5/ Compared to LVEF and GLS, #MAPSE is ⬇️dependent on image quality. Figure shows feasible #AutoMAPSE despite poor img. Thus:
#MAPSE➕#TEE➕AI 🟰 Hands-FREE #EchoFirst.
This could make continuous monitoring of LV function possible in #ICU #CriticalCare
6/
❓RESEARCH QUESTION❓
We asked if the combination of #MAPSE➕#TEE➕AI (i.e. #AutoMAPSE) could be used for continuous monitoring of LV function in #ICU
#CriticalCare
7/ METHODS AND RESULTS
We monitored 50 #ICU patients after cardiac surgery.
Every 5 mins, we recorded a set of hands-free 2C and 4C views that comprised 10 ❤️-beats (Figure below).
To ⬆️precision, we report 1 "measurement" as the avg MAPSE of 10 beats from 1 specific wall.
8/ We deemed #AutoMAPSE feasible if it estimated MAPSE from ≥1 out of 4 LV walls.
In fig, the blue dot shows feasible #AutoMAPSE
This was because experiments have shown that the #MAPSE of any wall reflect global LV function, not regional LV function: doi.org/10.1152/japplp…
9/ For monitoring, the same wall must be reassessed over time.
Thus, we defined 'monitoring feasibility' as
🟢'Excellent' if the same wall could be monitored >90% of the time.
🟡'Good' if the same wall could be monitored 50-90% of the time.
🔴'Poor' if <50%.
10/
Of all 50 patients, we found the monitoring feasibility of #AutoMAPSE to be
🟢'Excellent' in 88%
🟡'Good' in 6%
🔴'Poor' in 6%
11/ To asses the precision🎯of #AutoMAPSE, we recorded a triplicate sets of image. We report precision as 'least significant change'.
To asses of bias and agreement, we also measured #MAPSE manually in these triplicate images.
12/ We found that #AutoMAPSE was more precise than manual measurements (least significant change 2.4 vs 2.7 mm).
Also, #AutoMAPSE can easily⬆️precision by instantaneously avg more measurements, something manual measurements cannot due to high workload.
13/ Compared with manual MAPSE, #AutoMAPSE had a no bias and good agreement (bias -0.2 mm, limits of agreement -3.4 to 2.9 mm)
This means that #AutoMAPSE and manual #MAPSE on avg measures the same values👍✅
14/ CONCLUSION
LV dysfunction in #CriticalCare is bad. We need continuous monitoring of LV function #ICU.
Continuous monitoring of LV function using #AutoMAPSE had🟢excellent feasibility. Compared with manual #MAPSE, #AutoMAPSE was more precise🎯 and had no bias.
End🧵
unroll @threadreaderapp
• • •
Missing some Tweet in this thread? You can try to
force a refresh
1/ 🎯Bland-Altman analysis is the common way to evaluate AI in #EchoFirst. However, BA analyses can be tricky to interpret. 🧵On the paradox of BA analysis for AI in #EchoFirst with examples from our studies on #AutoMAPSE 📜
#Foamed #Cardiotwitter rdcu.be/dH8aR
2/ BA analysis is an important approach used for comparing 2⃣measuement methods for measuring the same unit, like #MAPSE (mm) and #AutoMAPSE (mm). ‼️Correlation (r) is inappropriate yet commonly used because it is seductive. Read the reasoning in their📜: doi.org/10.1016/S0140-…
3/ The output in BA analysis is 1⃣bias &2⃣limits of agreement (LOA). Bias is the avg difference across all paired measurements of manual MAPSE and autoMAPSE. LOA=+/- 2SD of the bias in the study. The bias and LOA is presented in a BA plot from 📜: rdcu.be/dH0mY
🧵1/ #AutoMAPSE had high feasibility for estimating MAPSE from ≥1⃣wall. Many worry that #MAPSE is limited by RWMA, such as acute MI - however, evidence suggests that regional MAPSE reflects global LV function.
2/ Several studies have shown that in AMI, MAPSE is ⬇️even in walls remote to MI. This was confirmed experimentally by Berg et al, in an experiment occluding the LAD in pigs and measuring #MAPSE by #WhyCMR.
First paper:
Berg et al: doi.org/10.1080/140174… doi.org/10.1152/japplp…
3/ ⬇️MAPSE in remote walls is explained through the distribution of afterload. For a specific region to shorten, that region must overcome the opposing afterload imposed by its neighboring region, which is shortening in the opposite direction.
🧵We have now refined AutoMAPSE.!
🔓
⬆️New paper on #autoMAPSE showing that the newest version achieves an optimal balance between feasibility, analysis speed, and agreement.
A🧵on the clinical importance of these findings
#echofirst #CardioTwitter #POCUS doi.org/10.1016/j.ultr…
1/ Our design involved measurements from 🔟beats per #EchoFirst recording. This is very different from the usual practice of measuring 1-3 beats. What are the advantages?
2/ Measuring 1-3 beats in #Echofirst involves huge physiological variability. How?
❤️🫁 interactions can cause serious beat-to-beat variations in any echo measurement. Often overlooked, these changes may be misinterpreted as real. Acting on wrong info ⬆️the risk of harm.
The RVOT or PA Doppler has a lot of underutilized potential for hemodynamic monitoring of the RV #ThePeoplesVentricle
🧵On the physiology of the RVOT/PA Doppler. Simplified understandings. Corrections are welcome!😃
#FOAMcc #EchoFirst #Cardiotwitter #FOAMed #Medtwitter #POCUS
VTI reflects stroke volume, but values are much lower than for LVOT. Lower cut-off for 60-79 year olds is actually around 10 cm.
Monitoring changes are likely more useful. doi.org/10.1016/j.jcmg…
AcT and Vmax/AT both reflects RV afterload. Increased afterload reduces AT and Vmax/AT.
➡️AcT correlates with mPAP and PVR. doi.org/10.1161/01.CIR…
2/ The goal of monitoring is
1⃣to detecting small and early changes
2⃣so that therapies can help patients.
For successful monitoring, the measurements must be precise and acquired rapidly and effortlessly. Neither #EchoFirst nor #POCUS fulfil these criteria.
3/ Eyeballing is rapid, effortless, and OK for diagnosing LV dysfunction.
But eyeballing is NOT precise because it categorises LV function.
Changes in LVEF from 45 to 30% are important, yet still in the same category ➡️ undetectable by categorical assessment.