Calibrating Multi-modal Representations: A Pursuit of Group Robustness without Annotations.

Fine-tuning pre-trained vision-language models, like CLIP, has yielded success on diverse downstream tasks. However, several pain points persist for this paradigm: (i) directly tuning entire pre-trained models becomes both time-intensive and computationally costly. Additionally, these tuned models tend to become highly specialized, limiting their practicality for real-world deployment; (ii) recent studies indicate that pre-trained vision-language classifiers may overly depend on spurious features - patterns that correlate with the target in training data, but are not related to the true labeling function; and (iii) existing studies on mitigating the reliance on spurious features, largely based on the assumption that we can identify such features, does not provide definitive assurance for real-world applications.

Prior knowledge-guided vision-transformer-based unsupervised domain adaptation for intubation prediction in lung disease at one week.

Data-driven approaches have achieved great success in various medical image analysis tasks. However, fully-supervised data-driven approaches require unprecedentedly large amounts of labeled data and often suffer from poor generalization to unseen new data due to domain shifts. Various unsupervised domain adaptation (UDA) methods have been actively explored to solve these problems.

Breast Cancers Detected during a Decade of Screening with Digital Breast Tomosynthesis: Comparison with Digital Mammography.

Background Digital breast tomosynthesis (DBT) has been shown to help increase cancer detection compared with two-dimensional digital mammography (DM). However, it is unclear whether additional tumor detection will improve outcomes or lead to overdiagnosis of breast cancer. Purpose This study aimed to compare cancer types and stages over 3 years of DM screening and 10 years of DBT screening to determine the effect of DBT.

Learning Sequential Information in Task-based fMRI for Synthetic Data Augmentation.

Insufficiency of training data is a persistent issue in medical image analysis, especially for task-based functional magnetic resonance images (fMRI) with spatio-temporal imaging data acquired using specific cognitive tasks. In this paper, we propose an approach for generating synthetic fMRI sequences that can then be used to create augmented training datasets in downstream learning tasks. To synthesize high-resolution task-specific fMRI, we adapt the -GAN structure, leveraging advantages of both GAN and variational autoencoder models, and propose different alternatives in aggregating temporal information.

First Results of Pediatric Robotic Inguinal Hernia Repair with the Senhance Surgical System: A Matched Cohort Study.

Introduction: Inguinal hernia repair (IHR) is one of the most common procedures in pediatric surgery. In children, the application of robotic surgery is limited, meaning safety and efficacy is still to be assessed. This report is the first one worldwide that describes inguinal hernia repair in children using the Senhance Surgical System (SSS).