ZACH-ViT: A Zero-Token Vision Transformer with ShuffleStrides Data Augmentation for Robust Lung Ultrasound Classification

2025-10-20

AI Summary

Distinguishing cardiogenic pulmonary edema (CPE), non-cardiogenic pathologies (e.g., ARDS-like or interstitial lung disease), and normal lung tissue in lung ultrasound (LUS) videos remains challenging due to severe visual heterogeneity and high overlap between B-lines and pleural artifacts. To address this, we propose a lightweight, permutation-invariant Vision Transformer that eliminates positional encoding and the [CLS] token, adopting a zero-token hierarchical architecture. We further introduce ShuffleStrides—a novel data augmentation technique explicitly designed for probe-scan sequences—to enhance generalization under limited data. Evaluated on 380 clinical LUS videos, our model achieves a ROC-AUC of 0.79 (sensitivity: 0.60; specificity: 0.91), trains 1.35× faster than prior approaches, and uses only 40% of the parameters of current state-of-the-art models. To our knowledge, this is the first method enabling fully order-agnostic modeling of medical ultrasound video inputs.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Differentiating cardiogenic pulmonary oedema (CPE) from non-cardiogenic and structurally normal lungs in lung ultrasound (LUS) videos remains challenging due to the high visual variability of non-cardiogenic inflammatory patterns (NCIP/ARDS-like), interstitial lung disease, and healthy lungs. This heterogeneity complicates automated classification as overlapping B-lines and pleural artefacts are common. We introduce ZACH-ViT (Zero-token Adaptive Compact Hierarchical Vision Transformer), a 0.25 M-parameter Vision Transformer variant that removes both positional embeddings and the [CLS] token, making it fully permutation-invariant and suitable for unordered medical image data. To enhance generalization, we propose ShuffleStrides Data Augmentation (SSDA), which permutes probe-view sequences and frame orders while preserving anatomical validity. ZACH-ViT was evaluated on 380 LUS videos from 95 critically ill patients against nine state-of-the-art baselines. Despite the heterogeneity of the non-cardiogenic group, ZACH-ViT achieved the highest validation and test ROC-AUC (0.80 and 0.79) with balanced sensitivity (0.60) and specificity (0.91), while all competing models collapsed to trivial classification. It trains 1.35x faster than Minimal ViT (0.62M parameters) with 2.5x fewer parameters, supporting real-time clinical deployment. These results show that aligning architectural design with data structure can outperform scale in small-data medical imaging.

Problem

• Classifying lung ultrasound videos for pulmonary edema diagnosis
• Addressing visual variability in non-cardiogenic inflammatory patterns
• Developing permutation-invariant models for unordered medical imaging data

Innovation

• Zero-token Vision Transformer removes positional embeddings
• ShuffleStrides augmentation permutes sequences and frames
• Compact 0.25M-parameter model trains faster with fewer parameters