Can AI Transform Patient Care from Reactive Craft to Strategic Art? -


Personalized Analytics is becoming essential in healthcare, stemming from the movement from fee-for-service to a value-based market. The need to preempt and prevent disease on a more personal level, rather than merely reacting to symptoms, has created a significant opportunity for machine learning-based applications. This "analytics of one" approach (using advanced mathematical models and artificial intelligence techniques) is already impacting several key areas: Prime examples include cardiac imaging analysis that aides physicians in assessing conditions, including heart attacks and coronary artery disease, and retinal image analysis to detect diabetic retinopathy. The anticipated goal for AI in healthcare is to enhance and expand the "four Ps" of care delivery – predictive, preventative, personalized and participatory. Predictive: Predictions have existed in healthcare for some decades now, as statistical models based on structured data sources.

Patients Report Mixed Views on Health-Tech and AI


Nearly 50% of patients consider biometric monitoring devices and artificial intelligence (AI) a great opportunity, while 11% view the technologies as a great danger, according to the findings of a study published in npj Digital Medicine. Through open-ended questions, 47% of patients said they believe the technologies offer great opportunities and identified 47 potential benefits. The respondents said that health-tech could improve their follow-up and the reactivity of care (55%), reduce their burden of treatment (23%) and facilitate physicians' work (21%). "Coupled with the progress of AI, the thousands of data points collected from (biometric monitoring devices) may help in informing diagnosis, predicting patient outcomes and helping care professionals select the best treatment for their patients," the study authors wrote. One participant, a 35-year-old man with diabetes and Hashimoto's thyroiditis, said that new technologies are, "the only way for a physician to simultaneously take into account all multiple parameters necessary to adjust diabetes treatment: insulin sensitivity, duration of action, blood sugar levels, physical activity, continuous measurement…" But others viewed the technologies as a great danger.

An Introduction to AI and Machine Learning


Get a free IBM Cloud account. Machine learning is branching out across numerous fields, one of the most interesting fields is health care. In this tech talk, we will go through an overview of what Machine Learning and Artificial Intelligence are, explaining at a high level key concepts such as models and classifiers. After, we will go through an example of how to train a machine learning model to predict type 2 diabetes using synthesized patient health records. Anwesha will demo preparing data using Apache Spark, visualizing data relationships using PixieDust, training a model, and deploying it to receive predictions.

Correlating Twitter Language with Community-Level Health Outcomes

arXiv.org Machine Learning

We study how language on social media is linked to diseases such as atherosclerotic heart disease (AHD), diabetes and various types of cancer. Our proposed model leverages state-of-the-art sentence embeddings, followed by a regression model and clustering, without the need of additional labelled data. It allows to predict community-level medical outcomes from language, and thereby potentially translate these to the individual level. The method is applicable to a wide range of target variables and allows us to discover known and potentially novel correlations of medical outcomes with life-style aspects and other socioeconomic risk factors.

Errors-in-variables Modeling of Personalized Treatment-Response Trajectories

arXiv.org Machine Learning

Estimating the effect of a treatment on a given outcome, conditioned on a vector of covariates, is central in many applications. However, learning the impact of a treatment on a continuous temporal response, when the covariates suffer extensively from measurement error and even the timing of the treatments is uncertain, has not been addressed. We introduce a novel data-driven method that can estimate treatment-response trajectories in this challenging scenario. We model personalized treatment-response curves as a combination of parametric response functions, hierarchically sharing information across individuals, and a sparse Gaussian process for the baseline trend. Importantly, our model considers measurement error not only in treatment covariates, but also in treatment times, a problem which arises in practice for example when treatment information is based on self-reporting. In a challenging and timely problem of estimating the impact of diet on continuous blood glucose measurements, our model leads to significant improvements in estimation accuracy and prediction.

