Lili Zhao

This study demonstrates that oral acetazolamide significantly reduces opioid-associated sleep-disordered breathing (SDB) by reducing the plant gain, but without concurrently altering chemoresponsiveness or cerebrovascular responsiveness to carbon dioxide. This study is the first attempt to develop personalized therapies to mitigate breathing instability and SDB related to the ubiquitous use of opioid drugs.

Background and Aims: Falls are common and morbid for patients with cirrhosis, and patients have identified reducing fall risk as an unmet need and care priority. Approach and Results: This is a 24-week two-stage Sequential, Multiple Assignment Randomized Trial (SMART) at 4 centers in Michigan, Pennsylvania, and Texas enrolling 230 adults with cirrhosis and portal hypertension without prior overt hepatic encephalopathy (HE). There were 2 stages. In stage 1 (weeks 1–12), participants were randomized to lactulose or enhanced usual care (education on nutrition and fall prevention). In stage 2 (weeks 13–24), participants were randomized to TeleTai-Chi or enhanced usual care (exercise education). We evaluated the effectiveness of lactulose with or without TeleTai-Chi. The primary outcome was a hierarchical composite of death or transplantation, injurious falls, incident HE, and non-injurious falls. We used an intention-to-treat (ITT) win ratio (WR). Lactulose followed by TeleTai-Chi, compared with enhanced usual care, reduced injurious falls (4% vs. 12%) and non-injurious falls (19% vs. 32%), with no differences in overt HE (2% vs. 0%) or death/transplantation (0% vs. 0%). Compared with non-users, lactulose users experienced fewer injurious falls (3.5% vs. 8.5%), non-injurious falls (15.0% vs. 26.5%), and overt HE (1.8% vs. 2.6%), with no difference in death/transplantation or adverse events, resulting in a WR 2.3, 95% CI (1.3–4.0). Sequential lactulose/TeleTai-Chi reduced the risk of the primary outcome compared with usual care (WR 2.7, 95% CI (1.3–5.5). Conclusion: In patients with cirrhosis and portal hypertension without prior overt HE, lactulose reduced falls and fall-related harms. Sequential lactulose followed by TeleTai-Chi was also superior to enhanced usual care.

Objective. Pulmonary embolism (PE) has a high mortality rate, making early diagnosis critical for patient survival. While computed tomography (CT) pulmonary angiography effectively aids diagnosis and embolus localization, patients with iodine allergies and kidney disease experience delayed diagnosis and/or adverse effects. This exploratory preclinical study qualitatively describes PE-induced hemodynamic changes observable in CT perfusion imaging (CT-P), a computationally derived surrogate for perfusion imaging derived from non-contrast four-dimensional CT (4DCT), in a small pilot cohort. Approach . Two laboratory mice were imaged with a micro-4DCT scanner before and after the introduction of a synthetic PE, created using radiopaque alginate beads to enable embolus localization in 4DCT images. CT-P images were computed by segmenting the right and left lungs at the maximum inhale/exhale phase and quantifying the local tissue displacements using deformable image registration. Pre- and post-embolism CT-P heterogeneity was compared as a function of proximity to the embolism in the axial, coronal, and sagittal planes. CT-ventilation (CT-V) images were also computed using the Integrated Jacobian method to investigate co-localized ventilation-perfusion changes. Main results. In this two-animal pilot study, pre- and post-embolism imaging enabled observations not feasible in human studies. CT-P was observed to increase locally at the embolism site but decreased distally within the same lobe. Other lobes within the affected lung showed reduced perfusion, while the contralateral lung showed a compensatory increase mirroring the embolism location. Local perfusion variance near the embolism appeared greater in the unaffected regions. A lobe-wise CT-V/CT-P mismatch of 12.79% and 13.60% was observed pre- and post-embolization. Significance. As a hypothesis-generating preclinical study, these results demonstrate the methodological potential of CT-P to capture embolism-associated hemodynamic changes, motivating future studies with larger sample sizes.

