Skip to content
PaperFren

Can a neural network recognise which heart-ultrasound view it's seeing?

Open paper intelligence

A convolutional neural network named 15 standard echocardiogram views more accurately than board-certified experts, using tiny low-resolution images.

Source

Fast and accurate view classification of echocardiograms using deep learning

Madani A, Arnaout R, Mofrad M, et al. · NPJ digital medicine · 2018

doi.org/10.1038/s41746-017-0013-1Read the full paper ↗288 citationscc by

Study at a glance

Design
Computational / modelling — Supervised CNN trained on 15-view labelled echocardiogram frames, with train/validation/test split by patient study, a random-label negative control, and a comparison with 4 human experts.
N
N=267 · Echocardiogram studies from different patients: 240 for training and validation (213 train, 27 validation) plus 27 for testing, yielding 223,787 labelled frames; 4 board-certified echocardiographers each classified 1500 test images.
Population
Randomly selected clinical transthoracic echocardiograms from UCSF, 2000-2017, adults aged 20-96
Outcome
Per-view and overall classification accuracy, F-score and AUC on held-out images and videos; comparison with expert accuracy

Structured fields used in claim comparison tables when every cited study has a complete layer.

What they did

The authors took 240 randomly chosen clinical echocardiogram studies for training and validation and 27 more for testing, had an expert label every clip by view, and shrank each frame to a 60 × 80-pixel greyscale image. They trained a VGG-inspired convolutional network on 15 views, keeping training, validation and test images from different patients. They checked the model against random-label training, compared it with 4 board-certified echocardiographers on the same test images, and used occlusion and saliency maps to see which parts of the image it relied on.

What they found

On videos (classified by majority vote across frames) overall test accuracy was 97.8%. On single still images it was 91.7%, compared with 79.4% for the human experts on a subset of the same images. Training on randomly shuffled labels gave 6.9% accuracy, roughly chance for 15 classes. Errors mostly involved genuinely similar views such as apical two- versus three-chamber, and masking the heart itself, but not other image regions, sharply reduced accuracy.

The limits

What it doesn't show

All data came from one medical centre, labelled by a single echocardiographer, so there is no external validation or measure of label noise. The human comparison used only 4 experts viewing the same shrunken low-resolution stills the model saw, a format they do not normally work from, which may understate human skill. View classification is only a first step; the paper does not show the model improves diagnosis or patient outcomes.

Key terms

Convolutional neural network (CNN)
A neural network that learns small image filters, stacked in layers, to recognise visual patterns.
Data augmentation
Randomly rotating, shifting, zooming or flipping training images so the model learns features robust to these changes.
Negative control (random labels)
Training on shuffled labels to confirm that performance falls to chance and is not caused by a leak or artefact.
Occlusion test
Masking parts of an image to see which regions the model needs to classify it correctly.
Saliency map
A heat map of the input pixels that most influence the model's output.

Flashcards

1 / 11

0 of 11 answers reviewed

Research intelligence for this paper

See its role on concept claims, tensions it is part of, placement history, and related discoveries.

Open paper intelligence

Quiz yourself

1 / 6

What architecture inspired the model?

Common questions

Why keep images from the same patient out of both training and test sets?

Frames from one study look alike; mixing them would let the model score well by recognising the patient rather than the view, inflating test accuracy.

Did the model really beat the experts?

On single low-resolution stills, yes, but the experts normally use full-resolution video, so the comparison favoured the model's format.

Why is video accuracy higher than still-image accuracy?

A video is classified by a majority vote over many frames, so occasional single-frame mistakes are outvoted.

More on Convolutional networks