AarXiv6 Dec 2021Cited 1×

Hybrid guiding: A multi-resolution refinement approach for semantic segmentation of gigapixel histopathological images

André Pedersen, Erik Smistad, Tor V. Rise, Vibeke Grotnes Dale, Henrik Sahlin Pettersen, Tor-Arne S. Nordmo, David Bouget, Ingerid Reinertsen, Marit Valla

Abstract

segmentation of gigapixel histopathological images Authors: ['André Pedersen' 'Erik Smistad' 'Tor V. Rise' 'Vibeke Grotnes Dale' 'Henrik Sahlin Pettersen' 'Tor-Arne S. Nordmo' 'David Bouget' 'Ingerid Reinertsen' 'Marit Valla'] In this study, we used 624 whole slide images (WSIs) of breast cancer from a Norwegian cohort. We propose a cascaded convolutional neural network design, called H2G-Net, for semantic segmentation of gigapixel histopathological images. The design involves a detection stage using a patch-wise method, and a refinement stage using a convolutional autoencoder. To validate the design, we conducted an ablation study to assess the impact of selected components in the pipeline on tumour segmentation.

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

A cascaded CNN (H2G-Net) with detection and refinement stages achieves a Dice score of 0.933 for breast cancer whole‑slide image segmentation, outperforming single‑resolution methods.

What’s new

i.Proposed H2G‑Net, a cascaded CNN with detection and autoencoder refinement stages
ii.Introduced hierarchical sampling and deep heatmap refinement for better segmentation
iii.Demonstrated a random sampling scheme that balances multiple labels without storing patches
iv.Showed CPU‑only inference of a gigapixel slide in ~58 seconds

How they did it

The study used 624 breast‑cancer whole slide images from a Norwegian cohort. A patch‑wise detection network first generated tumor heatmaps, which were then refined by a convolutional autoencoder using hierarchical sampling. Performance was evaluated via an ablation study and an independent test set of 90 WSIs.

What they found

Best design achieved Dice = 0.933 on 90‑image test set
Outperformed MobileNetV2 patch‑wise high‑resolution (Dice = 0.872) and low‑resolution U‑Net (Dice = 0.874)
Segmentation of a representative x400 WSI took ~58 seconds on CPU

Caveats

✕Future work needed for more efficient patch generation and sampling
✕Improved clustering methods are still required

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Hybrid guiding: A multi-resolution refinement approach for semantic segmentation of gigapixel histopathological images

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

The paper, in five minutes

Similar papers in all four archives