SubtleHD (1.x)

K243250

Subtle Medical, Inc. · cleared 2025-02-12 · product code QIH · Radiology

Premarket evidence — what FDA accepted

Device typesamd

source quote (p.6)

“SubtleHD is Software as a Medical Device (SaMD) consisting of a software algorithm that enhances images taken by MRI scanners.”

Algorithmconvolutional neural network based filtering, cascade of filter banks, single neural network trained for adaptive noise reduction and sharpness increase, image-guided optimization process, nonlocal mean based denoising, unsharp masking based sharpening filters, deep learning processed image

source quote (p.6)

“SubtleHD software implements an image enhancement algorithm using a convolutional neural network based filtering. Original images are enhanced by running through a cascade of filter banks, where thresholding and scaling operations are applied. A single neural network is trained for adaptive noise reduction and sharpness increase. The parameters within the neural network were obtained through an image-guided optimization process. Additional nonlocal mean based denoising and unsharp masking based sharpening filters are applied to the deep learning processed image.”

Adaptive (vs locked)No

source quote (p.9)

“The algorithm will be locked with fixed model parameters prior to release.”

PCCPYes

source quote (p.2)

“FDA's substantial equivalence determination also included the review and clearance of your Predetermined Change Control Plan (PCCP).”

Cybersecurity addressedFDA source did not state this

Validation studies (4)

Bench

sample size not stated

endpoints: demonstrate that software requirements are implemented

Standalone

n=471 other

endpoints: evaluating L1 loss; SSIM; PSNR

Retrospective clinical

n=410 images

endpoints: Denoising (SNR); Sharpness (Image Intensity Slope Change); Sharpness (Image Intensity Change for Brains); Sharpness and Over Smoothing (Gradient Entropy)

Reader study (MRMC)

n=410 images

endpoints: Denoising (Signal-to-Noise Ratio); Overall Image Quality / Diagnostic Confidence; Visibility of Small Structures; Artifact Introduction

Reported performance (0 observations)

FDA source did not state a quantitative performance metric — non-reporting is itself the signal.

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

recalls in product code, 24mo

MAUDE reports in code, 12mo

—

vs code's own 3-yr baseline

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).