BACKGROUND: Adrenal vein (AV) sampling (AVS) is used to guide therapy in primary aldosteronism (PA). When a single AV is successfully cannulated, the relative aldosterone secretion index (RASI), which compares the aldosterone/cortisol ratio in that AV versus the periphery, has been proposed as sufficient for PA subtyping, particularly when <1. Data on RASI reliability have, however, been inconsistent. METHODS: This retrospective cohort study included patients with PA who underwent AVS before and after cosyntropin stimulation at a referral center between January 2015 and December 2024. To simulate partially successful AVS, RASI was calculated in patients with successful bilateral AV cannulation and compared across PA subtypes and postoperative outcomes, with assessment of distributional overlap. RESULTS: Of 460 patients (mean age 53±12 years; 58% men), bilateral AVS was successful in 437 patients at baseline and in all patients after cosyntropin stimulation. Without cosyntropin, 98% of dominant AV and 97% of nondominant AV RASI from lateralized PA overlapped with bilateral PA RASI. Similar patterns were observed postcosyntropin. In adrenalectomized patients, RASI did not distinguish between those with and without PA cure. Previously proposed RASI thresholds misclassified up to 74% of lateralized PA and 64% of bilateral PA. When corroborated with cross-sectional imaging, the prediction of correct lateralization improved, particularly when using RASI from cosyntropin-stimulated AVS. CONCLUSIONS: Considering the substantial overlap of RASI across PA subtypes, partially successful AVS has limited utility and is unreliable in guiding adrenalectomy for PA.

Purpose: Distinguishing high-risk intraductal papillary mucinous neoplasms (IPMNs) from low-risk lesions remains a clinical challenge, often resulting in unnecessary procedures due to limited specificity of current methods. While radiomics and deep learning (DL) have been explored for pancreatic cancer, cyst-level malignancy risk stratification of IPMNs remains untapped. Methods: Our multi-institutional assessed the feasibility of AI for predicting IPMN dysplasia grade using cyst-level image features using 359 T2-weighted (T2W) MRI images from seven centers. We developed and compared 2D and 3D radiomics-only, DL-only, and radiomics-DL fusion models using expert radiologist scoring as a baseline reference. Model performance was evaluated using held-out test data. Results: The radiomics-DL fusion model showed the highest discriminatory ability on the test set AUC of 69.2%, outperforming the radiomics-only model, AUC of 66.5%. Expert accuracy varied widely from 37.4% to 66.7%, and the inter-rater agreement varied as well with weighted Cohen’s kappa coefficients of 0.33–0.67. Conclusion: The fusion model, which combines DL with radiomics features from routine T2W MRI, shows promise for objective, cyst-level risk stratification of IPMNs in a multi-center cohort, outperforming radiomics-only models and nearly matching expert radiologists using only T2W and T1-weighted (T1W) sequences. While performance requires improvement for standalone clinical use, this approach offers a scalable, non-invasive method to potentially improve diagnostic accuracy and reduce unnecessary surgical interventions.

Background: The clinical applicability of large language models (LLMs) in high-stakes cardiac emergencies remains unexplored. This study evaluated how well advanced LLMs perform in managing complex catheterization laboratory (Cath lab) scenarios and compared their performance with that of interventional cardiologists. Methods and Results: A cross-sectional study was conducted from 20 June to 2 December 2024. Twelve challenging inferior myocardial infarction scenarios were presented to seven LLMs (ChatGPT, Gemini, LLAMA, Qwen, Bing, Claude, DeepSeek) and five early-career interventional cardiologists. Responses were standardized, anonymized, and evaluated by thirty experienced interventional cardiologists. Performance comparisons were analyzed using a linear mixed-effects model with correlation and reliability statistics. Physicians had an average reference score of 80.68 (95% CI 76.3–85.0). Among LLMs, ChatGPT ranked highest (87.4, 95% CI 82.5–92.3), followed by Claude (80.8, 95% CI 75.7–85.9) and DeepSeek (78.7, 95% CI 72.9–84.6). LLAMA (73.7), Qwen (66.2), and Bing (64.3) ranked lower, while Gemini scored the lowest (59.0). ChatGPT scored higher than the early-career physician comparator group (difference 6.69, 95% CI 0.00–13.37; p < 0.05), whereas Gemini, LLAMA, Qwen, and Bing performed significantly worse; Claude and DeepSeek showed no significant difference. Conclusions: This expanded assessment reveals significant variability in LLM performance. In this simulated setting, ChatGPT demonstrated performance comparable to that of early-career interventional cardiologists. These results suggest that LLMs could serve as supplementary decision-support tools in interventional cardiology under simulated conditions.

