Dogs That Detect Prostate Cancer

Prostate cancer is the second leading cause of cancer-related death in men worldwide. A more sensitive and specific diagnostic strategy for prostate cancer is needed beyond Prostate-Specific Antigen (PSA) screening. 

Relying on Dogs' Sensitive Noses to Diagnose Cancer

Widely used Prostate-Specific Antigen (PSA) screening tests may miss aggressive prostate cancer in men or suggest that a cancer is aggressive in low-risk cases. Alternative tests continue to be investigated. There is also a study showing that dogs can be trained to detect prostate cancer from urine samples with highly accurate results. However, large-scale cancer screening with dogs would be impractical.

Florin examines a carousel holding urine samples from men with or without prostate cancer. By stopping and looking at the sample, she indicates the cancer-positive sample and receives a reward for the correct choice.

Combining Dogs' Skills With Artificial Intelligence

Guest and colleagues set out to uncover new insights that could help improve prostate cancer diagnostic tests by combining artificial intelligence and other promising detection methods with dogs' powerful sense of smell.

The researchers trained two dogs to detect aggressive prostate cancer in urine samples. These dogs demonstrated, for highly aggressive prostate cancer with a Gleason score of 9, detection sensitivity of 71% (the ability to identify truly positive cases) and specificity of 70 – 76% (the ability to correctly identify negative cases).

(A) The dogs participating in the experiment: Florin and Midas. (B) The containers they sniffed for samples. (C) The arms holding the metal test containers containing the samples. (D) Comparison of biopsy-negative controls and cancer samples in the double-blind experiment.

The team also applied two laboratory detection methods to the urine samples: Gas Chromatography – Mass Spectrometry (GC-MS) of volatile compounds and analysis of microbial species naturally present in urine. Both methods revealed significant differences between cancer-negative and cancer-positive samples.

Finally, the researchers identified specific portions of spectroscopy data that contributed substantially to the dogs' diagnoses by running an artificial neural network (a type of artificial intelligence application) and using the dogs' data. The findings suggested that larger studies could further integrate these disparate methodologies to improve detection of advanced prostate cancer and help develop new diagnostic tools that replicate dogs' olfactory abilities.

A Comment From the Study Authors: “We have shown that it is possible to mimic the dog's performance in terms of sensors and brains; now it is time to put this technology into every smartphone.”