Scientists Invent An AI That Can Smell 17 Diseases From Your Breath, Including Cancers

Identifying a color is a easy task, shine a wavelength of 620 nano-meters and most people will say it looks red. But when it comes to figuring out how exactly a particular molecule will smell, complexity get increases to a next level, both for humans and machines too. The one reason it is so hard to identify is that there is no intermediate carrier of smell, like for sound we have vibration in air.

Now, a teams of scientists have developed a deep learning method that smells compounds in the human breath and detect illnesses. The AI can detect 17 diseases, including 8 different types of cancer, just from your breath. This AI is cable of recognizing chemical signatures of the diseases by their chemical structure.

It is created by international team of researchers from 5 countries universities -Loughborough University, Western General Hospital, the University of Edinburgh, and the Edinburgh Cancer Centre in the United Kingdom. The project was led by Professor Hassam Haick of the Technion-Israel Institute of Technology. The team is now developing Sniffphone for disease detection right through your smartphone.




Scientists Invent An AI That Can Smell 17 Diseases From Your Breath, Including Cancers
The GC-MS can analyse the air to discover thousands of different molecules (Shutterstock)

“The sense of smell is used by animals and even plants to identify hundreds of different substances that float in the air. But compared to that of other animals, the human sense of smell is far less developed and certainly not used to carry out daily activities,” researcher Andrea Soltoggio wrote on Smithsonian.com. “For this reason, humans aren’t particularly aware of the richness of information that can be transmitted through the air, and can be perceived by a highly sensitive olfactory system.”

This devise carries a sensor called nano-array which is controlled by AI software. It is made up of carbon nanotubes and tiny gold particles. It’s deep learning algorithm is capable of analyzing the human breath samples form the special chemical signatures that correspond to various diseases.

This system works because an average people exhale over a 100 unique chemicals called volatile organic compounds (VOCs). The team proved that each disease has a very specific chemical signature within a person’s VOC. The scientists used mass spectrometry to figure out a 13-component “breathprint” for each of the 17 diseases in the study.

Scientists Invent An AI That Can Smell 17 Diseases From Your Breath, Including Cancers
Simple representation of the process: from compounds in the air or breath samples to the visualisation of the detected substances.

Team first collected the breath samples from participants undergoing cancer treatment. The samples were then analysed by two teams of chemists and computer scientists. After the number of compounds were identified manually by the chemists, the data then passed to deep learning networks, so they can get train.

With every breath sample, AI start to learn more efficiently, soon it start to recognize specific patterns that revealed specific compounds in the breath. To increase the neural network’s efficiency, the team increased the original training data by using data augmentation. The convolutional neural network dataset was augmented 100 times.

Scientist uses NVIDIA Tesla GPUs and the cuDNN-accelerated Keras, and TensorFlow deep learning framework to process multiple different pieces of information at the same time.




We found that just as we each have a unique fingerprint, each of the diseases we studied has an unique breath print, a ‘signature’ of chemical components,” said Professor Haick. “We have a device which can discriminate between them, which is elegant and affordable.”

Breath is an excellent raw material for diagnosis,” Professor Haick told Haaretz. “It is available without the need for invasive and unpleasant procedures, it’s not dangerous, and you can sample it again and again if necessary.”

Besides cancers, the conditions the device can diagnose include Parkinson’s, multiple sclerosis. Crohn’s disease and kidney disease. The AI devise tested on 2,800 breath samples from 1,404 people in the U.S., Israel, France, Latvia and China and was able to correctly diagnose in almost 9 out 10 cases.

It’s the first time a device was created that can distinguish between different diseases in a breath sample. Artificial intelligence plays a large role in that. Professor Haick, a nanotechnology expert, explained its workings this way to Smithsonian.com:

Haick said their AI “nose” can be used in other industries as well, like security or quality control.

“Computers equipped with this technology only take minutes to autonomously analyze a breath sample that previously took hours by a human expert,” Soltoggio said. AI is making the whole process cheaper – but above all it is making it more reliable.

Scientists are testing the device on thousands of patients so it can give more accurate result. Application of this devise to real world hospital is a future thing as it is still in a development process, but researcher insures that it will soon get launch in market,

They think making this technology widespread could really impact the survival rates of patients with certain diseases by allowing for much earlier detection.

The work will be presented at the International Joint Conference on Neural Networks (IJCNN 2018), Rio de Janeiro, Brazil next month. The research paper was recently published on Research Gate.

You can read the study here, in the American Chemical Society Nano journal.

More in AI :

Forget Dance Classes, With This New AI Deepfake Your Dance Moves

World’s Smallest Medical Robot Hits Guinness Book Of World Records

This New AI Tells Whether You’re Gay or Straight Just By Your Photograph

Harvard And Google Created An AI That Predict Earthquake Aftershocks

Meet “Will” World’s First AI Digital Teacher, Teaching Kids About Energy

Researcher Created An AI Model That Detect Breast Cancer At It’s Early Stage

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.