Pancreatic cancer, projected to become the second-deadliest malignancy in Western societies by 2030, requires urgent innovations in early detection and risk stratification. Intraductal papillary mucinous neoplasms (IPMNs) represent critical precursor lesions, but current clinical guidelines demonstrate suboptimal accuracy in malignancy prediction, leading to either unnecessary surgeries or missed opportunities for early intervention. In this paper, we present Cyst-X, an artificial intelligence (AI) framework that accurately predicts IPMN malignant transformation using multicenter magnetic resonance imaging (MRI) data. Unlike most previous approaches that rely on computed tomography (CT), our method capitalizes on the superior soft tissue contrast of MRI, allowing for more precise identification of subtle imaging biomarkers. We developed and validated deep learning models on 723 T1-weighted and 738 T2-weighted MRI scans from 764 patients among seven international institutions, demonstrating significantly superior performance (AUC=0.82) compared to current clinical (Kyoto) guidelines (AUC=0.75) and expert radiologists. AI-based imaging features correlate strongly with clinically recognized malignancy markers, providing potential biologically relevant insights. This approach holds promise to significantly refine clinical decision-making, reduce unnecessary surgeries, and better identify high-risk IPMN patients for timely intervention. Our approach integrates a novel pancreas segmentation network with robust classification models that identify subtle imaging biomarkers associated with malignancy risk. Importantly, we demonstrate that these models retain high performance in a privacy-preserving federated learning setting, where institutions collaboratively train AI models without exchanging patient data to address key regulatory and ethical barriers. We publicly release the Cyst-X dataset--the first large-scale, multi-center pancreatic cyst MRI collection--to accelerate research in this field. This study addresses a critical clinical need while establishing technical foundations for privacy-preserving AI in radiology that could transform pancreatic cancer management through earlier intervention and reduced unnecessary procedures. The dataset can be accessed at https://osf.io/74vfs/, and the source code for our deep learning segmentation and classification models is available at https://github.com/NUBagciLab/Cyst-X.

Background/Objectives: Preclinical animal studies and clinical case reports have shown evidence of histotripsy being capable of inducing anti-tumor immune responses strong enough to inhibit tumor growth of off-target tumors. Previous studies exploring histotripsy immune stimulation have used a single therapy dose. This study aims to explore how histotripsy dose affects local tumor immune stimulation in a murine melanoma model. Methods: C57BL/6J mice bearing subcutaneous B16F10 tumors were treated with histotripsy using an ultrasound-guided 8-element 1 MHz transducer operating at a 100 Hz pulse repetition frequency (PRF) and >30 MPa peak-negative pressure. The histotripsy dose was defined by the number of pulses (8, 20, 40, or 100) per focal location (ppl). Tissue damage and residual tissue structure were measured histologically and scored by a trained pathologist. The longitudinal effect of histotripsy dosing was assessed using tumor growth and survival. Acute immune stimulation was measured at days 2 and 7 post-treatment via immunofluorescence staining of the treated tumor. Results: Histotripsy doses at 20, 40, and 100 ppl achieved significant tumor necrosis within the target region (>75%), with residual structure decreasing as the dose increased. Overall, the greatest tumor control was observed in mice that received the 40 ppl dose compared to untreated mice. This correlates with the 40 ppl dose also having the largest increase in CD45+ immune cells and CD8⁺ T cells 7 days post-treatment compared to untreated mice. Conclusions: The effect of histotripsy dosing on immune infiltration and tumor growth highlights the significant impact of histotripsy dose on clinical effects.

Amyotrophic lateral sclerosis (ALS) is a rare and fatal neurodegenerative disease with a median survival of only 2 to 4 years from diagnosis. Improved tools are needed to shorten diagnostic delays and improve prognostication to benefit clinical care. Herein, we profiled whole blood gene expression by RNA sequencing in a large cohort of ALS participants (n = 422) versus controls (n = 272). Several machine learning classifiers trained on our detailed gene expression dataset accurately predicted case-control status, including in a fully independent external test cohort, achieving an area under the receiver operating characteristic curve of 0.894 with the best performing model. Integrating gene expression features with clinical variables improved our ability to discriminate ALS cases into shorter, intermediate, and longer survival in an external dataset. Finally, we identified ALS-relevant pathways in our blood transcriptomics dataset as well as “core genes” that overlapped with gene expression changes occurring in the primary disease tissue, facilitating a drug perturbation analysis that identified several candidates. Overall, our results highlight the potential diagnostic and prognostic applications of whole blood gene expression data, with important implications for improving ALS clinical care.