Patients have reported sensations of seeing light flashes during radiation therapy, even with their eyes closed. These observations have been attributed to either direct excitation of retinal pigments or generation of Cherenkov light inside the eye. Both in vivo human and ex vivo animal eye imaging was used to confirm light intensity and spectra to determine its origin and overall observability. A time-gated and intensified camera was used to capture light exiting the eye of a patient undergoing stereotactic radiosurgery in real time, thereby verifying the detectability of light through the pupil. These data were compared with follow-up mechanistic imaging of ex vivo animal eyes with thin radiation beams to evaluate emission spectra and signal intensity variation with anatomic depth. Angular dependency of light emission from the eye was also measured. Patient imaging showed that light generation in the eye during radiation therapy can be captured with a signal-to-noise ratio of 68. Irradiation of ex vivo eye samples confirmed that the spectrum matched that of Cherenkov emission and that signal intensity was largely homogeneous throughout the entire eye, from the cornea to the retina, with a slight maximum near 10 mm depth. Observation of the signal external to the eye was possible through the pupil from 0° to 90°, with a detected emission near 2500 photons per millisecond (during peak emission of the ON cycle of the pulsed delivery), which is over 2 orders of magnitude higher than the visible detection threshold. By quantifying the spectra and magnitude of the signal, we now have direct experimental observations that Cherenkov light is generated in the eye during radiation therapy and can contribute to perceived light flashes. Furthermore, this technique can be used to further study and measure phosphenes in the radiation therapy clinic.

The MEDication-Indication (MEDI) knowledgebase has been utilized in research with electronic health records (EHRs) since its publication in 2013. To account for new drugs and terminology updates, we rebuilt MEDI to overhaul the knowledgebase for modern EHRs. Indications for prescribable medications were extracted using natural language processing and ontology relationships from six publicly available resources: RxNorm, Side Effect Resource 4.1, Mayo Clinic, WebMD, MedlinePlus, and Wikipedia. We compared the estimated precision and recall between the previous MEDI (MEDI-1) and the updated version (MEDI-2) with manual review. MEDI-2 contains 3031 medications and 186,064 indications. The MEDI-2 high precision subset (HPS) includes indications found within RxNorm or at least three other resources. MEDI-2 and MEDI-2 HPS contain 13% more medications and over triple the indications compared to MEDI-1 and MEDI-1 HPS, respectively. Manual review showed MEDI-2 achieves the same precision (0.60) with better recall (0.89 vs. 0.79) compared to MEDI-1. Likewise, MEDI-2 HPS had the same precision (0.92) and improved recall (0.65 vs. 0.55) than MEDI-1 HPS. The combination of MEDI-1 and MEDI-2 achieved a recall of 0.95. In updating MEDI, we present a more comprehensive medication-indication knowledgebase that can continue to facilitate applications and research with EHRs.

Pseudomonas aeruginosa is a common, opportunistic bacterial pathogen among patients with cystic fibrosis, asthma, and chronic obstructive pulmonary disease. During the course of these diseases, l-ornithine, a non-proteinogenic amino acid, becomes more abundant. P. aeruginosa is chemotactic towards other proteinogenic amino acids. Here, we evaluated the chemotaxis response of P. aeruginosa towards l-ornithine. Our results show that l-ornithine serves as a chemoattractant for several strains of P. aeruginosa, including clinical isolates, and that the chemoreceptors involved in P. aeruginosa PAO1 are PctA and PctB. It seems likely that P. aeruginosa’s chemotactic response to l-ornithine might be a common feature and thus could potentially contribute to pathogenesis processes during colonization and infection scenarios.

Objective. Analyze the performance of electrical impedance tomography (EIT) in an innovative porcine model of subclinical hemorrhage and investigate associations between EIT and hemodynamic trends. Approach . Twenty-five swine were bled at slow rates to create an extended period of subclinical hemorrhage during which the animal’s heart rate (HR) and blood pressure (BP) remained stable from before hemodynamic deterioration, where stable was defined as <15% decrease in BP and <20% increase in HR — i.e. hemorrhages were hidden from standard vital signs of HR and BP . Continuous vital signs, photo-plethysmography, and continuous non-invasive EIT data were recorded and analyzed with the objective of developing an improved means of detecting subclinical hemorrhage— ideally as early as possible . Main results . Best area-under-the-curve (AUC) values from comparing bleed to no-bleed epochs were 0.96 at a 80 ml bleed (∼15.4 min) using an EIT-data-based metric and 0.79 at a 120 ml bleed (∼23.1 min) from invasively measured BP—i.e. the EIT-data-based metric achieved higher AUCs at earlier points compared to standard clinical metrics without requiring image reconstructions . Significance. In this clinically relevant porcine model of subclinical hemorrhage, EIT appears to be superior to standard clinical metrics in early detection of hemorrhage.

Rare diseases affect over 300 million people worldwide, yet patients often endure years-long diagnostic delays that limit timely intervention and trial opportunities. Computational rare disease recognition (RDR) remains constrained by knowledge resources that are often incomplete, heterogeneous, and dependent on extensive multi-disciplinary expert curation that cannot scale. Large language models (LLMs) applied directly for end-to-end diagnosis or disease discrimination face similar knowledge bottlenecks while also raising concerns around cost, reproducibility, and data governance. Here, we introduce GEN-KnowRD, a knowledge-layer-first framework that leverages LLMs to generate schema-guided rare disease profiles, systematically assesses their quality, and constructs a computable knowledge base (PheMAP-RD) for local deployment. GEN-KnowRD integrates this knowledge into lightweight inference pipelines for both general-purpose disease screening and specialized early discrimination from longitudinal electronic health records. In tests using six public benchmarks for general-purpose screening (9,290 patients spanning 798 rare diseases), GEN-KnowRD substantially improved disease ranking versus 1) a state-of-the-art, HPO-centered diagnostic framework (up to 345.8% improvement in top-1 success), 2) advanced end-to-end LLM reasoning (up to 129.1% improvement), and 3) a variant of GEN-KnowRD instantiated with expert-curated knowledge rather than LLM-generated profiles. In two real-world cohorts for early diagnosis of idiopathic pulmonary fibrosis (511 patients) as a use case, GEN-KnowRD also demonstrated robust discrimination performance gains, supporting effective RDR during the pre-diagnostic window. These findings demonstrate that repositioning LLMs from diagnostic reasoning to the knowledge layer—decoupling knowledge construction from patient-level inference—yields stronger RDR, while providing scalable, continuously updatable, and reusable infrastructure for diagnosis, screening, and clinical research across the rare disease landscape.