University of Illinois System

TROPICANA (TROPIcal Cyclones in ANthropocene: physics, simulations & Attribution) was a one-month programme gathering 60 scientists at the Institut Pascal (University Paris-Saclay), where we reflected on how to advance knowledge on the impact of Climate Change on Tropical Cyclones, and how to produce information relevant for mitigation, future risk assessment and adaptation. I will present several collective outcomes from the programme, including two white papers on defining Tropical Cyclones seeds and defining CYCLOPS (surface flux-driven cyclones outside the tropics), two reviews on African Easterly Waves and Tropical Cyclones features driving impacts, as well as a new Tropical Cyclones attribution methodology.

The aim of this work is to fabricate high-power, single mode vertical cavity surface emitting lasers (VCSELs). Multiple junction VCSELs provide a pathway towards scaling optical output power. However, especially in the case of oxide confinement, these lasers have a greater tendency to exhibit multimode behavior when compared to their single junction counterparts. To achieve high-power, single mode multi-junction VCSELs, more advanced single mode strategies must be employed. A two-dimensional complex refractive index model was applied to model the transverse modes of photonic crystal VCSELs in order to determine the most suitable designs. Oxide confined VCSELs were fabricated using 850 nm single and double junction wafers, as well as 940 nm double and triple junction wafers. We report a maximum single mode power of 8.0 mW from the 850 nm double junction VCSELs and 10.7 mW from the 940 nm triple junction VCSELs. Double junction implant/oxide VCSELs were also fabricated and produced a peak single mode power of 8.1 mW.

The resilience of quantum two-photon interference against optical loss and background sets it apart from classical single-photon interference for the purpose of sensing. However, achieving fast measurements at the nanometer scale with two-photon interference is challenging, as attaining high resolution typically requires either ultrabroadband photons or many probe photons. To realize nanometer-scale measurements in a matter of seconds or less, we greatly increase the per-photon information content via highly non-degenerate energy entanglement. Our fast measurements thereby enable us to probe time-varying signals at the nanometer scale. To further develop our capabilities, we have drawn on single-photon imaging techniques to implement a robust protocol for reconstructing periodic signals from our relatively sparse detection data, complete with amplitude and frequency information. With this analysis regimen, we have successfully extracted complex multi-frequency signals as well as kHz-range acoustic vibrations. We review our method and present the results of an example measurement to illustrate our capabilities.

Single-Molecule Localization Microscopy (SMLM) holds great promise for clinical pathology by enabling nanoscale visualization of disease biomarkers. However, its clinical translation is hindered by low throughput, high cost, and manual workflows. We present PathSTORM II, an automated, high-throughput nanoscopy platform designed to overcome these limitations. PathSTORM II features a large 300 × 300 µm² field of view, fully automated staining and imaging, and integrated label-free quantitative phase imaging (QPI). Its modular, cost-effective design leverages components adapted from astronomy to reduce complexity and expense. PathSTORM II brings super-resolution imaging closer to routine clinical use, enabling scalable and accessible diagnostics.

Photoacoustic computed tomography (PACT) combines the advantages of optical and ultrasound imaging, offering significant potential for both preclinical and clinical applications. The development and refinement of PACT imaging systems and reconstruction methods should ideally be guided by performance measures that reflect the clinical utility of the images produced. However, widely employed conventional image quality (IQ) measures, including mean squared error (MSE) and structural similarity index (SSIM), do not always correlate with the utility of images with consideration of a specified diagnostic task, such as lesion detection. To address this, task-based IQ measures can be employed, which directly quantify the utility of an image for a specified clinical purpose. However, currently, such task-based IQ measures are not widely utilized in the photoacoustics research community. In this study, a task-based IQ measure is introduced and employed to quantify the performance of PACT reconstruction methods with respect to a clinically significant task, namely binary signal detection. Through systematic virtual imaging studies that involve realistic numerical breast phantoms and both physicsand learning-based image reconstruction methods, the advantages of task-based over conventional IQ measures are demonstrated. Examples are provided in which conventional IQ measures suggest comparable performance across reconstruction methods, while the task-based IQ measure demonstrates substantial differences, particularly when clinically significant lesions are missed. These results highlight the limitations of conventional IQ metrics in characterizing the diagnostic performance of PACT methods and underscore the necessity of integrating task-based IQ assessment into the development and evaluation of PACT reconstruction methods for clinical translation.

Atopic dermatitis is a common disease that afflicts millions but is still diagnosed and treated in many cases by scoring and treatment by trial-and-error. We carried out a clinical study employing non-invasive multiphoton microscopy alongside spatial transcriptomic analysis with the goal of correlating in vivo imaging spatial parameters with gene expression analysis. The study found numerous strong correlations and tantalizingly demonstrated non-invasive multiphoton imaging capable of predicting spatial transcriptomic alterations.

Here, we examined the consequences of biologically relevant vitamin D deficiency, a known risk factor for aggressive prostate cancer, using ex vivo and in vivo models. Phenotypic and single-cell RNA sequencing of mouse prostate organoids showed that vitamin D deficiency stunted luminal cell differentiation more than androgen deficiency, which is a known driver of prostate development. Mice fed a vitamin D-deficient diet showed significantly altered expression of androgen-responsive genes in their prostate luminal cells, as determined by single-cell RNA sequencing. MDA-PCa-2b human prostate cancer cells, when maintained for 6 months in 1,25-dihydroxyvitamin D, lost the ability to form xenografts, despite normal proliferation in vitro. RNA sequencing showed that these cells also had disruptions in androgen signaling and multiple cancer-related pathways. This study offers new insights and validation of vitamin D's role in both benign and malignant prostate biology, underscoring its essential hormonal functions and supporting strategies for vitamin D supplementation to reduce prostate cancer risk in vulnerable populations.

The integration of foundation models has revolutionized robotics, endowing agents with unprecedented perception, planning and control capabilities. However, most current robotic models remain static: they are trained offline on fixed datasets and deployed without persistent updates, despite nonstationary environments, shifting task demands, and evolving user needs. Continual learning (CL) offers a principled solution by enabling robots to update from streaming experience while retaining previously acquired capabilities and avoiding catastrophic forgetting. Despite the accelerating progress of CL for robotics, an up-to-date review dedicated to this significant topic remains absent. To bridge this gap, we present the first comprehensive survey of Continual Learning for Robotics in the foundation model era, providing clear task formulations, a taxonomy of techniques, a review of benchmarks and environments, and a discussion of key challenges and future directions.