mlsanityData to 7 Oct 2026

Focus on What Matters: Two-Stage ROI-Aware Refinement for Anatomy-Preserving Fetal Ultrasound Reconstruction

arXiv 2604.23839 ↗First posted 26 Apr 2026cs.CV cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Measurement-critical ultrasound tasks often depend on a small anatomical region, making global reconstruction metrics an unreliable proxy for clinical fidelity. We propose an ROI-aware representation learning framework and instantiate it for first-trimester nuchal translucency (NT) screening under multi-hospital domain shift. A two-phase convolutional autoencoder (CAE) first learns a globally faithful 128-D latent code via MS-SSIM, then refines the NT ROI using intensity (L1) and normalized Sobel-edge constraints. To combine these heterogeneous objectives without manual tuning, we initialize loss weights via gradient-based calibration from per-term gradient magnitudes. Under strict hospital-wise evaluation with one hospital held out, ROI refinement improves both global and measurement-relevant quality: on the standard dev split it increases PSNR by +0.27 dB (val) and +0.29 dB (held-out test), reduces ROI MAE by 8.87% (val) and 6.43% (held-out test), and reduces ROI Edge-MAE by 11.10% on source hospitals and 4.90% on the unseen hospital. Beyond reconstruction, frozen-latent probes provide additional evidence of generalization: hospital provenance becomes less confidently predictable on the unseen site (0.556 to 0.541 max-softmax; 0.684 to 0.688 entropy) while OOD detection remains strong across site-held-out protocols (Mahalanobis AUROC up to 0.9956, with modest KNN gains in challenging splits). The same ROI-aware refinement principle is anatomy-agnostic and can be adopted for other fetal biometry targets (e.g., crown-rump length (CRL), nasal bone (NB)) and broader medical imaging settings where small ROIs dominate clinical decisions.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 35 of its 75 references that are in the map. First posted on arXiv 26 Apr 2026.

Builds on

35 of 75 references.

References · 35

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  2. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,510
  3. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,678
  4. 2016V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentationcited by 2,284
  5. 2021TransUNet: Transformers Make Strong Encoders for Medical Image Segmentationcited by 1,552
  6. 2018UNet++: A Nested U-Net Architecture for Medical Image Segmentationcited by 1,518
  7. 20163D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotationcited by 1,471
  8. 2018Attention U-Net: Learning Where to Look for the Pancreascited by 1,420
  9. 2016Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoisingcited by 1,352
  10. 2017Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semanticscited by 1,345
  11. 2017GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networkscited by 816
  12. 2018Multi-Task Learning as Multi-Objective Optimizationcited by 809
23 more
  1. 2017Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentationscited by 614
  2. 2013Network In Networkcited by 600
  3. 2024YOLOv9: Learning What You Want to Learn Using Programmable Gradient Informationcited by 422
  4. 2021Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysiscited by 418
  5. 2021Score-based diffusion models for accelerated MRIcited by 308
  6. 2022Diffusion Models for Medical Image Analysis: A Comprehensive Surveycited by 269
  7. 2021TransFuse: Fusing Transformers and CNNs for Medical Image Segmentationcited by 258
  8. 2017Tversky loss function for image segmentation using 3D fully convolutional deep networkscited by 228
  9. 2020A survey of loss functions for semantic segmentationcited by 216
  10. 2021UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformercited by 197
  11. 2018Boundary loss for highly unbalanced segmentationcited by 187
  12. 2019Reducing the Hausdorff Distance in Medical Image Segmentation with Convolutional Neural Networkscited by 116
  13. 2019SoftAdapt: Techniques for Adaptive Loss Weighting of Neural Networks with Multi-Part Loss Functionscited by 38
  14. 2019USE-Net: incorporating Squeeze-and-Excitation blocks into U-Net for prostate zonal segmentation of multi-institutional MRI datasetscited by 27
  15. 2024Centerline Boundary Dice Loss for Vascular Segmentationcited by 27
  16. 2024Scalable Multitask Learning Using Gradient-based Estimation of Task Affinitycited by 20
  17. 2022blob loss: instance imbalance aware loss functions for semantic segmentationcited by 18
  18. 2024Jacobian Descent for Multi-Objective Optimizationcited by 18
  19. 2024ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancementcited by 9
  20. 2025LDC-MTL: Balancing Multi-Task Learning through Scalable Loss Discrepancy Controlcited by 7
  21. 2025Automated Fetal Biometry Assessment with Deep Ensembles using Sparse-Sampling of 2D Intrapartum Ultrasound Imagescited by 3
  22. 2023Ultrasound Image Enhancement using CycleGAN and Perceptual Losscited by 2
  23. 2024MultiBalance: Multi-Objective Gradient Balancing in Industrial-Scale Multi-Task Recommendation Systemcited by 2

Built on by

Not cited yet.