Sandor,
I appreciate what you are saying, and perhaps we are having a discussion on
semantics.
"...that screening of only 15,000 Caucasians in the US would yield
300 growing melanocytic tumors in the uvea".
This sounds like 1 in 50 people have malignant melanomas. We have a lot less
than that here in Indiana.
If you say a lot of people have choroidal nevi (if you look hard enough and far
enough in the periphery) , and that some of them (I am surprised it is as high
as 18%) might show slow growth over years or decades, and some of these might
eventually become malignant melanomas, I could see that as valid. However, 300
some out of 15,000 is concerning.
I am under the impression a slow-growing nevus is a still a nevus (though a
technically "growing melanocytic tumor"), and of course should be followed.
When it turns fast-growing and malignant, then it is a melanoma.
I would not want to unduly alarm patients. Most ophthalmologists tell patients
"A nevus is like a freckle in the layer beneath the retina, and like most
freckles, usually nothing to worry about. But just to be safe, we will
photograph it and watch to see if it changes any in the future". A few have
offhandedly said "It might be a tumor that could be malignant", when it was
just a small, flat nevus. Some patients will obsess about that for the next
year until their annual exam.
Thanks for the articles from Dr. Shields and Dr. Shields, et. al. They know a
few things about ocular tumors. I will read those when I have time.
Tom
________________________________________
From: optimal-bounce@xxxxxxxxxxxxx <optimal-bounce@xxxxxxxxxxxxx> on behalf of
Sandor Ferenczy <sandorferenczy@xxxxxxxxx>
Sent: Friday, February 23, 2018 11:50 AM
To: optimal@xxxxxxxxxxxxx
Subject: [optimal] Re: [**External**] Google and Ophthalmic Imaging Neural Nets
No. we meant screening 15,000 caucasians could yield 300.
The differing numbers (ie 5/6 in a million vs. 324 in 15,000) are incidence
(diagnosed) vs. prevalence (actual occurrence)
General population screening is how we attempt bridge the gap between
occurrence & diagnosis.
In a multi-ethnic population, nevi occur in 5% of 2x 45 degree FOV fundus
images (relatively low coverage of the retina)
Diabetic studies have found 12% occurrence of nevi in wide-field imaging.
18% of detected nevi have been reported to show growth, and these with growth
then represent 3x higher metastatic risk.
in an otherwise unscreened, asymptomatic population:
12% of 15,000 is 1800 potentially with a melanocyctic tumor
18% of 1800 is 324 potentially with a growing melanocytic tumor & 3x higher
risk of metastatic disease.
Qiu M, Shields CL. Choroidal nevus in the United States population. Racial
disparities and associated factors in the National Health and Nutrition
Examination Survey. Ophthalmology 2015;122:2071-2083.
Silva PS, Cavallerano JD, Haddad NMN, et. al. Comparison of nondiabetic retinal
findings identified with nonmydriatic fundus photography vs ultrawide field
imaging in an ocular telehealth program. JAMA Ophthalmol 2016;134(3):330-334.
Shields CL, Shields JA, Kiratli H, et. al. Risk factors for growth and
metastasis of small choroidal melanocytic lesions. Ophthalmology
1995;102:1351-1361.
This simply showcases two things:
1) the disparity between natural occurrence of the tumors (whether benign or
malignant) and clinical diagnosis (ie we need better population screening)
2) the shortcomings of smaller FOV screening @ eye doctors offices. (< 50
degree FOV vs. wide field)
Caveat - we did this statistical analysis for grant applications for general
population screening. It is a statistical possibility, and in our minds a
reason to increased screening, but the hard general population data is not
known yet.
-sandor
On Feb 22, 2018, at 10:39 AM, Egnatz, Thomas <tegnatz@xxxxxxxxx> wrote:
Sandor,
Did you perhaps mean screening 15,000,000 Caucasians would yield 300
melanomas?
But that should only be 75 melanomas.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4399142/
Uveal melanoma represents 79-81% of ocular melanomas and 3-5% of all
melanomas.[1,2,3] In the United States, the incidence of uveal melanoma is
5/million population.
