A medical examiner puts on a pair of augmented-reality goggles and brings up a computer simulated image of a body that appears to hover steps away. Nearby on a metal autopsy table lies the body of a person brought into the lab after a fatal shooting. Instead of cutting into the victim, the examiner slices through the 3-D image, mapping the bullet’s trajectory and determining the cause of death without making a single incision.

This is one vision of the virtual future of autopsies, based on interviews with forensic and digital health-care experts: Using digital reconstructions and machine-learning algorithms to diagnose the cause of death, identify a victim, and even triage battlefield or motor-vehicle injuries in live patients by analyzing images of victims who died in similar incidents. It would mark a step change for the field of forensic science, where the standard methods of autopsy have remained nearly unchanged for a century.

Most medical examiners still cut into corpses to search for clues about the cause of death. The most common technology they use, the X-ray, was discovered in 1895 and produces a two-dimensional view that makes it hard to find evidence without an internal examination. Some researchers are on a mission to change that, seeking to put forensic medicine on the same high-tech path as clinical health care, where algorithms are being trained on millions of medical images to help with diagnostics, spotting subtle patterns easily missed by humans.

“The vision is to advance the science of forensic medicine using much more computational power and far fewer archaic, human methods,” says Chris Bain, a digital health professor at Monash University in Melbourne, Australia.

A prototype developed by researchers in Australia aims to create a 3-D image, eventually of an entire human body, that medical examiners can view and investigate using virtual reality goggles.



Photo:

Victorian Institute of Forensic Medicine

None of this would be possible without computed tomography, or CT scans that use rotating X-ray machines to create cross-sectional images of the body. CT scans are common in clinical practice, but still relatively rare in forensic pathology, especially in the U.S. where the medical examiner system is fragmented and some states have coroners who aren’t medically trained.

One global reference point for forensic imaging is the Victorian Institute of Forensic Medicine in southeastern Australia, which has built up a database of some 80,000 CT scans representing deaths ranging from traumatic injuries to homicide and suicide. Experts there are turning to machine learning to put millions of images to future use, from providing airtight evidence in criminal cases to quicker identification of victims of mass disasters.

The institute was founded in the 1980s following the notorious case of Lindy Chamberlain, who said that a dingo had made off with her baby. She was found guilty of murder but later acquitted amid criticism of the forensic evidence.

SHARE YOUR THOUGHTS

How do you think digital autopsies might change criminal investigations? Join the conversation below.

Today, the institute is one of a handful of expert centers around the world called in to help with disasters such as the 2004 Indian Ocean tsunami. It has been scanning deceased patients since the early 2000s, and its database is now one of the largest globally. Those images have reduced the autopsies the institute has performed to around 48% of cases now from 80% in 2005.

Researchers from the institute and Melbourne’s Monash University are working with the U.S. biomedical and defense company

Leidos Holdings Inc.

on one method for an incision-less forensic investigation. It involves creating a 3-D digital reconstruction of a shooting victim that they can slice in multiple planes and directions using advanced computer graphics, including augmented reality.

Computer algorithms can then help determine the bullet’s trajectory, find bullet fragments, and create a 3-D-printed model that can potentially be used as evidence in a courtroom.

“In my world, it’s about making a jury understand. In a clinical world, it’s about making a patient understand. Visualization is always much better than words,” says Richard Bassed, deputy director of the Victorian Institute.

Craniofacial reconstruction, used in forensics for identification purposes, is typically done by molding clay around skulls. Machine-learning could change that.



Photo:

Victorian Institute of Forensic Medicine

Another group of researchers is working on a prototype device that projects a 3-D image of a human head that can be viewed and manipulated using virtual reality goggles. Images from medical-imaging technology are typically viewed on 2-D computer screens. The goal is to eventually create a 3-D image of an entire body, paving the way for virtual autopsies.

Another area ripe for new technology is postmortem facial reconstruction, used for identification purposes, which has traditionally been carried out by putting clay on skulls in a process known as forensic art. “No one knows if it’s accurate or not. There has never been a way to put science around it,” says Dr. Bassed, a forensic dentist who spent months identifying the victims of the 2004 tsunami using dental records, DNA and fingerprints.

See More From The Future of Everything

Researchers are aiming to teach a machine to reconstruct a face using images of deceased people to learn the patterns between the nose, the muscles and the bones. The image database could also be used to develop a facial recognition tool to identify mass fatalities from a license photo. The field is still in its early stages, but experts say the dead can potentially offer lessons for the living.

The New Mexico Decedent Image Database, a forensic initiative based at the University of New Mexico, took more than 15,000 scans of dead people created by the state medical examiner between 2010 and 2017 and collated them with metadata on peoples’ life and death. The data set was made public in February and the organization has already had more than 250 inquiries from researchers world-wide. One researcher in France requested forensic data on knee fractures to help improve auto safety in vehicle crashes.

“We now have a big enough data set that [researchers] can really make some hay with it,” says Natalie Adolphi, head of forensic imaging at the University of New Mexico. A key to developing AI systems is having enough images to train and then validate the algorithms, she says.

Forensic scientists in New Mexico use special software that creates 3-D renderings of postmortem CT scans. The reconstructions can be used instead of autopsy photos in courtrooms to illustrate injuries.



Photo:

Natalie Adolphi

In New Mexico, medical examiners recently enhanced CT scans of a large toolbox filled with concrete to locate bullet fragments and the partial remains of a human body inside the box, gaining new insight into a crime. They say the evidence would have been almost impossible to retrieve without imaging technology.

They are also using image-viewing software to create 3-D renderings of postmortem CT images, which are starting to be used in medical and legal investigations, and may also increasingly replace gory autopsy photos shown in courtrooms.

The 3-D reconstructions allow measurements that are as accurate as directly measuring bone and tissue in a body, researchers say. The technology will improve understanding of human variation, perhaps exponentially, because it’s easier for researchers to access the university’s data set than human cadavers, says Heather Edgar, a forensic anthropologist with the New Mexico Office of the Medical Investigator.

Write to Rachel Pannett at rachel.pannett@wsj.com

Copyright ©2020 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Source link