Unsupervised Discovery, Control, and Disentanglement of Semantic Attributes With Applications to Anomaly Detection.

Our work focuses on unsupervised and generative methods that address the following goals: (1) learning unsupervised generative representations that discover latent factors controlling image semantic attributes, (2) studying how this ability to control attributes formally relates to the issue of latent factor disentanglement, clarifying related but dissimilar concepts that had been confounded in the past, and (3) developing anomaly detection methods that leverage representations learned in the first goal. For goal 1, we propose a network architecture that exploits the combination of multiscale generative models with mutual information (MI) maximization. For goal 2, we derive an analytical result, lemma 1, that brings clarity to two related but distinct concepts: the ability of generative networks to control semantic attributes of images they generate, resulting from MI maximization, and the ability to disentangle latent space representations, obtained via total correlation minimization. More specifically, we demonstrate that maximizing semantic attribute control encourages disentanglement of latent factors. Using lemma 1 and adopting MI in our loss function, we then show empirically that for image generation tasks, the proposed approach exhibits superior performance as measured in the quality and disentanglement of the generated images when compared to other state-of-the-art methods, with quality assessed via the Fréchet inception distance (FID) and disentanglement via mutual information gap. For goal 3, we design several systems for anomaly detection exploiting representations learned in goal 1 and demonstrate their performance benefits when compared to state-of-the-art generative and discriminative algorithms. Our contributions in representation learning have potential applications in addressing other important problems in computer vision, such as bias and privacy in AI.

Deep embeddings for novelty detection in myopathy.

We address the challenge of finding anomalies in ultrasound images via deep learning, specifically applying this to screening for myopathies and finding rare presentations of myopathic disease. Among myopathic diseases, this study focuses on the use case of myositis given the spectrum of muscle involvement seen in these inflammatory muscle diseases, as well as the potential for treatment. For this study, we have developed a fully annotated dataset (called "Myositis3K") which includes 3586 images of eighty-nine individuals (35 control and 54 with myositis) acquired with informed consent. We approach this challenge as one of performing unsupervised novelty detection (ND), and use tools leveraging deep embeddings combined with several novelty scoring methods. We evaluated these various ND algorithms and compared their performance against human clinician performance, against other methods including supervised binary classification approaches, and against unsupervised novelty detection approaches using generative methods. Our best performing approach resulted in a (ROC) AUC (and 95% CI error margin) of 0.7192 (0.0164), which is a promising baseline for developing future clinical tools for unsupervised prescreening of myopathies.

Sensor-based measurement of critical care nursing workload: Unobtrusive measures of nursing activity complement traditional task and patient level indicators of workload to predict perceived exertion.

OBJECTIVE: To establish the validity of sensor-based measures of work processes for predicting perceived mental and physical exertion of critical care nurses. MATERIALS AND METHODS: Repeated measures mixed-methods study in a surgical intensive care unit. Wearable and environmental sensors captured work process data. Nurses rated their mental (ME) and physical exertion (PE) for each four-hour block, and recorded patient and staffing-level workload factors. Shift was the grouping variable in multilevel modeling where sensor-based measures were used to predict nursing perceptions of exertion. RESULTS: There were 356 work hours from 89 four-hour shift segments across 35 bedside nursing shifts. In final models, sensor-based data accounted for 73% of between-shift, and 5% of within-shift variance in ME; and 55% of between-shift, and 55% of within-shift variance in PE. Significant predictors of ME were patient room noise (ß = 0.30, p < .01), the interaction between time spent and activity levels outside main work areas (ß = 2.24, p < .01), and the interaction between the number of patients on an insulin drip and the burstiness of speaking (ß = 0.19, p < .05). Significant predictors of PE were environmental service area noise (ß = 0.18, p < .05), and interactions between: entropy and burstiness of physical transitions (ß = 0.22, p < .01), time speaking outside main work areas and time at nursing stations (ß = 0.37, p < .001), service area noise and time walking in patient rooms (ß = -0.19, p < .05), and average patient load and nursing station speaking volume (ß = 0.30, p < .05). DISCUSSION: Analysis yielded highly predictive models of critical care nursing workload that generated insights into workflow and work design. Future work should focus on tighter connections to psychometric test development methods and expansion to a broader variety of settings and professional roles. CONCLUSIONS: Sensor-based measures are predictive of perceived exertion, and are viable complements to traditional task demand measures of workload.

Distributed consensus on camera pose.

Our work addresses pose estimation in a distributed camera framework. We examine how processing cameras can best reach a consensus about the pose of an object when they are each given a model of the object, defined by a set of point coordinates in the object frame of reference. The cameras can only see a subset of the object feature points in the midst of background clutter points, not knowing which image points match with which object points, nor which points are object points or background points. The cameras individually recover a prediction of the object's pose using their knowledge of the model, and then exchange information with their neighbors, performing consensus updates locally to obtain a single estimate consistent across all cameras, without requiring a common centralized processor. Our main contributions are: 1) we present a novel algorithm performing consensus updates in 3-D world coordinates penalized by a 3-D model, and 2) we perform a thorough comparison of our method with other current consensus methods. Our method is consistently the most accurate, and we confirm that the existing consensus method based upon calculating the Karcher mean of rotations is also reliable and fast. Experiments on simulated and real imagery are reported.

Bayesian model selection for mining mass spectrometry data.

A procedure for learning a probabilistic model from mass spectrometry data that accounts for domain specific noise and mitigates the complexity of Bayesian structure learning is presented. We evaluate the algorithm by applying the learned probabilistic model to microorganism detection from mass spectrometry data.