Today it would seem that machines are always above the care of doctors, with the unfailing consistencies that only the computer can achieve, detecting dangerous lesions and questionable moles. This month, Google carried out a study showing that its AI programs are more effective than doctors to detect breast cancer on mammograms.
However, what these studies show for many healthcare professionals is not just the AI promise, but also its potential threat. They claim that the nuanced judgmental abilities of the nurses and doctors are not so easily digitalized despite the obvious capacity of algorithms to compress data. And this technology may exacerbate existing problems in places where technology companies support medical RNs.
The main criticism for Google’s mammograms paper is that the company tries to simplify an already very contested operation. Christie Aschwanden reported in Wired earlier this month that early breast cancer scans could be as dangerous as they are useful. doctors have been saying for years now.
Research have shown that you can show a group of doctors the same lesions early on and get completely different responses to whether they are cancer. And even if they believe that this is what a lesion means-and it’s right to diagnose-it’s impossible to know if this cancer is a life hazard to anyone. Adamson claims this leads to an overdiagnosis: “Calling cancer something that would not affect people in their lives if you weren’t going to look for them.
If you call something cancer, it starts a series of painfully expensive and life-changing medical interventions. In breast cancer, radiation therapy, chemotherapy, breast tissue removal (lumpectomy) and the removal of one or both breasts (mastectomy) may be included. These are not hurrying decisions.
But in Google’s study, Adamson says, the complexities of such a diagnosis are not sufficiently considered. First of all, the researchers of the organization have based their algorithm on images which are already or not cancerous. Second, only binary results are created by a Google algorithm: yes, this is cancer, or no, it doesn’t. As Adamson argues in a recent article, the third option that represents the grey field of diagnosis should be uncertainty, which prolongs instead of concludes the debate.