Australia’s Monash University is digitally analyzing the past to create a
futuristic approach to solving gun crimes
Confronting gun violence
and the massive toll it takes on a daily basis remains front-and-center in
public life. Mass shootings may grab the most headlines, but law enforcement
knows it is the steady stream of individual cases that contribute to the United
States’ disproportionately high rate of gun violence. Even when there are no
fatalities, each incident of gun violence is a tragedy that has devastating
consequences and changes life forever for those caught in its aftermath.
Fortunately, there is a
consensus that employing smart technology to improve the capabilities of law
enforcement is a common-sense approach everyone can agree on. A study by the
United States Sentencing Commission found that gun offenders are statistically
the most likely to recommit crimes, with 45.5 percent of firearms offenders being
re-convicted versus only 27.6 percent for other offenders. This increases the
urgency of bringing individual firearms offenders to justice.
While roughly 20 percent
of gun murders are solved within 24 hours, the overall likelihood of solving a
firearm homicide sits at just 46 percent, significantly lower than the 75
percent rate for non-firearm homicides. Even more disturbing is evidence
showing that the chances of solving a shooting is actually declining. Equipping
investigators with the tools they need to solve cases quickly not only
increases their likelihood of success but also helps them take dangerous
suspects off the street before they offend again.
Around much of the
country, police departments are overburdened, their capabilities limited by
budget cuts and ever-expanding mandates. To make headway in the fight against
crime, it is more necessary than ever before to make sure that law enforcement
works smarter, not just harder. New digital tools represent a unique
opportunity for police to stay one step ahead and make real progress in
securing communities nationwide.
While TV and film may
glamorize old-fashioned police work, today’s investigators know that innovative
scientific approaches can make or break a case. Just as DNA testing made a
rapid leap from the university lab to the crime lab, the convergence of 3D
modeling, machine learning, and augmented reality has the potential to take
cutting-edge science and apply it in the real world. A team from Monash
University – Australia’s largest research university – is at the forefront of
By analyzing and
compiling a subset of over 75,000 postmortem computed tomography (CT) scans
from the Victorian Institute of Forensic Medicine (VIFM), the Monash team is
working to apply machine learning to create a digital 3D model of human anatomy
that is capable of identifying entry and exit wounds in shooting victims,
giving law enforcement instant insight that can be used in their investigation.
By recording the trajectory of the projectile through the body (identifying and
localizing projectile fragments in the process) the scan can determine key
information, such as whether the wounds were self-inflicted.
The VIFM includes Monash
University’s Department of Forensic Medicine and performs autopsy services for
all deaths reported to the Victorian State Coroner. With
an archive that increases by 7,000 cases a year, all causes of death are
represented in the comprehensive database, including traumatic injury, homicide,
and suicide. The unrivaled scope of the project means that all ages, ethnic
groups, and genders are represented within the archive, making it an extremely
useful tool moving forward.
The size and scope of
the database does present a drawback to those seeking to harness the full
extent of its insights. Containing more than two million digital photographic
images, including external and internal injuries, pathological conditions and
external whole-body images, means that comprehensive analysis of its entire
content is well beyond human capabilities. Machine learning, however, has the
potential to analyze and catalog these images in a fraction of the time that it
would take human experts.
This is a massive step
forward in a field that has employed the same observer-dependent basic analytic
techniques for much of the last 100 years.
Before the advent of CT
imaging, x-rays were used to produce a 2D view of the subject, which made
localizing projectiles and fragments difficult without conducting an internal
examination. Typically, the trajectory of foreign objects is determined using
long probes to assess a projectile’s path. Another significant drawback is that
current imaging techniques can’t differentiate between bullet fragments and
other foreign metal objects, such as pacemakers or dental fillings.
Monash Forensic researchers,
working in conjunction with Monash Faculty of IT experts in AI, are making 3D
digital reconstructions of shooting victims. These reconstructions allow investigators
to view victims along multiple planes and from different vantage points, using
advanced computer graphics and augmented reality. Advanced machine learning can
then be applied to determine trajectory and projectile fragmentation, resulting
in a 3D-printed model that can potentially be submitted as evidence in court.
Medical practitioners currently visualize 3D
data scans on 2D computer screens to investigate disease, injury, or to plan an
intervention. While augmented reality (AR) technology has allowed practitioners
to better visualize and interact with 3D data, it remains challenging and is
not always precise. Monash University researchers have developed the 3D_Slicer,
a patented prototype that allows users to visualize and interact with 3D
medical images, such as CT or MRI scans, with precision, and provide a true 3D
representation of a body or organ scan.
When the scans are run
thorough machine-learning algorithms, researchers will be able to recognize repeating
patterns from observing what happens when different types of bullets strike the
body. This will make it possible for investigators to understand the type of
weapon that was used, even if no bullet fragments are recovered and if there is
no evidence found at the crime scene.
With this knowledge, law
enforcement will eventually be able to compare the weapon and ammunition used
with other crimes and also run results through a weapons database that will
help them to quickly identify potential subjects.
Once fully realized, the
scan will also be able to determine how the gun was held as it was fired, the
range from which the gun was fired, and even the height of the shooter – all of
which could expedite autopsies and help investigative efforts.
Despite underpinnings of
scientific rigor, modern forensic evidence can run into problems in the
courtroom since its interpretation often hinges on the subjective opinions of
experts. Recent studies have called into question many of the long-established
assumptions that had lent credibility to forensic techniques, such as the
analysis of hair samples, bite marks, and blood spatter.
opens the door to legal attacks on the objectivity and validity of the evidence
being presented. Eliminating subjectivity also eliminates potential biases.
Automated localization of the femur. Green represents the machine-learning
interpretation. The blue line is the ground truth determined independently by
three trained image-analysis specialists. Video credit: Carlos Pena-Solorzano
& Matthew Dimmock | VIFM
This multi-disciplinary team
of scientists at Monash University is working to outsource analysis from the
subjective human brain to advanced machine learning algorithms that will allow
investigators to approach cases with greater objectivity and neutrality.
Gaining a better
understanding of exactly how bullets interact with the tissue they come into
contact with may also expand the range of applications for this technology. For
example, team members envision the 3D scan as a useful aid in military
frontline care. Leidos — a leading international IT and engineering company — is
helping to fund the work through the Monash Institute of Medical
Engineering, with ultimate military health applications in mind.
When fully developed, this
scan alone will be able to reduce the need for comprehensive autopsies,
allowing police to better allocate resources by differentiating between murders
and suicides while also boosting the veracity of evidence used in court. By
providing instant insight into the type of weapon used and an analysis of the
bullet’s trajectory, law enforcement will be immediately equipped with
everything that they need to begin their investigation on the right foot.
Gun violence shows
humanity at its lowest, but the scientific ideal that keep us pushing for
progress and knowledge show us at our best. Integrating scientific approaches
into crime investigations has yielded spectacular results and has helped bring
justice in cases that would have otherwise gone cold long ago. The 3D scanning
technology will continue this legacy.
The togetherness, unity,
and solidarity amid gun tragedies offers a blueprint for how Americans can
unite to solve this pressing issue. Innovative scientific breakthroughs like
the 3D scan from Monash University won’t end crime but they will go a long way
in making communities safer and more secure.
Dr. Chris Bain is the inaugural Professor of Practice in Digital Health at Monash
University in the Faculty of Information Technology. He has more than 25 years
of experience in the health industry, including 12 in clinical medicine. He’s
led numerous software development and implementations projects in the clinical
and management support areas, resulting in a range of prototype and fully
implemented systems. A large part of his work involves leading the university's
initiatives in digital health, working in collaboration with other faculties.