arXiv subjects

Marco Allinovi

Publications and source records attributed to Marco Allinovi.

4 recordsLinked to original sources

arXiv2026

CPSNet: Physics-Inspired Label-Free Deep Unfolding for Lung Ultrasound B-Line Detection

In this paper, we propose CPSNet, a label-free deep unfolding framework for lung ultrasound image analysis. CPSNet unfolds a Cauchy proximal splitting algorithm into a forward-backward style network architecture, incorporating skip connections to iteratively enhance noisy Radon domain images. We introduce the Radon-Based Neighbor Reconstruction Loss, a novel loss function that enforces reconstruction consistency between subsampled and reconstructed image pairs in the Radon domain, while applying regularization terms to enhance robustness against noise. Integrating the Cauchy penalty into the loss function preserves the statistical influence of the prior while enabling the network to learn flexibly. Trained in an unsupervised manner without ground truth images, CPSNet is evaluated using structural similarity index, Proxy peak signal-to-noise Ratio, and relative L2 norm, demonstrating stable and effective performance in lung ultrasound image reconstruction. Applied to B-line detection, CPSNet achieves greater stability, adaptability and efficiency compared to traditional and object detection methods, effectively preserving line structures and minimizing false detections. This study underscores CPSNet's potential as a reliable and efficient solution for lung ultrasound-based diagnostics, supporting more accurate and efficient clinical decision-making. The code developed for this paper is available at https://github.com/TQ-001/DUCPS.

eess.IV↗

arXiv2024

A Multimodal Approach for Fluid Overload Prediction: Integrating Lung Ultrasound and Clinical Data

Managing fluid balance in dialysis patients is crucial, as improper management can lead to severe complications. In this paper, we propose a multimodal approach that integrates visual features from lung ultrasound images with clinical data to enhance the prediction of excess body fluid. Our framework employs independent encoders to extract features for each modality and combines them through a cross-domain attention mechanism to capture complementary information. By framing the prediction as a classification task, the model achieves significantly better performance than regression. The results demonstrate that multimodal models consistently outperform single-modality models, particularly when attention mechanisms prioritize tabular data. Pseudo-sample generation further contributes to mitigating the imbalanced classification problem, achieving the highest accuracy of 88.31%. This study underscores the effectiveness of multimodal learning for fluid overload management in dialysis patients, offering valuable insights for improved clinical outcomes.

eess.IV↗

arXiv2023

DUBLINE: A Deep Unfolding Network for B-line Detection in Lung Ultrasound Images

In the context of lung ultrasound, the detection of B-lines, which are indicative of interstitial lung disease and pulmonary edema, plays a pivotal role in clinical diagnosis. Current methods still rely on visual inspection by experts. Vision-based automatic B-line detection methods have been developed, but their performance has yet to improve in terms of both accuracy and computational speed. This paper presents a novel approach to posing B-line detection as an inverse problem via deep unfolding of the Alternating Direction Method of Multipliers (ADMM). It tackles the challenges of data labelling and model training in lung ultrasound image analysis by harnessing the capabilities of deep neural networks and model-based methods. Our objective is to substantially enhance diagnostic accuracy while ensuring efficient real-time capabilities. The results show that the proposed method runs more than 90 times faster than the traditional model-based method and achieves an F1 score that is 10.6% higher.

eess.IV↗

arXiv2023

A Semi-supervised Learning Approach for B-line Detection in Lung Ultrasound Images

Studies have proved that the number of B-lines in lung ultrasound images has a strong statistical link to the amount of extravascular lung water, which is significant for hemodialysis treatment. Manual inspection of B-lines requires experts and is time-consuming, whilst modelling automation methods is currently problematic because of a lack of ground truth. Therefore, in this paper, we propose a novel semi-supervised learning method for the B-line detection task based on contrastive learning. Through multi-level unsupervised learning on unlabelled lung ultrasound images, the features of the artefacts are learnt. In the downstream task, we introduce a fine-tuning process on a small number of labelled images using the EIoU-based loss function. Apart from reducing the data labelling workload, the proposed method shows a superior performance to model-based algorithm with the recall of 91.43%, the accuracy of 84.21% and the F1 score of 91.43%.

eess.IV↗