1. If a tumor is incorrectly identified (and it happens a lot) no one realizes it until much later. In many cases the radiologist never knows what the outcome of the detection was. The patient could have been unnecessarily operated, put on year long drug schedule for no reason etc. Unlike anesthesiology where the results of any action are more or less in your face.
So all these AI startups are going to be training on very incomplete data.
2. More detection more profit.
There is no incentive to minimize false detections. The entire system from the radiologist, to the doc/hospital visits, to surgeons, to pharma, to the equipment manufacturers from false detections. Nobody is getting paid to reduce false detections.
So all these med AI startups are going to do quite well for themselves whether they actually benefit the public or not.
1. If a tumor is incorrectly identified (and it happens a lot) no one realizes it until much later. In many cases the radiologist never knows what the outcome of the detection was. The patient could have been unnecessarily operated, put on year long drug schedule for no reason etc. Unlike anesthesiology where the results of any action are more or less in your face.
So all these AI startups are going to be training on very incomplete data.
2. More detection more profit. There is no incentive to minimize false detections. The entire system from the radiologist, to the doc/hospital visits, to surgeons, to pharma, to the equipment manufacturers from false detections. Nobody is getting paid to reduce false detections.
So all these med AI startups are going to do quite well for themselves whether they actually benefit the public or not.