Dogs and AI to detect cancer
There are many ways to 𝐬𝐜𝐫𝐞𝐞𝐧 𝐟𝐨𝐫 𝐜𝐚𝐧𝐜𝐞𝐫, and many have disadvantages such as exposure to ionizing radiation, in the case of mammograms and computed-tomography scans, or the invasive nature of the examination, as in the case of colonoscopy. A s𝐢𝐦𝐩𝐥𝐞, 𝐧𝐨𝐧-𝐢𝐧𝐯𝐚𝐬𝐢𝐯𝐞, 𝐚𝐧𝐝 𝐬𝐞𝐥𝐟-𝐚𝐝𝐦𝐢𝐧𝐢𝐬𝐭𝐞𝐫𝐞𝐝 test has been developed to detect certain cancers in 𝐞𝐱𝐡𝐚𝐥𝐞𝐝 𝐛𝐫𝐞𝐚𝐭𝐡, combining detection by 𝐝𝐨𝐠𝐬' 𝐬𝐞𝐧𝐬𝐞 𝐨𝐟 𝐬𝐦𝐞𝐥𝐥 𝐚𝐧𝐝 𝐀𝐈.
Indeed, cancers give rise to 𝐩𝐚𝐫𝐭𝐢𝐜𝐮𝐥𝐚𝐫 𝐦𝐨𝐥𝐞𝐜𝐮𝐥𝐚𝐫 𝐩𝐫𝐨𝐟𝐢𝐥𝐞𝐬 in exhaled breath, the ability of dogs to detect these profiles has been demonstrated in several types of biological samples, and AI makes it possible to optimize 𝐫𝐞𝐬𝐮𝐥𝐭 𝐚𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐚𝐜𝐜𝐮𝐫𝐚𝐜𝐲. Exhaled breath samples from 1,386 participants were tested for breast, lung, prostate and colorectal cancer. Overall, 𝐬𝐞𝐧𝐬𝐢𝐭𝐢𝐯𝐢𝐭𝐲 𝐰𝐚𝐬 𝟗𝟑.𝟗% 𝐚𝐧𝐝 𝐬𝐩𝐞𝐜𝐢𝐟𝐢𝐜𝐢𝐭𝐲 𝟗𝟒.𝟑%. Sensitivity was similar for all four types of cancer, and sensitivity was already very good for 𝐞𝐚𝐫𝐥𝐲-𝐬𝐭𝐚𝐠𝐞 cancers (stages 0-2).
These results show that this 𝐛𝐢𝐨-𝐡𝐲𝐛𝐫𝐢𝐝, 𝐦𝐮𝐥𝐭𝐢-𝐜𝐚𝐧𝐜𝐞𝐫 𝐬𝐜𝐫𝐞𝐞𝐧𝐢𝐧𝐠 system has high sensitivity and specificity, enabling even early-stage cancers to be detected. In addition to its efficacy, the method is 𝐢𝐧𝐞𝐱𝐩𝐞𝐧𝐬𝐢𝐯𝐞. This method paves the way for a new type of cancer screening.