier for us, who the samples are sent to”, says Kerstin Montelius.
The idea of developing a technical solution for finding human remains was raised during the media coverage of the fire in London’s Grenfell Tower in 2017. A fire broke out in the high-rsie Grenfell Tower in London in 2017. Seventy-two people lost their lives. Photo credit Christopher Walls Seventy-two people lost their lives, and the search was hampered by the powerful fire.
“In a situation like that it would have been possible to use drones equipped with gas sensors, which could have searched through the building and found the bodies in a safe manner.”
Monumental task
Unlike a dog which can be affected by the surroundings, miss things or just have a bad day, gas sensors are consistent. The courts get an independent instrument whose finds could be considered proof in a court of law. But a biological system based on millions of years of evolution is hard to emulate.
Dogs have more than 200 million odour receptors in their nose, and their sense of smell takes up about one-third of their brain. Humans have about 5 million odour receptors, and our sense of smell takes up just 5 percent of our brain. With this in mind, one can imagine the enormity of the dog’s olfactory world. In other words, digitally emulating the dog’s sense of smell is a massive challenge.
The smell of decay
A decaying corpse consists of many different tissues that smell differently, and the smell varies depending on the degree of decay. The olfactory capability is, reportedly 10 000 to 100 000 times better than that of people. Photo credit iStock For the gas sensor to determine whether the remains are human and not from some other animal, every odour must have a “fingerprint” that can be identified and categorised. This can be done using gas chromatography – mass spectrometry, or GC-MS.
“Currently, there are no portable detection systems capable to outperform the speed, versatility, ruggedness, and discriminating power of dogs. With our research, we aim to develop an air analysis device based on smart sensors and data engineering that may effectively help the searching and overcome the legal and ethical constraints around dog training”, says Donatella Puglisi.
Researchers involved at LiU are Donatella Puglisi, Guillem Domènech-Gil and the graduated student Johanna Eriksson.
The portable gas sensor system can be trained through machine learning to identify human remains. Photo credit Olov Planthaber
