In the medical arena, dogs have shown an ability to detect some cancers at their earliest stages, when the possibility for cure is highest. Some studies have documented the ability of dogs to distinguish people with cancers from healthy controls through sniffing their breath or skin with an accuracy between 88% and 97%.(2)
"The dog's nose is the gold standard for chemical trace detection," explains Brent Craven, a researcher with the Gas Dynamics Lab and the Applied Research Lab at Penn State University. Craven and his team, which includes Drs. Gary Settles and Eric Paterson of the Mechanical Engineering Department, are using computer models to study the canine sense of smell to help develop 'artificial sniffer' technologies.
"We are trying to figure out what specifically makes dogs so efficient in this area," says Craven. He has been analyzing the internal aerodynamics and transport phenomena in the canine nose using computational fluid dynamics (CFD). Large parallel simulations on high-performance computers at Penn State with AcuSolve (ACUSIM Software, Inc.) CFD software yield terabytes of data. One challenge has been the inability of visualization software to handle the very large models required for this study.
"The complexity of the dog's nose rivals the human lung with its many different branching airways, and it's essential that we understand exactly how it functions," says Craven. "Constructing a 3D virtual computer model and running simulations with such a complicated model can require nearly 100 million cells – not all CFD visualization software can handle such a large data set."
To meet that challenge, Craven and other researchers at Penn State turned to EnSight Gold scientific visualization software from Computational Engineering International, Inc. (Apex, NC). EnSight Gold supports very large models that contain millions or billions of nodes, providing intensive parallel processing and rendering capabilities for applications such as airflow animation.
"EnSight offers unique rendering capabilities and can handle very large models, that's why we are using it here in the lab," Craven says. "Furthermore, the capability to fully automate data analysis and visualization in EnSight by writing script programs in the popular programming language Python is very powerful."
His artificial sniffer project began three years ago with high-resolution MRI scans of a cadaver dog. Craven then reconstructed those into a virtual computer model of a canine nose and, using experimental sniffing data from live dogs, was able to simulate the airflow through it. The model can steadily inspire, expire, or sniff so Craven can analyze the fluid dynamics of trace detection in nature’s best sniffer.
"One of the questions we are trying to answer is why dogs sniff so fast," says Craven. Using the EnSight software, they are able to analyze simulation results of their model of a Labrador retriever sniffing at 5 Hz (5 sniffs per second), the nominal sniff frequency of a dog that size. The eventual goal is to create a complete virtual dog’s nose, complete with virtual scent receptors that are located in the back of the nose.
The computer model is able to show the role of the various anatomical structures of the canine nose in respiration and olfaction. These include the nasal vestibule, which filters, distributes inspired air within the nasal cavity, and directs an expired air jet; the respiratory airways, which filter, warm/cool, humidify inspired air, and dehumidify expired air; and the olfactory airways, where chemical trace detection occurs.
Craven credits a multi-disciplinary team involving several universities for the success they have had, including anatomy and physiology faculty Drs. Ed Morrison and Eleanor Josephson from the University of Auburn, bioengineering professor Andrew Webb and Dr. Thomas Neuberger at Penn State, and mechanical engineering faculty Drs. Gary Settles and Eric Paterson. "The scope of the sniffer project encompasses many areas of science and engineering, from anatomy, biology, and even neuroscience to bioengineering, mechanical engineering, and chemical engineering," he explains. "Our modeling and rendering techniques are drawing the interest of researchers in many of these different fields."
Craven and his colleagues recently published some of their results in the November 2007 issue of the Anatomical Record in an article titled “Reconstruction and Morphometric Analysis of the Nasal Airway of the Dog (Canis familiaris) and Implications Regarding Olfactory Airflow." Their future efforts are focused on incorporating nostril motion, vapor and particle deposition, and receptor models into their virtual dog’s nose. Experimental validation (MRI velocimetry) of these efforts is also under way.
The ultimate goal of creating an efficient artificial sniffer is not to replace dogs, says Craven, but to complement them in a broad range of applications such as cancer detection, explosives detection, search and rescue, drug interdiction, as well as military applications.
In fact, the Defense Advanced Research Projects Agency is encouraging research organizations with expertise to participate in its new "RealNose" program, designed to build an artificial nose based on the actual olfactory receptors of a real dog that can smell a wide range of chemicals with the same accuracy and reliability.
"There is tremendous potential for artificial sniffer technology. They can be handheld, like 'dustbusters', or they can be mounted on robots for mobility to go out and search for trace chemicals," Craven says. "As engineers, we can certainly design an artificial nose, but advances in simulation and visualization technology are now making it possible for us to analyze the real thing and to create a true biomimetic design."
(1) "Naturalistic quantification of canine olfactory sensitivity," Sensory Research Institute, Florida State University, Tallahassee, FL, published online 6 September 2005 at www.sciencedirect.com
(2) SAGE Publications (2006, January 6). Can Dogs Smell Cancer? ScienceDaily
PSU CFD model of canine expiration.