The objective assessment of image quality (IQ) has been advocated for the analysis and optimization of medical imaging systems. One method of obtaining such IQ metrics is through a mathematical observer. The Bayesian ideal observer is optimal by definition for signal detection tasks, but is frequently both intractable and non-linear. As an alternative, linear observers are sometimes used for task-based image quality assessment. The optimal linear observer is the Hotelling observer (HO). The computational cost of calculating the HO increases with image size, making a reduction in the dimensionality of the data desirable. Channelized methods have become popular for this purpose, and many competing methods are available for computing efficient channels. In this work, a novel method for learning channels using an autoencoder (AE) is presented. AEs are a type of artificial neural network (ANN) that are frequently employed to learn concise representations of data to reduce dimensionality. Modifying the traditional AE loss function to focus on task-relevant information permits the development of efficient AE-channels. These AE-channels were trained and tested on a variety of signal shapes and backgrounds to evaluate their performance. In the experiments, the AE-learned channels were competitive with and frequently outperformed other state-of-the-art methods for approximating the HO. The performance gains were greatest for the datasets with a small number of training images and noisy estimates of the signal image. Overall, AEs are demonstrated to be competitive with state-of-the-art methods for generating efficient channels for the HO and can have superior performance on small datasets.

Medical imaging systems are often evaluated and optimized via objective, or task-specific, measures of image quality (IQ) that quantify the performance of an observer on a specific clinically-relevant task. The performance of the Bayesian Ideal Observer (IO) sets an upper limit among all observers, numerical or human, and has been advocated for use as a figure-of-merit (FOM) for evaluating and optimizing medical imaging systems. However, the IO test statistic corresponds to the likelihood ratio that is intractable to compute in the majority of cases. A sampling-based method that employs Markov-Chain Monte Carlo (MCMC) techniques was previously proposed to estimate the IO performance. However, current applications of MCMC methods for IO approximation have been limited to a small number of situations where the considered distribution of to-be-imaged objects can be described by a relatively simple stochastic object model (SOM). As such, there remains an important need to extend the domain of applicability of MCMC methods to address a large variety of scenarios where IO-based assessments are needed but the associated SOMs have not been available. In this study, a novel MCMC method that employs a generative adversarial network (GAN)-based SOM, referred to as MCMC-GAN, is described and evaluated. The MCMC-GAN method was quantitatively validated by use of test-cases for which reference solutions were available. The results demonstrate that the MCMC-GAN method can extend the domain of applicability of MCMC methods for conducting IO analyses of medical imaging systems.

Last year, when the Food and Drug Administration approved an Apple Watch feature that notified users if they had an irregular heart rhythm, the information tech industry hailed it as a watershed moment in consumer-focused health care. Cardiologists, on the other hand, warned that the app could lead to privacy violations, unwarranted worrying and wasteful or even dangerous medical care. It might have been good to have an authoritative assessment of the new technology's pros and cons. But in the United States, at least, that no longer happens. In Britain, France and the European Union, government agencies examine the ethical, social and economic impact of artificial intelligence and other big new technologies used in health care and elsewhere.

Decoding of the human genome still poses puzzles that might be solved with the help of artificial intelligence.

IMAGE: Treating diseases and "improving " genetic material: KIT researchers investigate potential contributions of AI and the resulting ethical problems. Decoding of the human genome still poses puzzles that might be solved with the help of artificial intelligence. New therapeutic approaches to treating severe diseases appear possible as do non-medical "improvements" of the genetic material. With funds of the Federal Ministry of Education and Research (BMBF), technology assessment experts of Karlsruhe Institute of Technology (KIT) study which applications are realistic and which ethical issues they may entail. "Modern genome research works on understanding and predicting how genetic differences between human beings determine complex features, such as predispositions to frequent diseases," says Harald König of KIT's Institute for Technology Assessment and Systems Analysis (ITAS).

To gain a better understanding of the emerging opportunities, challenges, and implications resulting from developments in artificial intelligence (AI), the Comptroller General of the United States convened the Forum on Artificial Intelligence, which was held on July 6 and 7, 2017, in Washington, D.C. GAO issued a technology assessment in March 2018 summarizing the results of this forum. Forum participants noted a range of opportunities and challenges related to AI, as well as areas needed for future research and for consideration by policymakers. Regarding opportunities, investment in automation through AI technologies could lead to improvements in productivity and economic outcomes, according to a forum participant. AI can also be used to gather an enormous amount of data from multiple sources and detect abnormalities faster than humans can, and it could be used to help solve some of the world's most complex and pressing problems. The participants also highlighted a number of challenges related to AI.

A quartet of tech experts arrived at a little-noticed hearing at the U.S. Capitol in May with a message: Quantum computing is a bleeding-edge technology with the potential to speed up drug research, financial transactions and more. To Rep. Adam Kinzinger, though, their highly technical testimony might as well have been delivered in a foreign language. "I can understand about 50 percent of the things you say," the Illinois Republican confessed. Kinzinger's quip drew chuckles from his peers on the House Energy and Commerce Committee, but it also illustrated an unavoidable challenge on Capitol Hill. Increasingly, members of Congress are confronting a wide array of complex policy debates posed by inventions like artificial intelligence and problems like the rise of Russian propaganda online.

When the draft version of a federal encryption bill got leaked this month, the verdict in the tech community was unanimous. Critics called it ludicrous and technically illiterate--and these were the kinder assessments of the "Compliance with Court Orders Act of 2016," proposed legislation authored by the offices of Senators Diane Feinstein and Richard Burr. The encryption issue is complex and the stakes are high, as evidenced by the recent battle between Apple and the FBI. Many other technology issues that the country is grappling with these days are just as complex, controversial, and critical--witness the debates over law enforcement's use of stingrays to track mobile phones or the growing concerns around drones, self-driving cars, and 3-D printing. Yet decisions about these technical issues are being handled by luddite lawmakers who sometimes boast about not owning a cell phone or never having sent an email.

This paper describes a heuristic Bayesian method for computing probability distributions from experimental data, based upon the normal distribution form of the influence diagram. An example illustrates its use in medical technology assessment. This approach facilitates the integration of results from different studies, and permits a medical expert to make proper assessments without considerable statistical training. There has been extensive research on the construction and manipulation of expert systems using probabilities as a measure for uncertainty. These systems are capable of recognizing considerable dependence and of learning from unreliable observations.