Medicine normally rather prevents than cure, and cure in case of diagnosed disease. Unfortunately, still nowadays, doctors and surgeons attempt to cure rather than preventing, which lead generally to diverse extensive costs (Physical, emotional, monetary, and budgetary) to the patients, to their surroundings, and to the society they live in. This is especially the case in the oncology area. In the U.S on average 439,2 per 100,000 men and women are diagnosed of cancer every year, and 163,5 per 100,000 men and women every year die from it . Roughly speaking, 1 man of 2 and 1 women out of 3 who got diagnosed by cancer will die from it. One of the main causes of this high rate of death can be due to the fact that doctors diagnose the disease way too late (National Cancer Institute, n.d).
In this regards, Artificial intelligence might be a partial if not a absolute solution. In 2016, Watson, an AI machine of IBM saved a Japanese female patient of 60 years old by diagnosing a rare leukaemia (blood cancer). Other tests realised that Watson could detect skin cancer on patients at a precision of 90%, while oncologist detects at a rate of 85% of all cases. What’s more interesting is mixing Watson and the oncologists’ opinion increase this rate to 95% of detecting capacity (Sicara, 2019).
Furthermore, a team in AI laboratory of the MIT and of the Massachusetts General Hospital (MGH) has created and trained a deep learning model capable of predicting whether a patient is highly likely of developing a breast cancer in the future, based on breast radio (mammograms). The algorithm was trained on 60 000 cases and had learnt to detect subtle patterns signals in mammary tissues to predict a future cancer ( Conner et al., 2019).
Nowadays, medical machine learning are far from being perfect, as there is a fear of « adversarial attacks », which is the manipulations of fragments of data that can alter the behaviours of AI. In other words, by wrongly coding the input, the machine learning would provide wrong output meaning in the medical area that it would diagnose a disease on patient that has none (Sicara, 2019; Finlayson et al., 2019).
However, due to the true potential that AI presents, do you think that, in the future, it would replace to a certain extend or completely doctors when it comes to diagnoses of disease? Would AI contribute to the prevention of potential diseases?
Sources:
Conner, A., Gordon, S., & Gordon, R. (2019) Using AI to predict breast cancer and personalize care. MIT CSAIL. Retrieved from: https://www.csail.mit.edu/news/using-ai-predict-breast-cancer-and-personalize-care
Finlayson, S. G., Bowers, J. D., Ito, J., Zittrain, J. L., Beam, A. L., & Kohane, I. S. (2019). Adversarial attacks on medical machine learning. Science, 363(6433), 1287-1289.
National Cancer Institute. (n.d). Cancer statistics. Retrieved from: https://www.cancer.gov/about-cancer/understanding/statistics
Sicara. (2019). Quand les médecins arrêteront de nous soigner. Retrieved from:
https://www.sicara.fr/parlons-data/2019-05-20-intelligence-artificielle-medecine-predictive
Hi Marc!
I really think that this could be the future especially because cancer is so dangerous, and a timely diagnosis could really save a lot of lives!!!!
However, I still think that AI would never replace doctors completely. A doctor’s experience could never be replaced by an AI, regardless with how much data you train it. Especially for very specific diseases where not much data has been gathered. The combination of a doctor’s experience with the precision of AI could be one of the greatest inventions of humans, and save so many lives!!!