Raw LLM Responses

I am a trainee registrar (resident) in the UK. I used to be very worried about AI. The more I work in this speciality the more I realise it's an opportunity and not a threat. 1. Our jobs are not just looking at pixels on a screen. Its about justifying the exam, choosing the correct modality, phase and timing. Nothing is black and white. Things are not as simple as yes or no. Its about appreciating the clinical history, old imaging and managing a finite resource. Most of the leading companies in the AI Radiology field I have spoken to are surprisingly uninformed about what we do. 2. X rays are a dying art form increasingly performed by radiographers. CT/MRI is widely available and the bread and butter of radiology. 3. US/Interventional radiology still require a person to perform and has high user variability. 4. Medico legally complexities - too much to cover. 5. Testing algorithms from one patient cohort/area has mixed results when applied to other patient groups. Though, I don't doubt huge training sets and deep learning can overcome this. I don't doubt it will be very useful for high volume and simple imaging like mammograms or looking for lesions on CXR. Radiologists are future proofing themselves by making sure more of their work involves patient contact/physical procedures or complex imaging like MR. There is a lot of hype and many predictions about our replacement are now overdue. We have a saying in radiology - those who use AI will replace those who don't.