Search

Deep learning models have shown promise in histopathology image analysis, but their opaque decision-making process poses challenges in high-risk medical scenarios. Here we introduce HIPPO, an explainable AI method that interrogates attention-based multiple instance learning (ABMIL) models in computational pathology by generating counterfactual examples through tissue patch modifications in whole slide images. Applying HIPPO to ABMIL models trained to detect breast cancer metastasis reveals that they may overlook small tumors and can be misled by non-tumor tissue, while attention maps$\unicode{x2014}$widely used for interpretation$\unicode{x2014}$often highlight regions that do not directly influence predictions. By interpreting ABMIL models trained on a prognostic prediction task, HIPPO identified tissue areas with stronger prognostic effects than high-attention regions, which sometimes showed counterintuitive influences on risk scores. These findings demonstrate HIPPO's capacity for comprehensive model evaluation, bias detection, and quantitative hypothesis testing. HIPPO greatly expands the capabilities of explainable AI tools to assess the trustworthy and reliable development, deployment, and regulation of weakly-supervised models in computational pathology.

Histopathology remains the gold standard for diagnosis of various cancers. Recent advances in computer vision, specifically deep learning, have facilitated the analysis of histopathology images for various tasks, including immune cell detection and microsatellite instability classification. The state-of-the-art for each task often employs base architectures that have been pretrained for image classification on ImageNet. The standard approach to develop classifiers in histopathology tends to focus narrowly on optimizing models for a single task, not considering the aspects of modeling innovations that improve generalization across tasks. Here we present ChampKit (Comprehensive Histopathology Assessment of Model Predictions toolKit): an extensible, fully reproducible benchmarking toolkit that consists of a broad collection of patch-level image classification tasks across different cancers. ChampKit enables a way to systematically document the performance impact of proposed improvements in models and methodology. ChampKit source code and data are freely accessible at https://github.com/kaczmarj/champkit ., Comment: Submitted to NeurIPS 2022 Track on Datasets and Benchmarks. Source code available at https://github.com/kaczmarj/champkit

The stochastic motions of a diffusing particle contain information concerning the particle's interactions with binding partners and with its local environment. However, accurate determination of the underlying diffusive properties, beyond normal diffusion, has remained challenging when analyzing particle trajectories on an individual basis. Here, we introduce the maximum likelihood estimator (MLE) for confined diffusion and fractional Brownian motion. We demonstrate that this MLE yields improved estimation over traditional mean square displacement analyses. We also introduce a model selection scheme (that we call mleBIC) that classifies individual trajectories to a given diffusion mode. We demonstrate the statistical limitations of classification via mleBIC using simulated data. To overcome these limitations, we introduce a new version of perturbation expectation-maximization (pEMv2), which simultaneously analyzes a collection of particle trajectories to uncover the system of interactions which give rise to unique normal and/or non-normal diffusive states within the population. We test and evaluate the performance of pEMv2 on various sets of simulated particle trajectories, which transition among several modes of normal and non-normal diffusion, highlighting the key considerations for employing this analysis methodology.

Experiments Along the Coexistence near Tricriticality (EXACT) will test exact predictions made by Renormalization Group theory by mapping the phase diagram of liquid He-3 and He-4 mixtures at the tricritical point, Tcp=0.867 K. A PdMn based High Resolution Thermometer (HRT) will be utilized by EXACT to make accurate measurements with a resolution that has never been attained for the temperature range 0.7-1 K. The basic design of this mini high resolution thermometer comprises a sensing element whose magnetic susceptibility changes with temperature, a thermal connector, magnetic shielding, and some permanent magnets to apply a constant magnetic field. In this study, we will quantitatively determine the resolution of possible sensing element candidates of 0.15%, 0.20%, and 0.25% ppm Mn and compare them with an annealed group of PdMn with the corresponding concentrations to see how this processing technique affects sensitivity.