Last week, I published a white paper arguing that the "transformation event" in digital pathology will be when AI can read slides, and write reports, and present them for validation to the pathologist. (FDA has recently approved software that can do this for certain applications within radiology, while saving 40% of radiologist time to review rather than write.)
Blog here. ("Is digital pathology close to a billion-dollar explosion in value?") For more about advances in pathology reporting by AI, blog here. ("How far away are AI generated path reports? Closer than you knew!")
Stage 1; A few days later, a graphic occurred to me and I scribbled it down to remember it:
Stage 2. I spent a few minutes making a powerpoint crude mock up of my idea:
Stage 3. Here is the AI (Chat GPT0 graphic design, after a few comments and tweaks by me.
The idea is that in current digital pathology, the lab proudly presents the pathologist with a sharp clear DICOM image on the monitor. He then reviews the case, types or dictates a report, and signs it out. The result, on a good day, is level 2 or maybe 3 on a hierarchy of report quality (Srigley report level, 1 to 6). Click to enlarge.
The new vision is that in certain areas such as prostate needle biopsies - the AI will soon read the slides, write the report, and present to the pathologist to validate, verify, sign out. The report is ALWAYS a Srigley level 6 (ICD10 perfect, SNO-MED perfect, etc).
Draft: Prostate needle biopsy has 7 cores, measuring 1x4mm to 2x10mm. Four cores have tumor. Tumor area is 30% in total. Gleasan grade is 7 (4+3). Perineural invasion is seen at two points (highlighted).
Or side by side:
Or with annotation:
