Nature Biomedical Engineering

Accurate polyp segmentation from colonoscopy images is critical for colorectal cancer prevention, yet the generalization of deep learning models under domain shift remains insufficiently explored. We propose Boundary-Explicit Guided Attention U-Net (BEGA-UNet), a boundary-aware segmentation architecture that introduces explicit edge modeling as a structural inductive bias to enhance both segmentation accuracy and cross-domain robustness. The framework integrates three components: an Edge-Guided Module (EGM) with learnable Sobel-initialized operators to capture boundary cues, a Dual-Path Attention (DPA) module that processes channel and spatial attention in parallel, and a Multi-Scale Feature Aggregation (MSFA) module to encode contextual information across multiple receptive fields. Evaluated on the combined Kvasir-SEG and CVC-ClinicDB benchmarks, BEGA-UNet achieves 88.53% Dice and 82.51% IoU, outperforming representative convolutional and transformer-based baselines. More importantly, cross-dataset evaluation demonstrates strong robustness under domain shift, with BEGA-UNet retaining 83.2% of its in-distribution performance--substantially higher than U-Net (64.5%), Attention U-Net (47.5%), and TransUNet (53.1%). In a zero-shot setting on an entirely unseen dataset, the model further maintains 72.6% performance retention. Comprehensive ablation studies indicate that explicit boundary modeling plays a central role in improving generalization, while multi-scale context aggregation further stabilizes performance across domains. Feature distribution analyses support this observation by showing that edge-oriented representations exhibit markedly reduced cross-domain variability compared to appearance-driven features. Overall, BEGA-UNet provides an effective and interpretable solution for robust polyp segmentation, demonstrating that explicit boundary modeling serves as a critical inductive bias for ensuring reliability under clinical domain shifts.

Most routine clinical markers are interpreted using population-based reference intervals, despite being regulated around patient-specific homeostatic setpoints. This mismatch obscures physiologic shifts, inhibiting detection of early disease signatures. Here, we develop a novel Bayesian inference method that adaptively constructs personalized reference intervals using each patients existing health records. In analysis of >100 million lab tests in >800,000 patients, these personalized intervals can be accurately constructed with only minimal prior data, meaning this method can be applied near universally. We show that across 43 common lab markers, patient setpoints are strongly associated with future morbidity, with signal strength increasing as more test data is collected. Deviation from personalized reference intervals provides strong and novel risk signatures across diverse disease states, including hypothyroidism, hematologic cancers, kidney disease, and pregnancy complications. Importantly, personalized reference intervals capture a different risk signature to existing population-based approaches, with the highest risk patients being those who deviate from both intervals simultaneously. In a targeted clinical use case study of iron infusion, use of personalized reference intervals greatly improved prediction of treatment efficacy and allowed precise tracking of treatment responses. Our results illustrate how existing health records can be used to construct personalized benchmarks for nearly all common clinical tests, driving a new paradigm for precision laboratory medicine.

Deep learning models utilizing longitudinal healthcare data have significantly advanced epidemiological research. However, contemporary transformer-based models increasingly rely on computationally intensive pre-training steps that entail processing massive real-world datasets with cost-prohibitive hardware. We introduce the Temporal Encoder with Late Fusion (TELF), a lightweight end-to-end predictive model featuring an encoder-only architecture for processing medical codes, followed by post-encoder concatenation with demographic variables. TELF learns code embeddings on-the-fly, thereby bypassing the resource-intensive pre-training bottleneck. Furthermore, its late-fusion design preserves the integrity of the temporal attention mechanism before integrating static demographic predictors. We evaluated TELF using an administrative claims database across three distinct cohorts: pancreatic cancer (n=53,661), type 2 diabetes (n=78,756), and heart failure (n=72,540). TELF consistently outperformed traditional machine learning baselines, including XGBoost, LightGBM, and logistic regression. Specifically, TELF achieved AUCs of 0.9150, 0.8199, and 0.8721 for pancreatic cancer, type 2 diabetes, and heart failure, respectively, compared with 0.9044, 0.7908, and 0.8535 for XGBoost and 0.9014, 0.7800, and 0.8466 for logistic regression. Beyond predictive superiority, TELF's isolated temporal attention mechanism enables population-level motif mining. By extracting high-attention temporal sequences, we mapped aggregated patient journey pathways, revealing interpretable clinical trajectories preceding disease onset. Collectively, these results demonstrate that TELF provides a resource-efficient and accessible framework for advanced temporal modeling in clinical and epidemiological research.

We developed an integrated platform combining high-throughput automated biofabrication, systematic patient-derived tissue experiments, and specialized artificial intelligence to enable patient-specific computational "digital twins" for thrombosis prediction. Our automated manufacturing platform fabricates 80 fully assembled, patient-specific vessel-on-chips within 10 hours from clinical imaging--a [~]100-fold improvement over manual methods--achieving sub-micron precision through novel two-stage pneumatic motion control and integrated optical feedback. Using these chips, we systematically captured thrombosis across 491 high-fidelity videos spanning 6 patient-derived vascular geometries, 5 distinct anatomical injury sites, and 14 anticoagulant/antiplatelet interventions, establishing a "physical twin" experimental corpus. We trained CLoT (Cascade Learner of Thrombosis), a conditional video diffusion model efficiently adapted via lightweight Low-Rank Adaptation (LoRA) to generate realistic thrombosis videos conditioned on patient-specific geometry, injury location, and drug treatment. Rigorous benchmarking against state-of-the-art commercial models (Sora, Wan, Kling, Seedance, Hailuo, Hunyuan) reveals CLoT achieves 7.38-fold superior temporal biological consistency and 5.3-fold higher spatial morphological fidelity. Prospective validation on unseen patients demonstrates >90% temporal accuracy. This integrated paradigm--combining automated fabrication with domain-specialized generative AI--establishes proof-of-concept for personalized medicine enabled by digital twins trained on human-derived vascular anatomy, enabling pre-treatment antithrombotic evaluation while providing a replicable template for translating tissue engineering into clinical practice.

Understanding how cardiomyocyte structure governs contractile function is fundamental to cardiac biology and disease modeling, yet current approaches rely on time-resolved imaging and computationally intensive analysis. Here, we present a generative artificial intelligence (AI) framework that directly predicts contractile behavior of human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) from single static images. Our approach integrates a U-Net-based generator with a patch-based generative adversarial network (GAN) discriminator to translate morphological and sarcomere structural features into pixel-resolved contraction heatmaps. This U-Net-GAN model achieved high predictive accuracy, with structural similarity index (SSIM) values up to 0.84 using combined morphological and structural inputs. To further enhance performance and generalizability, we incorporated synthetic cell-function pairs generated via a generative AI StyleGAN2 framework, improving prediction accuracy and perceptual similarity. Importantly, region-specific and whole-cell analyses revealed that AI predictions capture biologically meaningful structure-function relationships, with sarcomere organization strongly associated with both contractile output and prediction fidelity. Reconstruction error emerged as an interpretable metric reflecting localized inefficiencies in sarcomere-to-contraction coupling. Together, this framework establishes a scalable and interpretable strategy for inferring cardiomyocyte function from static morphology, eliminating the need for time-lapse imaging. More broadly, this work positions generative AI as a powerful tool for bridging cellular structure and function, enabling high-throughput functional phenotyping and advancing in vitro cardiac modeling.

Vocal biomarkers, encompassing voice and speech, have largely been developed for individual conditions in isolation, limiting their generalizability across diseases and recording settings. To address this, we introduce VoiceFM, a contrastive model that learns general-purpose clinical voice representations by aligning audio embeddings with rich clinical metadata. Using the Bridge2AI-Voice dataset (984 primarily English-speaking adult participants, 846 used for training and 138 held out as a temporally separated validation cohort, 40,056 recordings totaling 176 hours across 5 academic medical centers), VoiceFM pairs a fine-tuned Whisper large-v2 encoder with a tabular transformer over 44 clinical features via symmetric InfoNCE loss. Linear probes on frozen VoiceFM embeddings achieve mean AUROC 0.952 +/- 0.005 across five evaluation tasks (control vs disease screening plus four disease categories), significantly outperforming Frozen Whisper (0.926 +/- 0.013, p = 0.013), Frozen HuBERT (0.885 +/- 0.017, p = 0.0009), and the contrastively trained VoiceFM-HuBERT (0.938 +/- 0.006, p = 0.012). On the 138-participant held-out cohort, VoiceFM-Whisper achieves AUROCs of 0.99 for Alzheimer's/dementia/MCI and 0.89 for airway stenosis, demonstrating that the learned representations generalize to participants the model has never seen. VoiceFM representations transfer to three external datasets without retraining and improve few-shot classification. Recording task attribution identifies a small set of speech tasks that match or exceed the full battery's performance, suggesting shorter screening protocols are feasible. Trained predominantly on English audio, VoiceFM transfers without fine-tuning to Spanish-language Parkinson's disease (PD) detection (NeuroVoz, 107 participants, AUROC 0.93 +/- 0.02), with the signal dominated by articulatory rather than phonatory features. A fine-tuned classifier achieves participant-level AUROC 0.87 (sustained 0.85, countdown 0.80) on the mPower smartphone study (585 held-out participants). Together, these results show that contrastive alignment between voice and rich clinical metadata can serve as the basis for a clinical voice foundation model, producing a single set of transferable representations that generalize across diseases, languages, recording conditions, and patients enrolled after model freeze.

