It’s good that encouraging cost-benefit calculations are being done in scientific journals and that problems like the initial investment costs mentioned at the beginning are being recognized. Perhaps that initial cost is also eased by Aiforian’s smoothing of billing (which didn’t ease an investor’s mind regarding rapid revenue development). Scanners, etc.
The computational savings presented in the article are based solely on the digital pathology workflow, such as replacing physical microscopy, improving logistics, and saving lab technician/pathologist working hours, and no numerical monetary benefits have been calculated from the direct treatment effects of AI-based predictive models. Understandable, since they are still being developed—an example from a couple of years ago is Mayo’s colon cancer predictor, QuantCRC, which determines which patients are given chemotherapy and which are not. You could calculate the price savings of chemotherapy and hospital days from that, but try calculating the human “savings” that result from it. Another example is the recently announced Stratipath Breast collaboration. Modeling savings will be possible once these are put into use.
Those predictive models will undoubtedly be part of everyday medicine someday, and then it won’t be a question of whether digital pathology is worth it; it will be an essential part of treatment decisions. And there will be a lot of those models. The use of a model participating in a treatment decision will presumably be paid for significantly more than just calculating a conventional routine cancer cell index. At that point, the stalling by individual pathologists saying “I don’t want to and I don’t have to” will no longer work. Feel free to guess when this will be everyday reality. Of course, they are already qualitatively mentioned in the article.
There are growing pains. Adoption delays and model approvals take time. Scaling is then easy for SaaS, except that infrastructure bottlenecks like scanner issues must be cleared out of the way first. You can buy a cheap scanner that requires manual labor, which gradually eats up the savings—depending, of course, on the labor cost of a given country—or an automated, expensive one. That is probably why one of the founders of Aiforia is also discussing these scanner issues with low costs Aiforia - Lääketieteen kuva-analyysiohjelmistoja - #1027 käyttäjältä jerej at approx. 37:00