Incorporating simulated spatial context information improves the effectiveness of contrastive learning models.

Visual learning often occurs in a specific context, where an agent acquires skills through exploration and tracking of its location in a consistent environment. The historical spatial context of the agent provides a similarity signal for self-supervised contrastive learning. We present a unique approach, termed environmental spatial similarity (ESS), that complements existing contrastive learning methods. Using images from simulated, photorealistic environments as an experimental setting, we demonstrate that ESS outperforms traditional instance discrimination approaches. Moreover, sampling additional data from the same environment substantially improves accuracy and provides new augmentations. ESS allows remarkable proficiency in room classification and spatial prediction tasks, especially in unfamiliar environments. This learning paradigm has the potential to enable rapid visual learning in agents operating in new environments with unique visual characteristics. Potentially transformative applications span from robotics to space exploration. Our proof of concept demonstrates improved efficiency over methods that rely on extensive, disconnected datasets.

CD70-specific CAR-NK cells expressing IL-15 for the treatment of CD19-negative B cell malignancy.

Chimeric antigen receptor (CAR)-NK cells can eliminate tumors not only through the ability of the CAR molecule to recognize antigen expressed cancer cells but also through NK cell receptors themselves. This overcomes some of the limitations of CAR-T cells, paving CAR-NK cells for safer and more effective off-the-shelf cellular therapy. In this study, CD70, a pan-target of lymphoma, specific fourth-generation CAR with 4-1BB co-stimulatory domain and IL-15 was constructed and transduced into cord blood-derived NK cells by Baboon envelope pseudotyped lenti-vector. CD70-CAR NK cells displayed superior cytotoxic activity in vitro and in vivo against CD19 negative B-cell lymphoma when compared to non-transduced NK cells and CD19-specific CAR-NK cells. Importantly, mice received two doses of CD70-CAR NK cells showed effective eradication of tumors, accompanied by increased concentration of plasma IL-15 and enhanced CAR-NK cell proliferation and persistence. Our study suggests that repetitive administration-based CAR NK-cell therapy has clinical advantage compared to single dose of CAR-NK cells for the treatment of B-cell lymphoma.

ABCD of IA: A multi-scale agent-based model of T cell activation in inflammatory arthritis.

Biomaterial-based agents have been demonstrated to regulate the function of immune cells in models of autoimmunity. However, the complexity of the kinetics of immune cell activation can present a challenge in optimizing the dose and frequency of administration. Here, we report a model of autoreactive T cell activation, which are key drivers in autoimmune inflammatory joint disease. The model is termed a multi-scale Agent-Based, Cell-Driven model of Inflammatory Arthritis (ABCD of IA). Using kinetic rate equations and statistical theory, ABCD of IA simulated the activation and presentation of autoantigens by dendritic cells, interactions with cognate T cells and subsequent T cell proliferation in the lymph node and IA-affected joints. The results, validated with in vivo data from the T cell driven SKG mouse model, showed that T cell proliferation strongly correlated with the T cell receptor (TCR) affinity distribution (TCR-ad), with a clear transition state from homeostasis to an inflammatory state. T cell proliferation was strongly dependent on the amount of antigen in antigenic stimulus event (ASE) at low concentrations. On the other hand, inflammation driven by Th17-inducing cytokine mediated T cell phenotype commitment was influenced by the initial level of Th17-inducing cytokines independent of the amount of arthritogenic antigen. The introduction of inhibitory artificial antigen presenting cells (iaAPCs), which locally suppress T cell activation, reduced T cell proliferation in a dose-dependent manner. The findings in this work set up a framework based on theory and modeling to simulate personalized therapeutic strategies in IA.

Molecular recognition and activation of the prostacyclin receptor by anti-pulmonary arterial hypertension drugs.

The prostacyclin (PGI2) receptor (IP) is a Gs-coupled receptor associated with blood pressure regulation, allergy, and inflammatory response. It is a main therapeutic target for pulmonary arterial hypertension (PAH) and several other diseases. Here we report cryo-electron microscopy (cryo-EM) structures of the human IP-Gs complex bound with two anti-PAH drugs, treprostinil and MRE-269 (active form of selexipag), at global resolutions of 2.56 and 2.41 angstrom, respectively. These structures revealed distinct features governing IP ligand binding, receptor activation, and G protein coupling. Moreover, comparison of the activated IP structures uncovered the mechanism and key residues that determine the superior selectivity of MRE-269 over treprostinil. Combined with molecular docking and functional studies, our structures provide insight into agonist selectivity, ligand recognition, receptor activation, and G protein coupling. Our results provide a structural template for further improving IP-targeting drugs to reduce off-target activation of prostanoid receptors and adverse effects.