Glaucoma phenotyping remains challenging due to disease heterogeneity and single-modality limitations. We introduce a visual field (VF)-guided entangled learning framework that integrates structural and functional signals during training to learn functionally informed macular retinal nerve fiber layer (mRNFL) representations while enabling OCT-only inference. In 5,372 paired MEEI examinations, VF-guided phenotyping identified 9 clinically distinct mRNFL phenotypes with divergent progression rates (MD slopes -0.2 to -1.8 dB/year, P <0.001), improving clustering over OCT-only by 22% (FCM) and 11% (GMM). External evaluation in 74,077 UK Biobank images confirmed generalizability, with improved risk association (r=-0.33 vs r=0.04). Genetic analyses identified 12 additional glaucoma loci compared with OCT-only phenotyping. VF-guided entangled learning improves clinically and genetically coherent mRNFL phenotyping with broad applicability to multimodal medical imaging.

Although GLP1/GIP receptor agonists demonstrate unprecedented weight loss efficacy, their rapid clinical adoption has revealed significant real-world tolerability challenges. To evaluate their dynamic safety profiles, we developed a macro to micro pharmacovigilance framework by combining global FAERS reports with local UT Physician EHR. Macroscopically, we distilled 17 shared adverse events across the drug class from FAERS with disproportionality analysis. Microscopically, local EHR data (289,655 longitudinal treatment sessions across 71,316 patients) revealed 51.6% of GLP1 sessions terminated within 90 days. Furthermore, temporal stratified logistic regression demonstrated that initial exposure (0 to 30 days) correlated strongly with nausea and vomiting, which attenuated in extended sessions, whereas extended exposure (>2 years) uncovered late onset risks, notably incident hepatic steatosis. Ultimately, this time aware framework reveals that GLP1 safety profiles are profoundly duration dependent, providing critical insights into both acute intolerances and long-term medication safety.

Estimating heterogeneous treatment effects (HTE) from observational health data is essential for precision medicine, yet existing methods often struggle with high-dimensional covariates and time-to-event outcomes common in electronic health records (EHRs). We propose SurvTEDVAE, a disentangled variational autoencoder designed for causal survival analysis. The model learns latent representations corresponding to instrumental factors, confounders, and outcome-dependent risk factors, and integrates a survival likelihood to model time-to-event outcomes with censoring. The learned representations are used to estimate conditional average treatment effects using downstream causal estimators. We evaluated SurvTEDVAE using a semi-synthetic ACTG dataset and a high-dimensional EHR-based hypertension cohort with over 20,000 covariates. Across both datasets, SurvTEDVAE achieved lower estimation error for heterogeneous treatment effects compared with meta-learning and causal survival forest approaches. These results demonstrate that disentangled representation learning can improve causal effect estimation for survival outcomes in high-dimensional real-world health data.

