Accuracy of #AI based measurements on echocardiography www.jacc.org/doi/10.1016/... #cardiosky @jaccjournals.bsky.social
23.09.2025 18:14
👍 1
🔁 2
💬 0
📌 0
Accuracy of #AI based measurements on echocardiography www.jacc.org/doi/10.1016/... #cardiosky @jaccjournals.bsky.social
More about likelihood ratios in diagnostic testing:
Jaeschke R, et al, JAMA, 1994.
www.pubmed.gov/8309035
Interpretation for both LR+ and ILR- in terms of how conclusive the test changes pretest to posttest likelihood of a positive (or negative) finding:
>10 = large
5-10 = moderate
2-5 = small but sometimes important
<2 = inconclusive
Inverse negative likelihood ratio (ILR-) = 1/LR- = specificity / (1-sensitivity) = 92.3% / (1-41.2%) = 1.6 = how many more times likely it is that you do NOT have the disease in question thanks to the test
Positive likelihood ratio (LR+) = sensitivity / (1-specificity) = 41.2% / (1-92.3%) = 5.4 = how many more times likely it is that you have the disease in question thanks to the test
However, sens and spec values are highly dependent on disease prevalence in the population being studied. The measures of test performance that are independent of population disease prevalence are likelihood ratios, which tell us how many times more likely a diagnosis is after applying a test.
In Apple’s white paper on their Hypertension Notification Feature for Apple Watch, they report low sensitivity (41.2%) and high specificity (92.3%). They rightfully chose a threshold for notification while seeking to keep the false positive rate low.
The white paper: www.apple.com/health/pdf/H...
Read on if you are interested in (the Apple Watch) diagnostic test performance.
It's all about the likelihood ratios.
I still applaud Apple and think the Hypertension Notification feature will be useful for the individual and society. Any nudge that can help early identification of hypertension is valuable, especially when it is passively integrated into everyday use of a consumer device as implemented by Apple.
In other words, you should NOT rest on your laurels and assume that you don't have hypertension. Get your blood pressure tested regularly regardless, because hypertension is very common, deadly in the long term, and highly treatable.
If you do NOT get a Hypertension Notification despite using your watch, then you are NO MORE LIKELY to be free of hypertension than before you used your watch (inverse negative likelihood ratio 1.6, which is inconclusive).
If you get an Apple Watch Hypertension Notification, then you are 5 TIMES MORE LIKELY to have hypertension than if you didn't use your watch (LR+ 5.4). This is moderately useful test information. You should definitely get your blood pressure tested.
A thread..
#CardioSky
#MedSky
#Apple
#AppleWatch
Very much so, in a way that has surprised me. We are currently studying this prospectively.
Not at all. I rather like the approach, which can be done without compromising scientific rigor, as in the following paper from our group.
PubMed.gov/39846063
Mentor – senior person that broadly knows your field and gives you specific advice Coach – does not need to know your field, and asks you questions to help you guide yourself to do what you want to do, without necessarily giving you specific advice Supervisor – the person you report to and are accountable to in your work, dependency relationship Boss – is often the Supervisor, but may be one or more steps up in hierarchy, the one who hires and fires and pays your salary Role model – someone you look to for inspiration and wish to emulate Advocate – promotes you and your interests, perhaps by praising your work in a meeting Sponsor – actively invests in your growth, perhaps by using their influence to provide opportunities for advancement even when you are not in the room
Different roles, often overlapping, all there hopefully to help.
#mentor
#coach
#supervisor
#boss
#rolemodel
#advocate
#sponsor
#MedSky
#AcademicSky
#mentorship
In our experience, "accelerated aging" measured in years of healthy human aging is a unit of measurement that has unsurpassed clarity of understanding among the general public (patients). We need to be more 'tabloid' to get our message across, imho.
John Greenwood bringing it home on #WhyCMR vs #CVnuc for myocardial perfusion imaging. #CardioSky
#CardioSky #WhyCMR #EchoFirst
Imaging of aortic regurgitation presented eloquently by @JoaoLCavalcante #CSANZ2025
🤝poor agreement b/t echo & CMR
💰big price to pay when AR labelled moderate
🧲use CMR to accurately assess AR
#ANZCMR #whyCMR @scmrorg.bsky.social
Thursday #CSANZ2025 is packed with so many great meetings! I’ll be going between the imaging, heart failure and genetics meetings. Which one will you attend? @anastasiasmihaili.bsky.social @mugander.bsky.social
We don’t yet have definitive data on it, but the non-LAVI (c) formula is likely best for short-term changes in PAWP, whereas (b) is just as good in stable hemodynamics, and (a) is good enough if you haven’t acquired PV velocities.
We have actually have three formulas for estimating PAWP.
From Lindow 2024 using:
a) E + LAVI
b) E + PVs + LAVI
and from Lindow 2025 using
c) E + PVs + PVd
Their accuracy, precision and prognostic association are presented in each paper.
Both HFpEF and HFrEF were included. However, it was limited to those with at most mild mitral regurgitation or stenosis, and sinus rhythm.
More details are in the paper.
academic.oup.com/ehjimp/artic...
The four fallacies of (dis)loyalty between you and an organization, and vice versa. Created by martin@ugander.com
Beware the Four Fallacies of (Dis)Loyalty.
In providing career guidance, mentoring and coaching to academics, aspects of loyalty and holding grudges (disloyalty) often come up. This has provided structure to some of these conversations.
#AcademicSky
#PhDSky
#LeadershipSky
#MentorSky
#CoachSky
Tirzepatide reduces LV mass and paracardiac adipose tissue, a potential mechanism for reduced HF events in SUMMIT @jaccjournals.bsky.social @chriskramermd.bsky.social www.jacc.org/doi/10.1016/...
Our latest work, led by Thomas Lindow, on estimating wedge pressure by #EchoFirst. #Cardiosky
LA volume is not necessarily needed. LA inflow (PVs, PVd) and outflow (E) velocities, which are proportional to pressure gradients, perform just as well.
A multi-institutional collaborative effort 🇦🇺🇸🇪
Elegant TTE color Doppler visualization of what looks to me like blood flow between aortic root and left atrium.
#EHJCVI 🔎 What’s the link between ECV, scar burden, myocyte volume, and diastolic dysfunction in moderate AR? #whyCMR shows these parameters increase with worsening diastolic dysfunction 📉Is there an association with mortality? Read more 👉 doi.org/10.1093/ehjc... #CardioSky @jgrapsa.bsky.social
Accuracy of LV mechanical dyssynchrony indices for mechanical characteristics of left bundle branch block using #WhyCMR feature tracking
academic.oup.com/ehjcimaging/...
@sarahmoharem.bsky.social @purviparwani.bsky.social @vassv.bsky.social
Prognostic Relevance & Lower Limit of Reference Range of LV GLS: Clinical Validation Study
#CardioSky #JACCIMG
www.jacc.org/doi/10.1016/...
@docstrom.bsky.social @davidwienermd.bsky.social @jgrapsa.bsky.social @vietheartpa.bsky.social @anastasiasmihaili.bsky.social @jaccjournals.bsky.social