-
CT body composition measures help predict lung cancer recurrence
If you need an automated AI platform to run locally on your laptop/desktop/HPC to assess muscle area, volume and radiodensity (and also volumetric measurements from multiple organs and tissues), then check out the DAFS platform from Voronoi Health Analytics:
[link=https://www.voronoihealthanalytics.com/%C2%A0]https://www.voronoihealthanalytics.com/ [/link] -
1 Reply
-
devils advocate here.
“XX predicts recurrence of cancer YYY” Where XX is some unexpected parameter – like presence of nearsightedness predicts recurrence of breast cancer in 10 years.”
this is a particular genre of publication.
My question is always “So what?” Just tell the patient “you are likely to have a breast cancer recurrence” and send them home to be tortured by more worries?
OR – recommend full mastectomies in every nearsighted woman? OR – recommend more chemo for every nearsighted woman.My point, if it is not clear, is that these random associations do not lead to anything actionable. So what if they have a higher risk of recurrence (especially at the levels reported in this article). What would you do differently? Will you double the number of screening exams these patients get? No. Will you drop all screening for those with normal sight? No.
Even some correlations that have an intuitive connection with the outcome being sought, are pretty much useless. I come at this from the standpoint of a cardiac imager who could never figure out what to do with papers that said “Coronary artery calcification increases the likelihood of cardiac events by XX%.” What do you do with that????
now let me argue with myself: I could say “well, it may lead to new lues about the reason for recurrence, and lead to research to further elucidate the mechanisms of recurrence”. Hmmmm.. two thoughts – 1) I cannot recall a single instance of such research leading to anything really useful 2) with Hazard ratios of 1.12 and 0.88, the effect (especially in a retrospective review) is highly likely to disappear with the next publication.moreover, in the description of this paper, it gives no 95% confidence intervals. Further it is a retrospective review – therefore highly subject to selection bias. There is no report of covariates being investigated. I would at least like to see smoking history included. (maybe the original paper had this, I am not interested enough to look, in all honesty)
I can see one use for this – pumping up the CV. Oh – and selling software.
The question remains – SO WHAT?