Neurosurgical and neuromodulation therapies such as deep brain stimulation (DBS) require millimeter-level accuracy to effectively target functional brain regions. Yet, many neuroanatomical boundaries remain invisible to current imaging and electrophysiology methods, limiting precision and contributing to suboptimal patient outcomes. Here, we introduce a self-supervised artificial intelligence (AI) framework that learns to delineate functional subregions directly from the spectral content of intraoperative local field potential (LFP) recordings, without the need for predefined biomarkers or anatomical labels. The framework identifies physiologic structure across the full spectrum of the signal and, through explainable AI (XAI), reveals the specific frequency components underlying these distinctions. Validated in the subthalamic nucleus (STN), the model aligned with clinically defined borders and rediscovered known beta oscillations. Applied to the motor thalamus in tremor patients, it consistently identified functional transitions corresponding to the ventral oralis posterior (Vop) and ventral intermediate (Vim) nuclei--regions where conventional methods fail to provide reliable boundaries. To assess clinical relevance, physiologically defined clusters were functionally evaluated using monopolar review data at their first DBS clinic postoperative visit, demonstrating distinct stimulation-response profiles across clusters and linking electrophysiologic segmentation to clinically meaningful programming outcomes. These findings demonstrate that intraoperative LFP recordings can be transformed into both a real-time guidance resource and a data-driven platform for biomarker discovery, establishing a foundation for more precise, individualized neuromodulation therapies and advancing our understanding of functional brain organization.

Massively multiplexed qPCR is primarily constrained by increasing primer dimer formation as the number of distinct primers in a single reaction increases. Previous multiplex primer design algorithms either fail to sufficiently suppress primer dimers at 100+ plex, or take exceedingly high amounts of computational resources to complete. Here, we present DIMPLE, a linear-runtime primer design algorithm that effectively generates 10,000+ primers to amplify thousands of potential amplicons in a single qPCR reaction. As one clinical demonstration of this algorithm, we designed an assay to detect 2,302 distinct KMT2A gene fusion subtypes using 204 primers in a single tube. In contrast to FISH and convention NGS approaches with 2% variant allele frequency (VAF) limit of detection, our DIMPLE qPCR assay was able to analytically detect gene fusions down to 0.05% VAF. We also constructed proof-of-concept multiplex qPCR panels for additional oncology gene fusions, multiplex pathogen detection, and DNA methylation markers. The scalability and low computational cost DIMPLE are complementary to new instrument platforms for massively multiplex qPCR readout for enabling rapid, point-of-care nucleic acid testing.

Synthetic data generation is a promising approach for biomedical data sharing and dataset augmentation, yet existing methods lack mechanisms to preserve statistical properties necessary for scientific analysis. To address this, we introduce RLSYN+REG, a reinforcement learning-driven generative model, which encourages that regression models trained on synthetic data reproduce the coefficients and predictions of their real-data counterparts. We evaluate RL-SO_SCPLOWYNC_SCPLOW+RO_SCPLOWEGC_SCPLOW on MIMIC-III and the American Community Survey (ACS) across regression model reproduction, fidelity to real data, and privacy. Synthetic data from RLSO_SCPLOWYNC_SCPLOW+RO_SCPLOWEGC_SCPLOW substantially improves upon that of RLSO_SCPLOWYNC_SCPLOW, raising correlations between real and synthetic regression coefficients from 0.054 to 0.600 on MIMIC-III and from 0.160 to 0.376 on ACS. Predictive performance also improves, reducing the gap between real-data baselines by 81.4% and 97.6% on MIMIC-III and ACS, respectively. These improvements come with negligible cost to fidelity or privacy and are robust to reductions in training data.

Angiogenesis-promoting macromolecules reduce adverse remodeling and preserve cardiac function in rodents following myocardial infarctions, yet repeatedly fail to translate across length scales in humans. Through mass transport studies in human and swine myocardium, we found that dense, anisotropic myocardial fibers limit therapeutic diffusion and convection to millimeter scales for existing approaches including bolus intramyocardial injections, shear-thinning hydrogels, and epicardial patches. Furthermore, distributions are confined to one dimension along fibers. To increase myocardial drug distribution to centimeter length scales in vivo in swine, we engineered a three-dimensional multi-injection drug delivery array. Our device performs up to 40 simultaneous 120 {micro}L injections of functional macromolecules, hydrogels, or mRNA lipid nanoparticles. Injections are precisely placed in relation to fiber alignment, achieving near-complete coverage of the left ventricular myocardium.