Four Steps To Implementing Artificial Intelligence In Clinical Settings – Flarrio


The clinical implementation of Artificial Intelligence (AI) is the most viable means of uniting the interests of the healthcare industry's capital constituents: the patient, the payer, and the provider. AI can improve healthcare outcomes while reducing costs when used to address patient compliance, chronic care management, genome sequencing, and physician diagnostics by classifying treatment options. Its widespread clinical deployment is poised to transform the healthcare industry into one that maintains wellness instead of merely combating illness. Maximizing AI's clinical value depends on the proper execution of four interrelated steps, each of which represents emerging developments within the industry: The proper implementation of each of these steps will ensure a future in which AI substantially contributes to decreased costs of chronic care and patient non-adherence, while achieving patient objectives in accordance with contemporary physician economics. Their implementation will also provide physicians with a vital support tool for conducting remote diagnostics, treatment classifications and accelerated care management.

CERTIFAI: Counterfactual Explanations for Robustness, Transparency, Interpretability, and Fairness of Artificial Intelligence models

arXiv.org Machine Learning

As artificial intelligence plays an increasingly important role in our society, there are ethical and moral obligations for both businesses and researchers to ensure that their machine learning models are designed, deployed, and maintained responsibly. These models need to be rigorously audited for fairness, robustness, transparency, and interpretability. A variety of methods have been developed that focus on these issues in isolation, however, managing these methods in conjunction with model development can be cumbersome and timeconsuming. In this paper, we introduce a unified and model-agnostic approach to address these issues: Counterfactual Explanations for Robustness, Transparency, Interpretability, and Fairness of Artificial Intelligence models (CERTIFAI). Unlike previous methods in this domain, CERTIFAI is a general tool that can be applied to any black-box model and any type of input data. Given a model and an input instance, CERTIFAI uses a custom genetic algorithm to generate counterfactuals: instances close to the input that change the prediction of the model. We demonstrate how these counterfactuals can be used to examine issues of robustness, interpretability, transparency, and fairness. Additionally, we introduce CERScore, the first black-box model robustness score that performs comparably to methods that have access to model internals.

Merging versus Ensembling in Multi-Study Machine Learning: Theoretical Insight from Random Effects

arXiv.org Machine Learning

A critical decision point when training predictors using multiple studies is whether these studies should be combined or treated separately. We compare two multi-study learning approaches in the presence of potential heterogeneity in predictor-outcome relationships across datasets. We consider 1) merging all of the datasets and training a single learner, and 2) cross-study learning, which involves training a separate learner on each dataset and combining the resulting predictions. In a linear regression setting, we show analytically and confirm via simulation that merging yields lower prediction error than cross-study learning when the predictor-outcome relationships are relatively homogeneous across studies. However, as heterogeneity increases, there exists a transition point beyond which cross-study learning outperforms merging. We provide analytic expressions for the transition point in various scenarios and study asymptotic properties.

Jennifer Lim, MD: The Impact of Artificial Intelligence in DR Screening


EyeArt is an artificial intelligence system that has been developed and tested to help bridge the gap in screening for diabetic retinopathy. As the numbers of people with diabetes continue to grow, there aren't enough ophthalmologists and other health care providers to adequately screen for diabetic eye disease. Chair, Professor of Ophthalmology, and Director of Retina Service at the University of Illinois at Chicago presented the results of a clinical trial of the EyeArt system at the 2019 Association for Research in Vision and Ophthalmology (ARVO) Imaging in the Eye Conference in Vancouver, BC. Lim told MD Magazine that the EyeArt system has a "great ability to be useful to detect referable DR [diabetic retinopathy] from non-referable DR and what that really means in the practical sense is that the EyeArt system and this artificial intelligence system is useful in order to help triage patients and screen for diabetic retinopathy." She added that her hope that EyeArt will contribute to reducing blindness in this population in the future.

As Artificial Intelligence Moves Into Medicine, The Human Touch Could Be A Casualty

NPR Technology

When Kim Hilliard shows up at the clinic at the New Orleans University Medical Center, she's not there simply for an eye exam. The human touches she gets along the way help her navigate her complicated medical conditions. In addition to diabetes, the 56-year-old has high blood pressure. She has also had back surgery and has undergone bariatric surgery to help her control her weight. Hilliard is also at risk of blindness, which can result from a condition called diabetic retinopathy.