Google's AI boosts accuracy of lung cancer diagnosis, study shows - STAT
benzonit at GMAIL.COM
Tue May 21 13:35:17 UTC 2019
There being a 4th and 5th:
4. What you are detecting is the disease in question. (If you find a bunch
of "stuff" which category contains all the disease in question (highly
sensitive) but not all detected "things" are the disease in question *as
stated in the premise for ordering the test* (you get no credit for
incidental findings) (low specificity) then you've created "customers" out
of "people." (Read *The Last Well Person.*)
5. That all cases so detected become, in the natural (undetected) course,
the disease in question. This is a fascinating area of research, that some
micro-disease is obliterated by the (immune system, withers on vine, etc??)
6. That the disease in question kills you before your natural time; that
you die *of *rather than *with*.
Counting never was my forte.
On Tue, May 21, 2019 at 7:20 AM Burke Harry <harry.burke at gmail.com> wrote:
> The assumptions of early detection are: (1) at some point in time the
> disease will become untreatable, (2) there is an effective treatment for
> the disease, and (3) if we detect the disease early, before it is
> untreatable, will be able to ameliorate the disease.
> Harry Burke
> On May 20, 2019, at 10:44 PM, Tom Benzoni <benzonit at GMAIL.COM> wrote:
> The question unanswered is the elephant in the room: so what?
> Does this make a patient-oriented difference or only a disease-oriented
> This may represent another cognitive error, lead time bias, or
> "How To Make People Sick for Longer Times And Increase Profits"
> E.g., say you have a tumor which grows from 1 to 5 cm in a year. It takes
> another year for the tumor to grow from 5 to 10 cm.People die at 10 cm. By
> changing the method of detection to a more sensitive one and finding the
> tumor a year earlier, we appear to make a difference, but maybe not the one
> we're sold.
> The person may have a year of sickness added onto the front of their
> disease and still die at 10 cm. It only appears you've doubled longevity.
> Remember the Will Rogers' effect.
> On Mon, May 20, 2019 at 6:19 PM HM Epstein <hmepstein at gmail.com> wrote:
>> Stat reports that — in early testing — Google’s AI achieved more reliable
>> lung cancer Dx vs. radiologists. “It detected 5% more cancers and cut
>> false positives... by 11% from reviewing a single scan. It performed on par
>> with the radiologists when prior images of patients were also included in
>> the evaluation.”
>> Website <http://hmepstein.com/> Twitter <https://twitter.com/hmepstein>
>> LinkedIn <https://www.linkedin.com/in/helenekepstein/>
>> Facebook <https://www.facebook.com/HeleneEpsteinAuthor>
>> To unsubscribe from IMPROVEDX: click the following link:
>> or send email to: IMPROVEDX-SIGNOFF-REQUEST at LIST.IMPROVEDIAGNOSIS.ORG
>> Moderator:David Meyers, Board Member, Society for Improving Diagnosis in
>> To learn more about SIDM visit:
> To unsubscribe from IMPROVEDX: click the following link:
> or send email to: IMPROVEDX-SIGNOFF-REQUEST at LIST.IMPROVEDIAGNOSIS.ORG
> Moderator:David Meyers, Board Member, Society for Improving Diagnosis in
> To learn more about SIDM visit:
Moderator: David Meyers, Board Member, Society to Improve Diagnosis in Medicine
More information about the Test