New implantable-electrode fabrication strategies make dense, ultrafine electrode arrays with lower tissue burden increasingly feasible, shifting a key bottleneck for scalable brain-computer interfaces from electrode placement to readout capacity. Electrode pooling, in which multiple electrodes share a readout channel, could relax this bottleneck by combining extracellular signals before acquisition, but it has remained unclear whether such compression preserves the neural population structure needed for behavioral decoding. Here we evaluate this question using software-emulated electrode pooling in mouse sensorimotor cortex during a cue-guided reach-and-grasp task using a high-density microwire array coupled to a CMOS microelectrode-array platform. Pooled recordings retain forelimb kinematic information more effectively than a channel-matched control that discards electrodes. Pooling reduces the separability of electrode-specific spikes and sorted units, indicating a loss of some neuronal detail, but the mixed signals still preserve task-aligned low-dimensional latent dynamics that support decoding. When readout capacity is fixed, this trade-off allows broader electrode coverage to contribute to behaviorally informative population sampling. Together, these results define electrode pooling as a design trade-off for scalable readout, in which some electrode-specific neuronal information is lost but the population dynamics needed for movement decoding remain accessible.

Large-scale clinical genome sequencing yields vast numbers of variants of unknown significance (VUSs). The high frequency of VUSs and the paucity of platforms to characterize their functional impact pose significant challenges for clinical decision making. Here, we present an integrated end-to-end platform, REVi-SCOPE (Rapid evaluation of variants in single cells by optogenetics and prime editing), for characterization of the impact of VUSs on cardiac physiology. Our strategy consists of (1) introduction of variants directly into wild-type (WT) human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) via prime editing; (2) optogenetic assessment of calcium and membrane voltage dynamics in single hiPSC-CMs within the pool of edited and unedited cells; and (3) in situ single-cell genotyping of the phenotyped hiPSC-CMs with single-allele resolution. By optimizing and integrating each of these steps, we created a platform that enables VUS characterization in 10 days. We validated the REVi-SCOPEs capabilities by analyzing the properties of established arrhythmogenic variants. We then used REVi-SCOPE to reveal the functional impact of a VUS, TRPM4A320V, identified in a child with a conduction block. Together, our results show that REVi-SCOPE enables functional characterization of VUSs linked to cardiac arrhythmias with unprecedented throughput.

Selecting an effective treatment relies on accurately anticipating patient's response to alternative interventions. However, forecasting longitudinal clinical trajectories remains difficult because electronic health records contain heterogeneous, irregularly sampled data over extended time periods. These issues are especially relevant for laboratory measurements, which are central for diagnostics, assessment of therapeutic responses, and tracking disease progression in routine clinical practice. However, existing deep learning methods for counterfactual prediction usually assume regularly sampled data, an assumption incompatible with the irregular, heterogeneous data-generation processes of real-world clinical practice. Here we present the Time-Aware G-Transformer, which integrates causal G-computation with time-aware attention to predict counterfactual outcomes on irregular data. By explicitly conditioning on the timing of future observations and encoding measurement patterns, the model captures temporal dynamics that previous methods overlook. Evaluated on synthetic tumor growth data and on 90,753 cancer patient trajectories from an academic medical center, our approach demonstrates superior long-horizon (> 1 day) prediction accuracy and uncertainty calibration compared to state-of-the-art baselines. These results demonstrate that embedding temporal relations directly into the attention mechanism enables robust integration of patient history data for evaluating potential treatment strategies in personalized medicine.

Neuropathy caused by chemotherapy is a common and debilitating side-effect of cancer treatment. With 30% of patients experiencing chronic neuropathy and with no good evidence-based treatments; early detection triggering chemotherapy regime modification remains the best option for prevention. Early detection is challenging because of a lack of diagnostic tools with sufficient longitudinal temporal precision and convenience for patient/clinical adoption. To tackle this problem, we developed SenseCheQ; enabling self-administered autonomous sensory testing which can be used by patients at home. We present the instrumental engineering approach taken to address the challenge, including haptic self-calibration combined with skin thermal-clamping protocols, and demonstrate robustly reliable performance in the face of environmental and user-related variance in home settings. We present prospective case studies of people having chemotherapy treatment for cancer, conducting regular unsupervised quantitative sensory testing to monitor their nerve function at home. These proof-of-principle studies show SenseCheQ can detect sub-clinical changes in nerve function, matching patient reported outcomes and lab-based sensory testing. This highlights SenseCheQs promise as a scalable biomarker platform for neuropathy-detection and therapeutic development.

