106261 research outputs found
Sort by
Dataset for Mastcam-Z Analog Spectral Imager (MASI): A Mastcam-Z Testbed and Field Instrument
No full textPlease cite as: Megan Barrington, Christian Tate, Alexander Hayes. (2025) Dataset for Mastcam-Z Analog Spectral Imager (MASI): A Mastcam-Z Testbed and Field Instrument. [dataset] Cornell University Library eCommons Repository. https://doi.org/10.7298/bjr2-4f92These files contain data supporting the results reported in Barrington et al. Mastcam-Z Analog Spectral Imager (MASI): A Mastcam-Z Testbed and Field Instrument. In Barrington et al., we describe the Mastcam-Z Analog Spectral Imager (MASI). Mastcam-Z is a stereoscopic, zoomable multispectral imaging system located on the Remote Sensing Mast (RSM) of the Perseverance Rover. Mastcam-Z is the first zoomable multispectral imaging system flown on a NASA spacecraft. The Mastcam-Z Analog Spectral Imager is a Mastcam-Z emulator built at Cornell University for three primary purposes: 1) to serve as a testbed for the pre-flight radiometric and geometric calibration of Mastcam-Z, 2) to characterize Mastcam-Z anomalies observed during calibration and operation, and 3) to act as a field instrument for collecting terrestrial analog multispectral data with the same resolution and spectral characteristics as Mastcam-Z images acquired on Mars. MASI is engineered using a combination of commercial off-the-shelf (COTS) components and Mastcam-Z flight spare hardware, and is calibrated using a similar set of algorithms to the flight instruments. We show that MASI produces reflectance values that are similar to Mastcam-Z laboratory and inflight values, and provide a detailed description of MASI’s hardware, software, and spectral data products.Mars 2020 Mastcam-Z Science Team Grant number: 1511125/NNN13D496
Fashion’s Data Doubles: How AI is Reshaping Modeling Work
No full textGenerative AI technologies are shifting conditions of work across different creative fields. Drawing on in-depth qualitative interviews, this research snapshot discusses how new applications of generative AI are affecting fashion models. First, we consider how AI technologies are extending the ability of fashion brands to manipulate models’ images. Next, we discuss key interview findings on the emerging impacts of AI on fashion models, including how models’ images and measurements are increasingly treated as data to extract, non-consensual alteration of models’ images, growing economic insecurity, and propagation of harmful beauty standards—issues that are felt widely but unevenly across the industry. The brief concludes by addressing efforts to strengthen AI governance in the modeling field and beyond
Corn silage: rain following dry conditions
No full textThe 2025 growing season has experienced many extremes. Recently several areas have received much needed rain and while it has not been enough to alleviate the dry conditions for most, it has been enough to “perk up” some corn fields. The following reminders and guidelines apply to corn that has a developed ear and in the maturation phase as this rain is received
DHA suppresses hormone-sensitive and castration-resistant prostate cancer growth by decreasing de novo lipogenesis.
No full textOBJECTIVE: De novo lipogenesis (DNL) is associated with prostate cancer (PCa) progression, while fatty acid synthase (FASN) overexpression is a hallmark of DNL. Palmitate, its main product, is a saturated fatty acid that supports PCa growth. Polyunsaturated fatty acids (PUFAs), which can be acquired from the microenvironment, undergo peroxidation more readily and affect membrane fluidity. Docosahexaenoic acid (DHA) is a prototype PUFA omega-3 produced inefficiently in human cells. Its levels are low in PCa cells compared to normal cells. We hypothesize that excess DHA may reprogram lipid metabolism and induce cell growth suppression. METHODS: Androgen-responsive LNCaP, castration-resistant cells C4-2 and 22Rv1, human PCa castration-resistant organoids, and prostate cancer xenografts were exposed to DHA. RESULTS: DHA accumulated into lipid droplets as triacylglycerols and cholesterol esters, led to increased phospholipid acyl chain unsaturation and altered phospholipid ratio, a known trigger of endoplasmic reticulum (ER) stress. DHA caused a decrease in sterol regulatory element-binding protein (SREBP) transcriptional program, which, in turn, led to decreased expression of FASN. The subsequent reduction in DNL caused downregulation of the androgen receptor (AR) and its splice variant AR-V7. In addition, β-oxidation was enhanced, and DHA was preferentially oxidized over palmitate. Glucose oxidation also increased in the presence of DHA. Finally, DHA led to ROS overproduction, oxidative damage, and ER stress. CONCLUSIONS: DHA reduces the growth of hormone-sensitive and castration-resistant PCa both in vitro and in vivo via deregulation of lipid metabolism.2026-05-1
Spinal CSF leaks in spontaneous intracranial hypotension: A single-institution analysis of incidence, typology and treatment outcomes.
