Kozminski Techblog

Cancer is the second leading cause of death globally and is responsible for an estimated 9.6 million deaths in 2018. Globally, about 1 in 6 deaths is due to cancer. To fight it, the key is detecting it early on. The task is extremely difficult, as cancer in the early stage of growth is extraordinarily difficult to detect. Standard screening methods such as radiological imaging can miss signs of cancer or return a false negative (as it does in 20-30% of cases). Hereditary testing is another detection method that determines genetic predisposition to cancer, but this does not provide much detail and cannot reveal if a person has cancer now. It takes a lot of time and resources to find developing cancer cells. But there might be a solution.

In recent study, researchers trained an algorithm to detect early signs of cancer cells forming. AI can not only greatly improve the accuracy of image detection for cancer, but could also open up entirely new fields between genomics and cancer screening.

Founded by six deep-learning experts from KAIST University in South Korea in 2013, Lunit trained their INSIGHT algorithm on chest x-rays and mammography images to detect lung and breast cancer. Boasting a 97% detection rate for lung and breast cancer, Lunit put their success down to training: “rather than guiding our algorithm to a specific location, we provided a region and said “there is a nodule there, try to find it,” and let the algorithm learn by itself,” says Brandon Suh, Lunit’s CEO.  “It is difficult for doctors to find small nodules hidden behind ribs or organs in chest x-rays” says Suh, but their algorithm searches for cancer patterns to dramatically reduce the chance of a false negative or missed case of cancer.

This unique approach to accurately detect and treat cancer is part of a wave of AI implementation in healthcare that does not look likely to stop. Wide-scale implementation of AI could lead to a fully proactive healthcare system, that responds to diseases preemptively istead of being focused on treating sick people. Unfortunately, there are some obstacles. Aside from a lack of data, mindset and professional changes are major obstacles for AI in healthcare. But if applications like Lunit can make themselves known, and more importantly understood by the healthcare community, then AI will be a powerful weapon in the fight against cancer.

Sources:

https://www.forbes.com/sites/charlestowersclark/2019/04/30/the-cutting-edge-of-ai-cancer-detection/#2582c7e27336

https://www.who.int/news-room/fact-sheets/detail/cancer

https://m.medicalxpress.com/news/2019-12-artificial-intelligence-previously-unknown-features.html