Spatial transcriptomic analysis drives PET imaging of tight junction protein expression in pancreatic cancer theranostics.

We apply spatial transcriptomics and proteomics to select pancreatic cancer surface receptor targets for molecular imaging and theranostics using an approach that can be applied to many cancers. Selected cancer surfaceome epithelial markers were spatially correlated and provided specific cancer localization, whereas the spatial correlation between cancer markers and immune- cell or fibroblast markers was low. While molecular imaging of cancer-associated fibroblasts and integrins has been proposed for pancreatic cancer, our data point to the tight junction protein claudin-4 as a theranostic target. Claudin-4 expression increased ∼16 fold in cancer as compared with normal pancreas, and the tight junction localization conferred low background for imaging in normal tissue. We developed a peptide-based molecular imaging agent targeted to claudin-4 with accumulation to ∼25% injected activity per cc (IA/cc) in metastases and ∼18% IA/cc in tumors. Our work motivates a new approach for data-driven selection of molecular targets.

The Influence of Etiology and Stimulation Target on the Outcome of Deep Brain Stimulation for Chronic Neuropathic Pain: A Systematic Review and Meta-Analysis.

OBJECTIVES: Deep brain stimulation (DBS) to treat chronic neuropathic pain has shown variable outcomes. Variations in pain etiologies and DBS targets are considered the main contributing factors, which are, however, underexplored owing to a paucity of patient data in individual studies. An updated meta-analysis to quantitatively assess the influence of these factors on the outcome of DBS for chronic neuropathic pain is warranted, especially considering that the anterior cingulate cortex (ACC) has emerged recently as a new DBS target. MATERIALS AND METHODS: A comprehensive literature review was performed in PubMed, Embase, and Cochrane data bases to identify studies reporting quantitative outcomes of DBS for chronic neuropathic pain. Pain and quality of life (QoL) outcomes, grouped by etiology and DBS target, were extracted and analyzed (α = 0.05). RESULTS: Twenty-five studies were included for analysis. Patients with peripheral neuropathic pain (PNP) had a significantly greater initial stimulation success rate than did patients with central neuropathic pain (CNP). Both patients with CNP and patients with PNP with definitive implant, regardless of targets, gained significant follow-up pain reduction. Patients with PNP had greater long-term pain relief than did patients with CNP. Patients with CNP with ACC DBS gained less long-term pain relief than did those with conventional targets. Significant short-term QoL improvement was reported in selected patients with CNP after ACC DBS. However, selective reporting bias was expected, and the improvement decreased in the long term. CONCLUSIONS: Although DBS to treat chronic neuropathic pain is generally effective, patients with PNP are the preferred population over patients with CNP. Current data suggest that ACC DBS deserves further investigation as a potential way to treat the affective component of chronic neuropathic pain.

Geographic Distribution of Facial Plastic and Reconstructive Surgeons.

Background: Distribution of the general otolaryngology workforce has been described, but not specifically for the facial plastic and reconstructive surgeon (FPRS) workforce. Objective: To describe the distribution of FPRS within the United States. Methods: The 2022 American Academy of Facial Plastic and Reconstructive Surgery (AAFPRS) registry was used to identify active FPRSs. Member addresses were converted into coordinates and overlayed onto a geographic representation of 2020 census data within ArcGIS software. A centroid model of U.S. counties was constructed to determine the average distances residents were from the nearest FPRS. Results: In total, 1312 AAFPRS active members practiced in 373 counties. Thirty-three percent of all residents (115 million) resided in counties without an FPRS and 15.3% of FPRSs practiced in New York City or the Greater Los Angeles Area, which accounted for 8% of the total U.S. population. The mean and median distances a resident in a county without an FPRS was from the nearest FPRS are 63 and 49 miles (101 and 79 kilometers), respectively. Conclusions: Metropolitan areas have greater concentrations of FPRSs than the national average and the distances U.S. residents are from FPRS services are quantifiable.

Programmable late-stage functionalization of bridge-substituted bicyclo[1.1.1]pentane bis-boronates.

