EchoSolv AS
K241245Echo IQ Ltd · cleared 2024-10-04 · product code POK · Radiology
Premarket evidence — what FDA accepted
Device typesamd
source quote (p.4)
“EchoSolv AS is a machine learning (ML) and artificial intelligence (AI) based decision support software indicated for use as an adjunct to echocardiography for assessment of severe aortic stenosis (AS).”
Algorithmmachine learning (ML) and artificial intelligence (AI) based
source quote (p.4)
“EchoSolv AS is a machine learning (ML) and artificial intelligence (AI) based decision support software indicated for use as an adjunct to echocardiography for assessment of severe aortic stenosis (AS).”
Adaptive (vs locked)No
source quote (p.6)
“The EchoSolv AS AI Model was developed on a dataset consisting of 631,824 individuals with 1,077,145 transthoracic echocardiograms (TTE). The dataset was randomly split (ratio 70:30 based on individuals) into two separate groups, training and test set. Data from 442,276 individuals (70%) were entered into the AI model to train the device to detect severe AS cases. The remaining 189,548 individuals (30%) were reserved for internal testing. Individual patients appeared only once in either the training or test dataset but not both.”
PCCPFDA source did not state this
Cybersecurity addressedFDA source did not state this
Validation studies (2)
Standalone
n=6,268 patients
endpoints: ROC curve; AUROC; Diagnostic likelihood ratios (DLR)
Reader study (MRMC)
n=200 cases · 1 site(s)
endpoints: ROC curves; AUROCs
Reported performance (8 observations)
sensitivity0.801CI 0.786-0.818
source quote (p.7)
“Sensitivity and specificity at the high probability threshold were 0.801(95%CI: 0.786-0.818) and 0.923 (95%CI: 0.915-0.932), respectively.”
specificity0.923CI 0.915-0.932
source quote (p.7)
“Sensitivity and specificity at the high probability threshold were 0.801(95%CI: 0.786-0.818) and 0.923 (95%CI: 0.915-0.932), respectively.”
aurocas written: “auc”0.948CI 0.943-0.952
source quote (p.7)
“The EchoSolv AS model achieved a native system performance of 0.948 (95% CI: 0.943-0.952) AUROC.”
aurocas written: “AUROC (unassisted reads)”0.865CI 0.837-0.893
source quote (p.9)
“AUROC for unassisted and assisted reads were 0.865 (95%CI: 0.837-0.893) and 0.883 (95%CI: 0.857-0.909), respectively.”
aurocas written: “AUROC (assisted reads)”0.883CI 0.857-0.909
source quote (p.9)
“AUROC for unassisted and assisted reads were 0.865 (95%CI: 0.837-0.893) and 0.883 (95%CI: 0.857-0.909), respectively.”
aurocas written: “Mean AUROC improvement (assisted vs unassisted)”0.018CI 0.037-0.001
source quote (p.9)
“When cardiologist readers were provided with EchoSolv AS to assist with their interpretation of a TTE, there was an improvement in all study endpoints: mean AUROC (0.018±0.010, 95%CI: 0.037-0.001; p=0.064).”
agreement_kappaas written: “Kappa (unassisted reads)”0.641CI 0.597-0.685
source quote (p.9)
“Reader concordance (agreement) was evaluated using Fleiss' Kappa. Kappa for unassisted and assisted reads were 0.641 (95%CI: 0.597-0.685) and 0.667 (95%CI: 0.623-0.711), respectively.”
agreement_kappaas written: “Kappa (assisted reads)”0.667CI 0.623-0.711
source quote (p.9)
“Reader concordance (agreement) was evaluated using Fleiss' Kappa. Kappa for unassisted and assisted reads were 0.641 (95%CI: 0.597-0.685) and 0.667 (95%CI: 0.623-0.711), respectively.”
Each value carries its own analysis unit and task — never compare or pool across devices. Source: 510(k) summary PDF.
Predicate network
Postmarket — what happened after clearance
0
recalls in product code, 24mo
0
MAUDE reports in code, 12mo
—
vs code's own 3-yr baseline
0
drift signals on this device
Recall and MAUDE counts are product-code-level (reports aren't reliably attributable to one device). Signals are descriptive observables with sources — never a judgment that the device is unsafe or drifting. Snapshot 2026-07-08.
Reimbursement — how devices like this got paid
Not yet tracked — no payment pathway indexed for this clearance (the reimbursement corpus is a growing seed set).