Tom
From: optimal-bounce@xxxxxxxxxxxxx <optimal-bounce@xxxxxxxxxxxxx> on behalf
of Sandor Ferenczy <sandorferenczy@xxxxxxxxx>
Sent: Wednesday, February 21, 2018 9:38 AM
To: optimal@xxxxxxxxxxxxx
Subject: [optimal] Re: [**External**] Google and Ophthalmic Imaging Neural
Nets
I more picture a self-guided imaging system next to the self-guided blood
pressure system @ the local pharmacy.
Imagine the power of screening tens of thousands of wide field, color images
with machine learning, abnormal findings whether vascular, nerve, retinal,
tumor can be forwarded directly to the existing reading centers @ the world's
best hospitals & patient be told to follow up with a local ophthalmologist.
In statistical work we have done, we estimate that screening of only 15,000
Caucasians in the US would yield 300 growing melanocytic tumors in the uvea.
...imagine...
No human could do intra ocluar screening as efficiently as a computer. we
just need to train them what to look for.
-sandor
On Wed, Feb 21, 2018 at 9:18 AM, joey hatfield <bioartevolution@xxxxxxxxx>
wrote:
Imagine, program guided self assessment fundus imaging on a phone that can
send a report to the physician. Evaluation of the posterior pole for this
application, diabetic changes, etc...AI is very real and being looked at
heavily by most major companies.
Joey Hatfield
Area Sales Consultant
Heidelberg Engineering
Mobile: 501-515-0697
email: jhatfield@xxxxxxxxxxxxxxxxxxxxxxxxx
10 Forge Parkway l Franklin, MA l 02038
Tel:
Fax:
www.heidelbergengineering.com
www.spectralis.info
On Feb 21, 2018, at 8:06 AM, Sandor Ferenczy <sandorferenczy@xxxxxxxxx> wrote:
Neither fake news or bad science.
Machine learning (a computer's progressive improvement at a task based on
training rather than explicit programming) is real, has been around since
the '50s, and is more and more being applied to medicine.
The most important part of this general public article are the second two
paragraphs:
The algorithms didn’t outperform existing medical approaches such as blood
tests, according to a study of the finding published in the journal Nature
Biomedical Engineering. The work needs to be validated and repeated on more
people before it gains broader acceptance, several outside physicians said.
But the new approach could build on doctors’ current abilities by providing
a tool that people could one day use to quickly and easily screen themselves
for health risks that can contribute to heart disease, the leading cause of
death worldwide.
-sandor
On Tue, Feb 20, 2018 at 9:34 PM, CPMC Ophthalmic Diagnostic Center
<cpmceyelab@xxxxxxxxxxxxxxxx> wrote:
I am unclear as to how these poor Topcon images would help screen for stroke
Unless they are referring to H. Plaque, Occlusions, etc and the usual signs
that would send a patient for a Carotid angiogram.
Fake news or bad science?
Denice Barsness, CRA, COMT, CDOS, FOPS
CPMC Dept of Ophthalmology/ The Eye Institute
Ophthalmic Diagnostic Services
711 Van Ness Avenue Suite 250
San Francisco CA 94109
415-600-5781
FAX 415-558-7011
From: Barsness, Denice
Sent: Tuesday, February 20, 2018 3:30 PM
To: CPMC Ophthalmic Diagnostic Center <cpmceyelab@xxxxxxxxxxxxxxxx>
Subject: FW: [**External**] [optimal] Google and Ophthalmic Imaging Neural
Nets
From: optimal-bounce@freelists.orgOn Behalf Ofjmc eye photo
Sent: Tuesday, February 20, 2018 3:29:24 PM (UTC-08:00) Pacific Time (US &
Canada)
To: optimal@xxxxxxxxxxxxx
Subject: [**External**] [optimal] Google and Ophthalmic Imaging Neural Nets
WARNING: This email originated outside of the Sutter Health email system!
DO NOT CLICK links if the sender is unknown and never provide your User ID
or Password.
Interesting article at link below. Article says images are scans, but the
notch says otherwise.
https://www.washingtonpost.com/news/the-switch/wp/2018/02/19/google-used-artificial-intelligence-to-predict-heart-attacks-with-the-human-eye/?utm_term=.27acdee9d2c2
Anyone out there participating is such studies?
Richard Press posted about a similar project re diabetes last march.
ALSO — Is Google a sustaining member of OPS?
john michael coppinger
jmc eye photo
