With digital lightboxes, 3d imaging, automated segmentation and other workflow improvements, a lot of the "low hanging fruit" has already been removed from the process.
When my doctor thought I might have a stress fracture, I went for x-rays. I with my stint 10 year previous in medical imaging could tell at a glance I didn't have a stress fracture. The radiologists report was a brief "does not present with stress fracture nor any other visible issue". AI is not going to eliminate this; a radiologist is still going to need to spend 30 seconds "sign" the diagnosis, it's not going to take much out of the system.
The real work in radiology is the hard cases, ones that require diagnosis and consulting with other specialists. If AI can help a orthopaedic surgeon plan a surgery for a car accident victim; then it will start replacing radiologists.
When my doctor thought I might have a stress fracture, I went for x-rays. I with my stint 10 year previous in medical imaging could tell at a glance I didn't have a stress fracture. The radiologists report was a brief "does not present with stress fracture nor any other visible issue". AI is not going to eliminate this; a radiologist is still going to need to spend 30 seconds "sign" the diagnosis, it's not going to take much out of the system.
The real work in radiology is the hard cases, ones that require diagnosis and consulting with other specialists. If AI can help a orthopaedic surgeon plan a surgery for a car accident victim; then it will start replacing radiologists.