Modular functionalization enables versatile exploration of chemical space and has been broadly applied in structure-activity relationship (SAR) studies of aromatic scaffolds during drug discovery. Recently, the bicyclo[1.1.1]pentane (BCP) motif has increasingly received attention as a bioisosteric replacement of benzene rings due to its ability to improve the physicochemical properties of prospective drug candidates, but studying the SARs of C2-substituted BCPs has been heavily restricted by the need for multistep de novo synthesis of each analogue of interest. Here we report a programmable bis-functionalization strategy to enable late-stage sequential derivatization of BCP bis-boronates, opening up opportunities to explore the SARs of drug candidates possessing multisubstituted BCP motifs. Our approach capitalizes on the inherent chemoselectivity exhibited by BCP bis-boronates, enabling highly selective activation and functionalization of bridgehead (C3)-boronic pinacol esters (Bpin), leaving the C2-Bpin intact and primed for subsequent derivatization. These selective transformations of both BCP bridgehead (C3) and bridge (C2) positions enable access to C1,C2-disubstituted and C1,C2,C3-trisubstituted BCPs that encompass previously unexplored chemical space.

Direct measurements of methane emissions from key facilities in Alberta's oil and gas supply chain.

The Canadian government aims to achieve a 40-45 % reduction of oil and gas (O&G) methane (CH4) emissions by 2025, and 75 % by 2030, although recent studies consistently show that Canada's federal inventory underestimates emissions by a factor of 1.4 to 2.0. We conducted aerial mass balance measurements at sixteen upstream O&G facilities in Alberta between September 29 and November 6, 2021, and our measurements revealed that emissions were, on average, 1.7 (standard deviation (SD): 0.6) times higher than the reported emissions for the same year. On a subsequent campaign from August 12 to September 27, 2022, we focused on understudied O&G sectors covering 24 midstream and end-use facilities. These sites were found to be emitting, on average, 3.4 (SD: 1.1) times more CH4 than reported. By extrapolating our measurements to Alberta, we found that underground gas storage contributed to 1.6 % of provincial O&G emissions, followed by natural gas power stations/refineries less than 1.0 %. The widespread underreporting of CH4 emissions highlights the necessity for more empirical measurements of midstream and end-use facilities.

Deep Rank-Consistent Pyramid Model for Enhanced Crowd Counting.

Most conventional crowd counting methods utilize a fully-supervised learning framework to establish a mapping between scene images and crowd density maps. They usually rely on a large quantity of costly and time-intensive pixel-level annotations for training supervision. One way to mitigate the intensive labeling effort and improve counting accuracy is to leverage large amounts of unlabeled images. This is attributed to the inherent self-structural information and rank consistency within a single image, offering additional qualitative relation supervision during training. Contrary to earlier methods that utilized the rank relations at the original image level, we explore such rank-consistency relation within the latent feature spaces. This approach enables the incorporation of numerous pyramid partial orders, strengthening the model representation capability. A notable advantage is that it can also increase the utilization ratio of unlabeled samples. Specifically, we propose a Deep Rank-consistEnt pyrAmid Model (), which makes full use of rank consistency across coarse-to-fine pyramid features in latent spaces for enhanced crowd counting with massive unlabeled images. In addition, we have collected a new unlabeled crowd counting dataset, FUDAN-UCC, comprising 4000 images for training purposes. Extensive experiments on four benchmark datasets, namely UCF-QNRF, ShanghaiTech PartA and PartB, and UCF-CC-50, show the effectiveness of our method compared with previous semi-supervised methods. The codes are available at https://github.com/bridgeqiqi/DREAM.

Mitigating Scar Tissue Formation in Tendon Injuries: Targeting HMGB1, AMPK Activation, and Myofibroblast Migration All at Once.

Tendon injuries, while prevalent, present significant challenges regarding their structural and functional restoration. Utilizing alpha-smooth muscle actin (α-SMA)-Ai9-scleraxis (Scx)-green fluorescent protein (GFP) transgenic mice, which exhibit both Scx (a tendon cell marker) and α-SMA (a myofibroblast marker), we explored the effects of metformin (Met) on tendon healing, repair, and its mechanisms of action. Our findings revealed that intraperitoneal (IP) injections of Met, administered before or after injury, as well as both, effectively prevented the release of HMGB1 into the tendon matrix and reduced circulating levels of HMGB1. Additionally, Met treatment increased and activated AMPK and suppressed TGF-ß1 levels within the healing tendon. Tendon healing was also improved by blocking the migration of α-SMA+ myofibroblasts, reducing the prevalence of disorganized collagen fibers and collagen type III. It also enhanced the presence of collagen type I. These outcomes highlight Met's anti-fibrotic properties in acutely injured tendons and suggest its potential for repurposing as a therapeutic agent to minimize scar tissue formation in tendon injuries, which could have profound implications in clinical practice.

