Search

Radiology reports are an instrumental part of modern medicine, informing key clinical decisions such as diagnosis and treatment. The worldwide shortage of radiologists, however, restricts access to expert care and imposes heavy workloads, contributing to avoidable errors and delays in report delivery. While recent progress in automated report generation with vision-language models offer clear potential in ameliorating the situation, the path to real-world adoption has been stymied by the challenge of evaluating the clinical quality of AI-generated reports. In this study, we build a state-of-the-art report generation system for chest radiographs, $\textit{Flamingo-CXR}$, by fine-tuning a well-known vision-language foundation model on radiology data. To evaluate the quality of the AI-generated reports, a group of 16 certified radiologists provide detailed evaluations of AI-generated and human written reports for chest X-rays from an intensive care setting in the United States and an inpatient setting in India. At least one radiologist (out of two per case) preferred the AI report to the ground truth report in over 60$\%$ of cases for both datasets. Amongst the subset of AI-generated reports that contain errors, the most frequently cited reasons were related to the location and finding, whereas for human written reports, most mistakes were related to severity and finding. This disparity suggested potential complementarity between our AI system and human experts, prompting us to develop an assistive scenario in which Flamingo-CXR generates a first-draft report, which is subsequently revised by a clinician. This is the first demonstration of clinician-AI collaboration for report writing, and the resultant reports are assessed to be equivalent or preferred by at least one radiologist to reports written by experts alone in 80$\%$ of in-patient cases and 60$\%$ of intensive care cases.

Polyborosiloxane (PBS) is an important rate-stiffening soft polymer with dynamic, reversible crosslinks used in applications ranging from self-healing sensing and actuation to body and structural protection. Its highly rate-dependent response, especially for impact-mitigating structures, is important. However, the large strain response of PBS has not been characterized over quasi-static to high strain rates. Currently, there are no constitutive models that can predict the strongly rate-dependent large-deformation elastic-viscoplastic response of PBS. To address this gap, we have developed a microstructural physics motivated constitutive model for PBS and similar soft polymers and polymer gels with dynamic crosslinks to predict their large strain, non-linear loading-unloading, and significantly rate-dependent response. We have conducted compression experiments on PBS up to true strains of $\sim$125$\%$ over a wide strain rate range of 10$^{-3}$ s$^{-1}$ to 10$^{3}$ s$^{-1}$. The model reasonably accurately captures the response of PBS over six decades of strain rates. We propose boron-oxygen coordinate-bond dynamic crosslinks with macroscopic relaxation timescale $\tau \approx 3$ s and temporary entanglement lockups at high strain rates acting as crosslinks with $\tau \approx 0.0005$ s as the two types of crosslink mechanisms in PBS. We have outlined a numerical update procedure to evaluate the convolution-like time integrals arising from dynamic crosslink kinetics. Experiments involving three-dimensional inhomogeneous deformations were used to verify the predictive capabilities of our model and its finite element implementation. The modeling framework can be adopted for other dynamically crosslinked rate-stiffening soft polymers and polymer gels that are microstructurally similar to PBS.