AdURA-Net: Adaptive Uncertainty and Region-Aware Network

TL;DR

AdURA-Net is a novel adaptive uncertainty-aware neural network designed for reliable thoracic disease classification, effectively handling uncertain labels in complex medical imaging datasets to improve clinical decision-making.

Contribution

The paper introduces AdURA-Net, combining adaptive dilated convolutions, multiscale deformable alignment, and a dual head loss to model uncertainty and anatomical complexity in medical images.

Findings

Improved handling of uncertain labels in thoracic disease classification.

Enhanced model confidence calibration for clinical applications.

Effective capture of anatomical complexities in medical images.

Abstract

One of the common issues in clinical decision-making is the presence of uncertainty, which often arises due to ambiguity in radiology reports, which often reflect genuine diagnostic uncertainty or limitations of automated label extraction in various complex cases. Especially the case of multilabel datasets such as CheXpert, MIMIC-CXR, etc., which contain labels such as positive, negative, and uncertain. In clinical decision-making, the uncertain label plays a tricky role as the model should not be forced to provide a confident prediction in the absence of sufficient evidence. The ability of the model to say it does not understand whenever it is not confident is crucial, especially in the cases of clinical decision-making involving high risks. Here, we propose AdURA-Net, a geometry-driven adaptive uncertainty-aware framework for reliable thoracic disease classification. The key highlights of the proposed model are: a) Adaptive dilated convolution and multiscale deformable alignment coupled with the backbone Densenet architecture capturing the anatomical complexities of the medical images, and b) Dual Head Loss, which combines masked binary cross entropy with logit and a Dirichlet evidential learning objective.