PET imaging of focused-ultrasound enhanced delivery of AAVs into the murine brain.

Rationale: Despite recent advances in the use of adeno-associated viruses (AAVs) as potential vehicles for genetic intervention of central and peripheral nervous system-associated disorders, gene therapy for the treatment of neuropathology in adults has not been approved to date. The currently FDA-approved AAV-vector based gene therapies rely on naturally occurring serotypes, such as AAV2 or AAV9, which display limited or no transport across the blood-brain barrier (BBB) if systemically administered. Recently developed engineered AAV variants have shown broad brain transduction and reduced off-target liver toxicity in non-human primates (NHPs). However, these vectors lack spatial selectivity for targeted gene delivery, a potentially critical limitation for delivering therapeutic doses in defined areas of the brain. The use of microbubbles, in conjunction with focused ultrasound (FUS), can enhance regional brain AAV transduction, but methods to assess transduction in vivo are needed. Methods: In a murine model, we combined positron emission tomography (PET) and optical imaging of reporter gene payloads to non-invasively assess the spatial distribution and transduction efficiency of systemically administered AAV9 after FUS and microbubble treatment. Capsid and reporter probe accumulation are reported as percent injected dose per cubic centimeter (%ID/cc) for in vivo PET quantification, whereas results for ex vivo assays are reported as percent injected dose per gram (%ID/g). Results: In a study spanning accumulation and transduction, mean AAV9 accumulation within the brain was 0.29 %ID/cc without FUS, whereas in the insonified region of interest of FUS-treated mice, the spatial mean and maximum reached ~2.3 %ID/cc and 4.3 %ID/cc, respectively. Transgene expression assessed in vivo by PET reporter gene imaging employing the pyruvate kinase M2 (PKM2)/[18F]DASA-10 reporter system increased up to 10-fold in the FUS-treated regions, as compared to mice receiving AAVs without FUS. Systemic injection of AAV9 packaging the EF1A-PKM2 transgene followed by FUS in one hemisphere resulted in 1) an average 102-fold increase in PKM2 mRNA concentration compared to mice treated with AAVs only and 2) a 12.5-fold increase in the insonified compared to the contralateral hemisphere of FUS-treated mice. Conclusion: Combining microbubbles with US-guided treatment facilitated a multi-hour BBB disruption and stable AAV transduction in targeted areas of the murine brain. This unique platform has the potential to provide insight and aid in the translation of AAV-based therapies for the treatment of neuropathologies.

Deep generative model deciphers derailed trajectories in acute myeloid leukemia.

Single-cell genomics has the potential to map cell states and their dynamics in an unbiased way in response to perturbations like disease. However, elucidating the cell-state transitions from healthy to disease requires analyzing data from perturbed samples jointly with unperturbed reference samples. Existing methods for integrating and jointly visualizing single-cell datasets from distinct contexts tend to remove key biological differences or do not correctly harmonize shared mechanisms. We present Decipher, a model that combines variational autoencoders with deep exponential families to reconstruct derailed trajectories (https://github.com/azizilab/decipher). Decipher jointly represents normal and perturbed single-cell RNA-seq datasets, revealing shared and disrupted dynamics. It further introduces a novel approach to visualize data, without the need for methods such as UMAP or TSNE. We demonstrate Decipher on data from acute myeloid leukemia patient bone marrow specimens, showing that it successfully characterizes the divergence from normal hematopoiesis and identifies transcriptional programs that become disrupted in each patient when they acquire NPM1 driver mutations.

An Investigation into the Surface Integrity of Micro-Machined High-Speed Steel and Tungsten Carbide Cutting Tools.

The performance and lifespan of cutting tools are significantly influenced by their surface quality. The present report highlights recent advances in enhancing the surface characteristics of tungsten carbide and high-speed steel cutting tools using a novel micro-machining technique for polishing and edge-honing. Notably, the main aim was to reduce the surface roughness while maintaining the hardness of the materials at an optimal level. By conducting a thorough analysis of surfaces obtained using different techniques, it was found that the micro-machining method effectively decreased the surface roughness of the cutting tools the most effectively out of the techniques investigated. Significantly, the surface roughness was reduced from an initial measurement of 400 nm to an impressive value of 60 nm. No significant change in hardness was observed, which guarantees the maintenance of the mechanical properties of the cutting tools. This analysis enhances the comprehension of surface enhancement methodologies for cutting tools through the presentation of these findings. The observed decrease in surface roughness, along with the consistent hardness, exhibits potential for improving tool performance. These enhancements possess the capacity to optimise manufacturing processes, increase tool reliability, and minimise waste generation.

