One of the stories that will haunt me forever is Gary Paulsen’s Newbery Award winning fiction book, Hatchet. In Hatchet, Brian is on his way to Canada in a two-seat prop plane with a taciturn pilot, and suddenly he smells flatulence. The pilot rubs his underarms and says, “Nothing is wrong.” The smells grows.
The pilot harrumphs and says, ”Must have been something I ate.” More bodily fluids, more smell. The worst smell Brian ever smelt. The 10-year-old boy turns away so as to not embarrass the pilot. The pilot reaches for the transmitter, and suddenly he’s dead. A heart attack. Would I know someone’s having a heart attack if they suddenly released the most darned awful rotten egg smelling stench?
Not from all the research I’ve done on heart attacks. Rather, gastrointestinal problems like ulcers, muscle spasms in the esophagus, a gallbladder attack, and pancreatitis – all of which make people smell like ghastly rotten eggs – feel like heart attacks. However, they’re actually just heartburn, and they don’t kill you.
AI That Sniffs Out Disease
Most diseases don’t exude any warning smells, whatsoever. No one including the victim know they’re ill, or that they’re going to be ill. This may change in the near future, however, as scientists are working to create different artificial intelligence devices imbued with prenaturally sharp “noses” capable of sniffing out danger.
Last week, James Gathany of Loughborough University told The Conversation (a magazine that interviews PhD researchers or academics), that he and his team were using artificial intelligence to analyze the chemical compounds in breath samples. A team of doctors, nurses, radiographers and medical physicists at the Edinburgh Cancer Centre collected breath samples from participants undergoing cancer treatment.
The samples were then analysed by two teams of chemists and computer scientists. Once a number of compounds were identified manually by the chemists, extremely powerful computers were fed that data. The computers used deep learning networks that were specifically engineered to “read” the traces left by odors.
The deep learning networks learned more and more from each breath sample, until they could recognise specific patterns that revealed specific compounds in the breath. In their study, Gathany and his team focused on a group of chemicals called aldehydes which have a sweet and sometimes pungent smell.
The odors of this chemical group range. Some aldehydes smell like vanilla and like almond. Others have the sharp scent of blue cheese, or the musky smell of the Himalayan musk deer, according to Chemistry LibreTexts. Smell one of those odors from someone and you know they’re stressed or ill.
The History of Disease-Detecting AI
Scientists have used AI to detect disease for a while. In 2013, the Quad-City Times rounded up the most recent relevant research in the field. Researchers in Philadelphia had recently tested an experimental device that sniffed out melanoma from the air around the skin. In Cleveland, researchers produced chemical sensors for acetone, a first step toward a breathalyzer for diabetes.
Other US researchers used trained dogs to sniff out prostate cancer and melanoma from organic compounds in urine samples and directly from the skin. Finally, in other American states, scientists dabbled with sniffing brain tests to detect lung cancer. Commonly, it takes regular doctors several hours to diagnose illness from sniffing odors. And even then they may be incorrect.
These wondrous devices smell in minutes. “Effectively, AI is making the whole process cheaper – but above all it is making it more reliable,” James Gathany wrote in his article for The Conversation. “Even more interestingly, this intelligent software acquires knowledge and improves over time as it analyses more samples.”
Smells of 17 diseases from 1,400 patients
In 2011, Professor Hossam Haick of the Technion-Israel Institute of Technology led an international team of 56 researchers in five countries. They confirmed that different diseases have different “chemical signatures” which you can “tear apart” in breath samples. A press release narrated how the doctor and his team studied more than 1,400 patients.
The study included the following 17 diseases, all different and unrelated from one another:
- Lung cancer
- Colorectal cancer
- Head cancer
- Neck cancer
- Ovarian cancer
- Bladder cancer
- Prostate cancer
- Kidney cancer
- Stomach cancer
- Crohn’s disease
- Ulcerative colitis
- Irritable bowel syndrome
- Parkinson’s disease (two types)
- Multiple sclerosis
- Pulmonary hypertension
- Chronic kidney disease
Professor Haick then collected samples from 14 departments at nine medical centers in five countries: Israel, France, the USA, Latvia, and China. The expert on non-invasive disease diagnosis concluded that each disease had its own smell.
What makes AI-sniffing devices different from regular doctors is that now anyone who has such a device can sniff disease, know for a certainty they or someone else are ill, and do so in minutes. Not only that but each illness carries a particular odor so you’ll know exactly what the disease is. Using that knowledge you can help yourself, or someone else, long before they fall dead at your feet. Or in the pilot’s seat as in Brian’s case.