Search for: All records

Due to their large sizes, volumetric scans and whole-slide pathology images (WSIs) are often processed by extracting embeddings from local regions and then an aggregator makes predictions from this set. However, current methods require post-hoc visualization techniques (e.g., Grad-CAM) and often fail to localize small yet clinically crucial details. To address these limitations, we introduce INSIGHT, a novel weakly-supervised aggregator that integrates heatmap generation as an inductive bias. Starting from pre-trained feature maps, INSIGHT employs a detection module with small convolutional kernels to capture fine details and a context module with a broader receptive field to suppress local false positives. The resulting internal heatmap highlights diagnostically relevant regions. On CT and WSI benchmarks, INSIGHT achieves state-of-the-art classification results and high weakly-labeled semantic segmentation performance. 

By providing a large gaseous volume for nuclear interactions while simultaneously recording the tracks of resulting reaction products, an active target serves as both a thick target and a detector. Once a reaction occurs, the emitted charged fragments strip electrons from the target gas along their path as they transverse the detector. Collection of these stripped electrons allow for detection of the product tracks. As beam intensity increases, the resulting ionization in the active target can significantly distort this collection of electrons. If left uncorrected, the resulting measurements could be wrong. In this paper, we investigate the impact of the space charge produced by heavy radioactive beams within the Active Target - Time Projection Chamber at Michigan State University. The beams are injected parallel to the electric field of the time projection chamber which is operated without a magnetic field for this experiment. We analyze the rate dependence of the space charge effects and demonstrate that they can be modeled and effectively corrected. 

Cationic water-soluble deep cavitands enable hierarchical assembly-based recognition, optical detection and extraction of perfluoroalkyl substances (PFAS) in aqueous solution. Recognition of the PFAS occurs at the lower rim crown of the cavitand, which triggers self-aggregation of a PFAS-cavitand complex, allowing extraction from water. In addition, when paired with an indicator dye that can be bound in the cavity of the host molecule, the PFAS-cavitand association causes a significant (>20-fold at micromolar [PFAS]) enhancement of dye fluorescence due to conformational rearrangement of the fluxional cavitand AMI, allowing optical sensing of PFAS. The cavitands are water-soluble, and the detection and recognition occur in purely aqueous solution. The association is most effective for long chain sulfonate PFAS, and as such, selective optical detection of perfluorooctanesulfonate is possible, with a LOD = 130 nM in buffered water, and 500 nM in real-world samples such as polluted canal water. By pairing the AMI host with multiple dyes in an array-based format, full discrimination of five other PFAS can be achieved at micromolar concentration via differential sensing. In addition, the aggregation process allows extraction of PFAS from solution, and a 99% reduction of PFOS concentration in water is possible with a single treatment of an equimolar concentration of AMI cavitand. The hierarchical nature of the cavitand recognition system allows both selective, sensitive optical detection and extraction of PFAS from water with a single scaffold. 

Large Language Models (LLMs) have achieved remarkable success in natural language tasks, yet understanding their reasoning processes re- mains a significant challenge. We address this by introducing XplainLLM, a dataset accom- panying an explanation framework designed to enhance LLM transparency and reliability. Our dataset comprises 24,204 instances where each instance interprets the LLM’s reasoning behavior using knowledge graphs (KGs) and graph attention networks (GAT), and includes explanations of LLMs such as the decoder- only Llama-3 and the encoder-only RoBERTa. XplainLLM also features a framework for gener- ating grounded explanations and the debugger- scores for multidimensional quality analysis. Our explanations include why-choose and why- not-choose components, reason-elements, and debugger-scores that collectively illuminate the LLM’s reasoning behavior. Our evaluations demonstrate XplainLLM’s potential to reduce hallucinations and improve grounded explana- tion generation in LLMs. XplainLLM is a re- source for researchers and practitioners to build trust and verify the reliability of LLM outputs. Our code and dataset are publicly available.