Bodily expressed emotion understanding through integrating Laban movement analysis.

Bodily expressed emotion understanding (BEEU) aims to automatically recognize human emotional expressions from body movements. Psychological research has demonstrated that people often move using specific motor elements to convey emotions. This work takes three steps to integrate human motor elements to study BEEU. First, we introduce BoME (body motor elements), a highly precise dataset for human motor elements. Second, we apply baseline models to estimate these elements on BoME, showing that deep learning methods are capable of learning effective representations of human movement. Finally, we propose a dual-source solution to enhance the BEEU model with the BoME dataset, which trains with both motor element and emotion labels and simultaneously produces predictions for both. Through experiments on the BoLD in-the-wild emotion understanding benchmark, we showcase the significant benefit of our approach. These results may inspire further research utilizing human motor elements for emotion understanding and mental health analysis.

Unlocking the Emotional World of Visual Media: An Overview of the Science, Research, and Impact of Understanding Emotion: Drawing Insights From Psychology, Engineering, and the Arts, This Article Provides a Comprehensive Overview of the Field of Emotion Analysis in Visual Media and Discusses the Latest Research, Systems, Challenges, Ethical Implications, and Potential Impact of Artificial Emotional Intelligence on Society.

The emergence of artificial emotional intelligence technology is revolutionizing the fields of computers and robotics, allowing for a new level of communication and understanding of human behavior that was once thought impossible. While recent advancements in deep learning have transformed the field of computer vision, automated understanding of evoked or expressed emotions in visual media remains in its infancy. This foundering stems from the absence of a universally accepted definition of "emotion," coupled with the inherently subjective nature of emotions and their intricate nuances. In this article, we provide a comprehensive, multidisciplinary overview of the field of emotion analysis in visual media, drawing on insights from psychology, engineering, and the arts. We begin by exploring the psychological foundations of emotion and the computational principles that underpin the understanding of emotions from images and videos. We then review the latest research and systems within the field, accentuating the most promising approaches. We also discuss the current technological challenges and limitations of emotion analysis, underscoring the necessity for continued investigation and innovation. We contend that this represents a "Holy Grail" research problem in computing and delineate pivotal directions for future inquiry. Finally, we examine the ethical ramifications of emotion-understanding technologies and contemplate their potential societal impacts. Overall, this article endeavors to equip readers with a deeper understanding of the domain of emotion analysis in visual media and to inspire further research and development in this captivating and rapidly evolving field.

Combined near infrared photoacoustic imaging and ultrasound detects vulnerable atherosclerotic plaque.

Atherosclerosis is an inflammatory process resulting in the deposition of cholesterol and cellular debris, narrowing of the vessel lumen and clot formation. Characterization of the morphology and vulnerability of the lesion is essential for effective clinical management. Here, near-infrared auto-photoacoustic (NIRAPA) imaging is shown to detect plaque components and, when combined with ultrasound imaging, to differentiate stable and vulnerable plaque. In an ex vivo study of photoacoustic imaging of excised plaque from 25 patients, 88.2% sensitivity and 71.4% specificity were achieved using a clinically-relevant protocol. In order to determine the origin of the NIRAPA signal, immunohistochemistry, spatial transcriptomics and spatial proteomics were co-registered with imaging and applied to adjacent plaque sections. The highest NIRAPA signal was spatially correlated with bilirubin and associated blood-based residue and with the cytoplasmic contents of inflammatory macrophages bearing CD74, HLA-DR, CD14 and CD163 markers. In summary, we establish the potential to apply the NIRAPA-ultrasound imaging combination to detect vulnerable carotid plaque and a methodology for fusing molecular imaging with spatial transcriptomic and proteomic methods.

Combined near infrared photoacoustic imaging and ultrasound detects vulnerable atherosclerotic plaque.

