MedSaab-US: A Backpropagation-Free Multi-Scale Wavelet-Saab Framework for Thyroid Nodule Segmentation in Ultrasound Images

2026-07-02 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Medical Imaging AI · Depth: Expert, quick

Summary

MedSaab-US is a novel backpropagation-free framework designed for thyroid nodule segmentation in ultrasound images, addressing the high parameter counts and GPU dependency of deep learning methods. Grounded in the Green Learning paradigm, MedSaab-US extracts multi-scale spatial-frequency features by integrating multi-level Discrete Wavelet Transform (DWT) with multi-scale channel-wise Saab transforms, using patch sizes of 5 x 5, 11 x 11, and 21 x 21 pixels. It employs Label-Assisted Greedy (LAG) feature selection to identify discriminative features, which are then classified pixel-wise by an XGBoost model. Evaluated on the TN3K dataset (2,879 training, 614 test images), MedSaab-US achieved a mean Dice coefficient of 0.4784 +/- 0.2190. The framework boasts a model footprint under 500K parameters and performs CPU-only inference in approximately 0.3 seconds per image, offering an efficient non-DL baseline.

Key takeaway

For Machine Learning Engineers deploying medical image segmentation in resource-constrained environments, MedSaab-US presents a compelling alternative to deep learning. Its backpropagation-free design, low parameter count (under 500K), and CPU-only inference (0.3 seconds per image) make it highly efficient. You should consider this Green Learning-based framework to achieve robust thyroid nodule segmentation without the heavy computational demands of traditional DL models, especially when GPU resources are limited.

Key insights

MedSaab-US provides an efficient, backpropagation-free alternative for thyroid nodule segmentation, overcoming deep learning's resource demands.

Principles

• Green Learning enables resource-efficient models.
• Analytical parameter derivation reduces training complexity.
• Iterative greedy construction bypasses backpropagation.

Method

MedSaab-US combines multi-level DWT with multi-scale Saab transforms for feature extraction. Label-Assisted Greedy (LAG) selects discriminative features, which an XGBoost classifier then uses for pixel-wise prediction.

In practice

• Deploy in resource-constrained medical settings.
• Use as a non-DL baseline for segmentation.
• Enable rapid CPU-only inference for diagnostics.

Topics

• Thyroid Nodule Segmentation
• Ultrasound Imaging
• Green Learning
• Wavelet Transform
• Saab Transform
• XGBoost

Best for: Computer Vision Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Learning a Sampling-Free Variational DNN Plugin from Tiny Training Sets to Refine OOD Segmentation With Uncertainty Estimation — VarDeepPCA refines OOD medical image segmentation using intrinsic geometric priors and sampling-free variational learning from tiny in-distribution datasets.
• SegMix:Shuffle-based Feedback Learning for Semantic Segmentation of Pathology Images — Shuffle-based feedback learning improves weakly supervised semantic segmentation in pathology by adapting to multi-scale image characteristics.
• Cross-Domain Feature Expansion for Tabular Medical Data via Knowledge Graphs Injection — MedKGTab expands tabular medical features by combining data-driven statistics with biomedical knowledge graphs for high-fidelity generation.
• Federated orthogonal learning for detection of liver lesions from multi-phase contrast-enhanced CT images — FedOG enhances federated learning for medical image segmentation by using orthogonal gradient decomposition to counter data heterogeneity and incomplete phases.
• SegWithU: Uncertainty as Perturbation Energy for Single-Forward-Pass Risk-Aware Medical Image Segmentation — SegWithU provides efficient, reliable uncertainty for medical image segmentation via a single-forward-pass perturbation energy model.
• Multi-layer feature aggregation network with residual module and attention mechanism for jaw cyst image segmentation — MFA-Net enhances medical image segmentation by integrating residual modules, channel-spatial attention, and multi-layer feature aggregation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.