Personalized biophysical modeling of the heart is a useful approach for noninvasively analyzing and predicting in vivo cardiac mechanics. Three main developments support this style of analysis: state-of-the-art cardiac imaging technologies, modern computational infrastructure, and advanced mathematical modeling techniques. In vivo measurements of cardiac structure and function can be integrated using sophisticated computational methods to investigate mechanisms of myocardial function and dysfunction, and can aid in clinical diagnosis and developing personalized treatment. In this article, we review the state-of-the-art in cardiac imaging modalities, model-based interpretation of 3D images of cardiac structure and function, and recent advances in modeling that allow personalized predictions of heart mechanics. We discuss how using such image-based modeling frameworks can increase the understanding of the fundamental biophysics behind cardiac mechanics, and assist with diagnosis, surgical guidance, and treatment planning. Addressing the challenges in this field will require a coordinated effort from both the clinical-imaging and modeling communities. We also discuss future directions that can be taken to bridge the gap between basic science and clinical translation.
Recognizing the limitations of computer-assisted tools for thyroid nodule diagnosis using static ultrasound images, this study developed a diagnostic tool utilizing dynamic ultrasound video, namely Thyroid Nodules Visualization (TNVis), by leveraging a two-stage deep learning framework that involved three-dimensional (3D) visualization. In this multicenter study, 4569 cases were included for framework development, and data from seven hospitals were employed for diagnostic validation. TNVis achieved a Dice similarity coefficient of 0.90 after internal testing. For the external validation, TNVis significantly improved radiologists’ performance, reaching an AUC of 0.79, compared to their diagnostic performance without the use of TNVis (AUC: 0.66; p < 0.001) and those with partial assistance (AUC: 0.72; p < 0.001). In conclusion, the TNVis-assisted diagnostic strategy not only significantly improves the diagnostic ability of radiologists but also closely imitates their clinical diagnostic procedures and provides them with an objective 3D representation of the nodules for precise and personalized diagnosis and treatment planning.
MitraClip is the sole percutaneous device approved for functional mitral regurgitation (MR; FMR) but MR recurs in over one third of patients. As device-induced mechanical effects are a potential cause for MR recurrence, we tested the hypothesis that MitraClip increases leaflet stress and procedure-related strain in sub-valvular left ventricular (LV) myocardium in FMR associated with coronary disease (FMR-CAD). Simulations were performed using finite element models of the LV + mitral valve based on MRI of 5 sheep with FMR-CAD. Models were modified to have a 20% increase in LV volume (↑LV_VOLUME) and MitraClip was simulated with contracting beam elements (virtual sutures) placed between nodes in the center edge of the anterior (AL) and posterior (PL) mitral leaflets. Effects of MitraClip on leaflet stress in the peri-MitraClip region of AL and PL, septo-lateral annular diameter (SLAD), and procedure-related radial strain (Err) in the sub-valvular myocardium were calculated. MitraClip increased peri-MitraClip leaflet stress at end-diastole (ED) by 22.3±7.1 kPa (p<0.0001) in AL and 14.8±1.2 kPa (p<0.0001) in PL. MitraClip decreased SLAD by 6.1±2.2 mm (p<0.0001) and increased Err in the sub-valvular lateral LV myocardium at ED by 0.09±0.04 (p<0.0001)). Furthermore, MitraClip in ↑LV_VOLUME was associated with persistent effects at ED but also at end-systole where peri-MitraClip leaflet stress was increased in AL by 31.9±14.4 kPa (p = 0.0268) and in PL by 22.5±23.7 kPa (p = 0.0101). MitraClip for FMR-CAD increases mitral leaflet stress and radial strain in LV sub-valvular myocardium. Mechanical effects of MitraClip are augmented by LV enlargement.
This retrospective study evaluated the efficacy of large language models (LLMs) in improving the accuracy of Chinese ultrasound reports. Data from three hospitals (January-April 2024) including 400 reports with 243 errors across six categories were analyzed. Three GPT versions and Claude 3.5 Sonnet were tested in zero-shot settings, with the top two models further assessed in few-shot scenarios. Six radiologists of varying experience levels performed error detection on a randomly selected test set. In zero-shot setting, Claude 3.5 Sonnet and GPT-4o achieved the highest error detection rates (52.3% and 41.2%, respectively). In few-shot, Claude 3.5 Sonnet outperformed senior and resident radiologists, while GPT-4o excelled in spelling error detection. LLMs processed reports faster than the quickest radiologist (Claude 3.5 Sonnet: 13.2 s, GPT-4o: 15.0 s, radiologist: 42.0 s per report). This study demonstrates the potential of LLMs to enhance ultrasound report accuracy, outperforming human experts in certain aspects.