Sleep disturbances are pervasive, debilitating non-motor symptoms of Parkinson's disease (PD), where sleep spindle deficits directly drive cognitive decline and disease progression. Current adaptive deep brain stimulation (aDBS) for PD is largely limited to motor symptom management, with no established technical foundation for sleep spindle-targeted closed-loop modulation. The functional role of the basal ganglia in human sleep spindle regulation remains incompletely characterized, and no robust cross-subject pipeline exists to decode these transient events from clinically implanted DBS electrodes. Here, we developed a connectomics-guided meta-learning framework for cross-subject sleep spindle decoding and anticipatory prediction, using whole-night synchronized basal ganglia local field potential and polysomnography data from 17 PD patients with bilateral DBS implants. Our framework achieved 92.63% accuracy for concurrent spindle decoding and 83.44% accuracy for 2-second-ahead prediction, with optimal signals localized to the limbic subthalamic nucleus and <50 ms total latency meeting real-time closed-loop requirements. This work defines the neuroanatomical substrate of basal ganglia spindle signaling in PD, establishes the cross-subject spindle decoding pipeline for clinical DBS systems, and provides a critical translational foundation for sleep-targeted closed-loop aDBS to mitigate PD non-motor burden.

Tau protein aggregation in the brain is a hallmark of Alzheimers disease (AD). Positron emission tomography (PET) is the only in vivo method to visualize tau pathology and estimate both its burden and regional distribution, but the use of tau-PET is constrained by high cost and limited accessibility. Here, we develop a deep learning model to synthesize tau-PET scans from more accessible data: structural magnetic resonance imaging (MRI), demographics, and when available, blood biomarkers. We included 5,191 participants across the AD continuum or with another neurological disorder from 13 cohorts (mean age 70 years, 51% female) and optimized a 3D U-Net neural network with residual and attention units for this task. In held-out test data, synthetic tau-PET reliably modeled tau burden, with correlations of R=0.77-0.86 with true tau-PET across individuals in common AD regions of interest. Spatial similarity between synthetic and true tau-PET was likewise high, with mean regional correlation of R=0.75. Synthetic scans also captured clinically meaningful prognostic information comparable to true tau-PET, including distinction between early (HR=12, p<0.001) and late (HR=45, p<0.001) stages of tau accumulation. These findings demonstrate that clinically informative synthetic tau-PET scans can be generated from widely available modalities using deep learning, potentially offering a scalable and cost-effective approach for estimating tau AD pathology in the brain.

Forceful and coordinated contractions of the uterine myometrium are essential for successful labor, delivery, and postpartum uterine involution. Failure of the uterus to generate or sustain contractile force (uterine atony) after delivery results in postpartum hemorrhage, a leading cause of maternal mortality globally. Paradoxically, uterine atony is exacerbated by prolonged oxytocin exposure used to induce or augment labor through a process of contractile desensitization. Despite its prevalent use in obstetrics, the direct impact of oxytocin desensitization on myometrial contractile force generation remains poorly defined. Current model systems are inadequate to address this gap: ex vivo myometrial tissue strips are limited by tissue availability, donor variability, and lack of genetic tractability, while existing in vitro models provide only indirect readouts of contractility without direct force quantification. Here, we introduce engineered myometrial microtissues (EMMIs), a platform enabling the direct, isometric measurement of contractile force in response to physiological agonists like oxytocin. By embedding and molding immortalized human myometrial smooth muscle cells within a collagen hydrogel, we induced significant structural and molecular maturation over six days. Upon maturation, EMMIs were characterized by circumferential cellular alignment, sustained expression of smoothelin, upregulation of connexin-43, and a transcriptomic shift toward a contractile phenotype. Mature EMMIs generated calcium-sensitive, dose-dependent contractions to oxytocin and potassium chloride. Genetic deletion of the oxytocin receptor abolished oxytocin-induced contractility, establishing receptor specificity. Finally, we utilized EMMIs to recapitulate clinical oxytocin desensitization, providing a direct link between prolonged oxytocin exposure and diminished contractile output. Together, these findings establish engineered myometrial microtissues (EMMIs) as a genetically manipulable, and reproducible system for investigating myometrial contractile physiology to improve obstetric outcomes. TeaserEngineered 3D uterine tissues quantify how labor-inducing drugs weaken contractions and drive maternal hemorrhage