Written for Honors thesis proposal, Cognitive Science (CogSci H195), UC Berkeley, Spring 2022.

Dogs possess sophisticated olfactory capabilities that grant them the ability to detect the presence of cancer (Taverna et al., 2015), malaria (Guest et al., 2019) and COVID-19 (Jendrny et al., 2020) in human individuals. The latter has garnered widespread buzz as canines seem capable of detecting even asymptomatic COVID cases (Dickey & Junqueira, 2021) and a handful of dogs are currently employed at airports and sporting venues to sniff out infections (Bellware & Suliman, 2021). Challenges in scaling are many and include the resource and time intensive training required, difficulties in maintaining the privacy of individuals identified as infected and the black box nature of canine disease detection.

The goal of my thesis will be to use a customized EEG headset to record brain activity from dogs while they are trained to detect certain scents. This data will be used to train a ML classifier with the aim of mapping EEG brainwaves to specific scents such that the presence of an odorant can be identified directly from the brain data. This will allow canines in the field to alert human guardians about the presence of a disease anonymously, as the information can be beamed directly to a smartphone, as well as improve upon the time required for training by providing quantifiable thresholds in scent detection accuracy. Harnessing the olfactory power of dogs could be a huge boon in the fight against COVID-19 as the ubiquity of our canine friends can allow for continuous, rapid and inexpensive testing as well as be readily extend to other diseases.

Prior research in the 1960’s by Walter Freeman in rabbits indicate that different odorants produce specific EEG activity that are mathematically distinguishable from each other (Freeman & Baird, 1987). However, such experiments were invasive and used intracranial electrodes that required surgery as well as sedating animal subjects to minimize movement artifacts. More recently, a 2020 Nature study was able to use EEG activity to identify specific processing features of visual information in canine cortices. The dogs involved in the study were bred to be lab subjects and were trained for 18 months to sit still for the experiments. The fur on their head also had to be tonsured off. (Kujala et al., 2020) My thesis will use surface level electrodes (IDUN Dryodes) that should be able to handle motion artifacts as well as sit atop a canine’s furry skull. To my knowledge, this would be the first investigation into canine olfaction using surface EEG. Data will be collected from at least 5 human-canine dyad volunteers from the California Search and Rescue Dog Association (CARDA). Canine subjects will be trained to detect a certain odour A from a line-up of 5 odour ports. Following training, threshold accuracy of 90% (9/10 correct trials) will be ascertained. Brain activity of successful canines will then be recorded for 50 correct trials using the same set-up. The same procedure will be repeated for a different odour B.

For each of the 50 trials, EEG activity at the time of correct odorant identification will be considered. Data will be split into training and test data-sets (40/10). Deep learning models (e.g. EEGNet) will be trained on the data with the goal of identifying the correct odorant sniffed - A or B. The predictive accuracy of the models will be estimated using the test data-sets.

References

  • Bellware, K., & Suliman, A. (2021, September 9). Coronavirus-sniffing dogs unleashed at Miami airport to detect virus in employees. The Washington Post. ↗
  • Dickey, T., & Junqueira, H. (2021). Toward the use of medical scent detection dogs for COVID-19 screening. Journal of Osteopathic Medicine, 121(2), 141–148. ↗
  • Freeman, W. J., & Baird, B. (1987). Relation of olfactory EEG to behavior: Spatial analysis. Behavioral Neuroscience, 101(3), 393–408. ↗
  • Guest, C., Pinder, M., Doggett, M., Squires, C., Affara, M., Kandeh, B., Dewhirst, S., Morant, S. V., D’Alessandro, U., Logan, J. G., & Lindsay, S. W. (2019). Trained dogs identify people with malaria parasites by their odour. The Lancet Infectious Diseases, 19(6), 578–580. ↗
  • Jendrny, P., Schulz, C., Twele, F., Meller, S., von Köckritz-Blickwede, M., Osterhaus, A. D. M. E., Ebbers, J., Pilchová, V., Pink, I., Welte, T., Manns, M. P., Fathi, A., Ernst, C., Addo, M. M., Schalke, E., & Volk, H. A. (2020). Scent dog identification of samples from COVID-19 patients – a pilot study. BMC Infectious Diseases, 20(1), 536. ↗
  • Kujala, M. V., Kauppi, J.-P., Törnqvist, H., Helle, L., Vainio, O., Kujala, J., & Parkkonen, L. (2020). Time-resolved classification of dog brain signals reveals early processing of faces, species and emotion. Scientific Reports, 10(1), 19846. ↗
  • Taverna, G., Tidu, L., Grizzi, F., Torri, V., Mandressi, A., Sardella, P., La Torre, G., Cocciolone, G., Seveso, M., Giusti, G., Hurle, R., Santoro, A., & Graziotti, P. (2015). Olfactory system of highly trained dogs detects prostate cancer in urine samples. The Journal of Urology, 193(4), 1382–1387. ↗