Also, how would you train AI for this?pic.twitter.com/DUMzBpf7Kv
আপনি আপনার টুইটগুলিতে ওয়েব থেকে এবং তৃতীয়-পক্ষ অ্যাপ্লিকেশনগুলির মাধ্যমে অবস্থান তথ্য যেমন শহর বা সুনির্দিষ্ট অবস্থান যোগ করতে পারবেন। আপনার কাছে আপনার টুইটের অবস্থান ইতিহাস মোছার বিকল্প থাকবে। আরও জানুন
Simple - you can’t. Best would be to triage it it as ‘abnormal’ for human review
Looking for potentially interesting abnormalities is exactly what we should be teaching algorithms to do. That's where the biggest societal impact is. Teaching them to give more detailed findings adds very little
Depends very heavily on the use case and the clinical context. We should explore both.
I'd be interested to hear about clinical contexts where detailed findings has large societal impact in the short-medium term. I haven't come up with them as yet.
Grade and stage of cancer?
No I don't think so - I studied that one v closely. Finding the nodules is the thing that saves 90% of time. Deciding whether to do a followup biopsy rarely needs computer help AFAICT - if it's small but looks worrying, you just schedule a followup scan. Do you agree?
Makes all the difference in most visceral tumours! Lung nodules relatively simple pathway.
Ah - but if blockbuster had adopted the new tech they would still be in business! That’s the key here.
Good point - remember that Blockbuster had online movie streaming service too....
Ultimately like with Netflix the customers (ie clinicians/patients) will decide the fate of imaging as AI advancements drastically reduce the need for Radiologist physician labor
They were just late to the game! Tried to hold on to physical media for too long. Same will happen to rads who don’t adopt and adapt. Those that move with the times won’t be replaced.
Absolutely
Because big companies will not want to take medicolegal responsibility for the report.
But they already are. AI is already used in diagnostics and there are already systems in place (most fall on the treating physician to know the vaildation of the machine/AI in question)
Nobody is! Please find an example of this if you are going to make this claim.
I was commenting on the libility issue Sriram mentioned in case you responded out of context. We literally use machine learning to diagnose brain tumors subtypes. Computers already make diagnosis in medicine was my point to his comment. We have already accepted that risk.
Ok - where’s the example?
Scan time and dose reduction probably...but automated acquisition?
I think that the human treatment and professionalism of the #radiographer will always be necessary.
#radiologia #TSID #AI
It's already being investigated - automated anatomy detection on scout views, automated sequence planning etc.
It seems good, but a few words, the human treatment of the #radiographer to the #patient, a few words of encouragement ... will always be more useful than the coldness of the automation. #HRAD #TSID #Radiology #Radiología
টুইটার তার ক্ষমতার বাইরে চলে গেছে বা কোনো সাময়িক সমস্যার সম্মুখীন হয়েছে আবার চেষ্টা করুন বা আরও তথ্যের জন্য টুইটারের স্থিতি দেখুন।