No full textOBJECTIVE: To report incidence, typology and treatment outcomes of spinal CSF leaks in patients with spontaneous intracranial hypotension (SIH). METHODS: In this IRB approved study, consecutive SIH patients with myelogram-confirmed spinal CSF leak location, who underwent treatment between 2021 and 2023 at a single institution were retrospectively analyzed. The outcome variable was definitive treatment of SIH, defined as clinical and/or radiographic resolution of symptoms. Leak type classification was: Type 1 = ventral dural tear, Type 2 = lateral dural nerve root sleeve tear, Type 3 = CSF-venous fistula (CVF). RESULTS: 32 SIH patients (average age 48 ± 15, 28 % male, 72 % female) were analyzed. A majority of them had a Type 1 CSF leak (59 %), followed by Type 3 (31 %) and Type 2 (9 %) leaks. Thoracic spine was the predominant location of the leaks (84 %); notably all CSF-venous fistulas were located there. Following trials of conservative management, all patients underwent treatment with EBP after leak site localization. 22/32 patients (69 %) had at least some resolution of symptoms following the first EBP. For 2/32 (6 %, both Type 2 leak), one targeted EBP provided definitive treatment. 30/32 (94 %, all leak types) had persistent clinical symptoms and had additional EBP(s). The mean number of EBPs per patient was 1.4 (range = 1-3). Following treatment failure of EBP(s), 10 patients with Type 3 leaks had transvenous embolization, which resulted in definitive treatment for 9 (90 %); 16 patients (leak Type 1 = 15, Type 2 = 1) had open dural surgery, which resulted in definitive treatment for 15 (94 %, all Type 1 leaks). CONCLUSION: Overall, our analysis is consistent with recent data demonstrating that SIH incidence is higher among female patients and that CVFs are slightly more prevalent than previously reported, seen in nearly a third of our patients. Thoracic spine is the predominant location of CSF leaks; all our CVFs were located there. On treatment modalities, while EBP remains an important tool offering immediate symptom relief to SIH patients in the short term, permanent closure of the CSF leak and complete resolution of symptoms is rarely achieved with EBP. Definitive treatment is more likely with targeted endovascular and surgical modalities.2026-05-2
How Much Does Distance Matter? Sectoral Differences in City-level Patent Collaboration in China
No full textThis study investigates how the effect of geographic distance on city-level patent collaboration varies across industries in China. While prior research has established that innovation collaboration generally declines with distance, little attention has been paid to whether this spatial decay effect is consistent across sectors. Drawing on evolutionary economic theory and the concept of technological regimes, this paper argues that industries differ in how spatially constrained their innovation processes are, depending on factors such as knowledge bases and R&D intensity. Using co-invented patent data from 2020, the study applies interaction models to capture how distance effects vary by sector. The results reveal substantial heterogeneity: science-based and capital-intensive industries such as machinery and chemistry are significantly more resilient to distance, while sectors like agriculture and health are more localized. Robustness checks confirm the stability of these patterns across model specifications and collaboration intensities. These findings underscore the need for differentiated spatial innovation policies and offer new insight into how industrial characteristics shape the geography of collaborative innovation
The Slippery Slope of Supply Chains: Evaluating the Use of Sanctions in Enforcing International Law for Commercial Actors
Full text link"You Knew What You Were Getting Into": Perspective Differences in Gauging Informed Consent
No full textWe examine differences between perceived and experienced consent in organizational contexts-specifically, the aspect of consent that reflects how informed consenters feel. We theorize that people tasked with soliciting consent overestimate the extent to which consenters feel fully informed of what they are agreeing to and thus feel they have truly consented. We provide support for these predictions across six pre-registered studies (N = 2,993) and eight supplemental pre-registered studies (N = 4,406) that establish causal and mediation evidence, downstream organizational consequences, and real-world relevance. This research reveals that even when an agreement meets the legal criteria for consent, there may be misaligned perceptions of employees' feelings of consent, with consequences for employees' relationship with their organization. The current studies offer a significant step forward in understanding the markedly understudied role of consent in organizations
AI Transcription of Historical Oology Cards: A Proof of Concept Using a Multimodal Large Language Model
Full text linkIn natural history museums, handwritten specimen data cards serve as repositories of field note-derived biological information. Digitizing these records is essential to preserve their content and facilitate worldwide research access. However, this process remains a bottleneck due to the complexities of historical handwriting. I evaluate the potential of Multimodal Large Language Models (MLLMs) to automate the transcription of historical oology cards from the Robert B. Lyle Collection at the Cornell University Museum of Vertebrates. Moving beyond traditional Optical Character Recognition (OCR) pipelines, experimentation tested whether GPT-4o could transcribe and interpret card data using two prompting strategies: a literal "strict transcription" directive and a structured "thinking archivist" directive. Analysis of 262 handwritten and typed cards reveals that the structured directive significantly improved semantic accuracy and reduced median human review time, yielding error-free transcripts for 74% of the dataset. Furthermore, the development of a "Hesitation Score" derived from token-level log-probabilities identified the cards most likely to contain errors. These results demonstrate that MLLMs can accelerate digitization by acting as intelligent transcribers, enabling a scalable, triage-based workflow that optimizes human effort for large-scale historical transcription