Atherosclerosis is an inflammatory process resulting in the deposition of cholesterol and cellular debris, narrowing of the vessel lumen and clot formation. Characterization of the morphology and vulnerability of the lesion is essential for effective clinical management. Photoacoustic imaging has sufficient penetration and sensitivity to map and characterize human atherosclerotic plaque. Here, near infrared photoacoustic imaging is shown to detect plaque components and, when combined with ultrasound imaging, to differentiate stable and vulnerable plaque. In an ex vivo study of photoacoustic imaging of excised plaque from 25 patients, 88.2% sensitivity and 71.4% specificity were achieved using a clinically-relevant protocol. In order to determine the origin of the near-infrared auto-photoacoustic (NIRAPA) signal, immunohistochemistry, spatial transcriptomics and proteomics were applied to adjacent sections of the plaque. The highest NIRAPA signal was spatially correlated with bilirubin and associated blood-based residue and inflammatory macrophages bearing CD74, HLA-DR, CD14 and CD163 markers. In summary, we establish the potential to apply the NIRAPA-ultrasound imaging combination to detect vulnerable carotid plaque.

Feasibility of [18F]fluoropivalate hybrid PET/MRI for imaging lower and higher grade glioma: a prospective first-in-patient pilot study.

PURPOSE: MRI and PET are used in neuro-oncology for the detection and characterisation of lesions for malignancy to target surgical biopsy and to plan surgical resections or stereotactic radiosurgery. The critical role of short-chain fatty acids (SCFAs) in brain tumour biology has come to the forefront. The non-metabolised SCFA radiotracer, [18F]fluoropivalate (FPIA), shows low background signal in most tissues except eliminating organs and has appropriate human dosimetry. Tumour uptake of the radiotracer is, however, unknown. We investigated the uptake characteristics of FPIA in this pilot PET/MRI study. METHODS: Ten adult glioma subjects were identified based on radiological features using standard-of-care MRI prior to any surgical intervention, with subsequent histopathological confirmation of glioma subtype and grade (lower-grade - LGG - and higher-grade - HGG - patients). FPIA was injected as an intravenous bolus injection (range 342-368 MBq), and dynamic PET and MRI data were acquired simultaneously over 66 min. RESULTS: All patients tolerated the PET/MRI protocol. Three patients were reclassified following resection and histology. Tumour maximum standardised uptake value (SUVmax,60) increased in the order LGG (WHO grade 2) < HGG (WHO grade 3) < HGG (WHO grade 4). The net irreversible solute transfer, Ki, and influx rate constant, K1, were significantly higher in HGG (p < 0.05). Of the MRI variables studied, DCE-MRI-derived extravascular-and-extracellular volume fraction (ve) was high in tumours of WHO grade 4 compared with other grades (p < 0.05). SLC25A20 protein expression was higher in HGG compared with LGG. CONCLUSION: Tumoural FPIA PET uptake is higher in HGG compared to LGG. This study supports further investigation of FPIA PET/MRI for brain tumour imaging in a larger patient population. CLINICAL TRIAL REGISTRATION: Clinicaltrials.gov, NCT04097535.

BBox-Guided Segmentor: Leveraging expert knowledge for accurate stroke lesion segmentation using weakly supervised bounding box prior.

Stroke is one of the leading causes of death and disability in the world. Despite intensive research on automatic stroke lesion segmentation from non-invasive imaging modalities including diffusion-weighted imaging (DWI), challenges remain such as a lack of sufficient labeled data for training deep learning models and failure in detecting small lesions. In this paper, we propose BBox-Guided Segmentor, a method that significantly improves the accuracy of stroke lesion segmentation by leveraging expert knowledge. Specifically, our model uses a very coarse bounding box label provided by the expert and then performs accurate segmentation automatically. The small overhead of having the expert provide a rough bounding box leads to large performance improvement in segmentation, which is paramount to accurate stroke diagnosis. To train our model, we employ a weakly-supervised approach that uses a large number of weakly-labeled images with only bounding boxes and a small number of fully labeled images. The scarce fully labeled images are used to train a generator segmentation network, while adversarial training is used to leverage the large number of weakly-labeled images to provide additional learning signals. We evaluate our method extensively using a unique clinical dataset of 99 fully labeled cases (i.e., with full segmentation map labels) and 831 weakly labeled cases (i.e., with only bounding box labels), and the results demonstrate the superior performance of our approach over state-of-the-art stroke lesion segmentation models. We also achieve competitive performance as a SOTA fully supervised method using less than one-tenth of the complete labels. Our proposed approach has the potential to improve stroke diagnosis and treatment planning, which may lead to better patient outcomes.