Artificial Intelligence Using Open Source BI-RADS Data Exemplifying Potential Future Use

OBJECTIVES

With much hype about artificial intelligence (AI) rendering radiologists redundant, a simple radiologist-augmented AI workflow is evaluated; the premise is that inclusion of a radiologist's opinion into an AI algorithm would make the algorithm achieve better accuracy than an algorithm trained on imaging parameters alone. Open-source BI-RADS data sets were evaluated to see whether inclusion of a radiologist's opinion (in the form of BI-RADS classification) in addition to image parameters improved the accuracy of prediction of histology using three machine learning algorithms vis-à-vis algorithms using image parameters alone.

MATERIALS AND METHODS

BI-RADS data sets were obtained from the University of California, Irvine Machine Learning Repository (data set 1) and the Digital Database for Screening Mammography repository (data set 2); three machine learning algorithms were trained using 10-fold cross-validation. Two sets of models were trained: M1, using lesion shape, margin, density, and patient age for data set 1 and image texture parameters for data set 2, and M2, using the previous image parameters and the BI-RADS classification provided by radiologists. The area under the curve and the Gini coefficient for M1 and M2 were compared for the validation data set.

RESULTS

The models using the radiologist-provided BI-RADS classification performed significantly better than the models not using them (P < .0001).

CONCLUSION

AI and radiologist working together can achieve better results, helping in case-based decision making. Further evaluation of the metrics involved in predictor handling by AI algorithms will provide newer insights into imaging.

Rigorous prognostication and modeling of gas adsorption on activated carbon and Zeolite-5A

Gas adsorption on various adsorbents is of highly important issue for the separation of gas mixtures in many industrial processes. In this work, estimation of pure gases (CH₄, N₂, CO₂, H₂, C₂H₄) adsorption on activated carbon (AC) and CO₂, CH₄, N₂ on Zeolite-5A adsorbent were studied by developing four different computing techniques, namely MLP-ANN, ANFIS, LSSVM, and PSO-ANFIS for a broad range of experimental data found in the literature. Temperature, pressure, pore size (only for AC) and kinetic diameter of adsorbed gases are considered as the inputs and the gas adsorption as the output parameters of the developed models. We also used several statistical and graphical tools to assess the accuracy and applicability of the proposed models. The results of the study suggest the reliability and validity of all the models developed for estimating the equilibrium adsorption of gases on the adsorbents. Also, it is found that of all the models developed, the ANN model estimates experimental data of the gas adsorption on AC more accurately due to its values of R² and AARD%, 0.9865 and 0.8948, respectively. Besides, PSO-ANFIS is the best model to prognosticate gas adsorption on zeolite 5A with R² and AARD%, 0.9897 and 0.9551, respectively.

Deep Reinforcement Learning in Medicine

Reinforcement learning has achieved tremendous success in recent years, notably in complex games such as Atari, Go, and chess. In large part, this success has been made possible by powerful function approximation methods in the form of deep neural networks. The objective of this paper is to introduce the basic concepts of reinforcement learning, explain how reinforcement learning can be effectively combined with deep learning, and explore how deep reinforcement learning could be useful in a medical context.

Progress in the Development and Challenges for the Use of Artificial Kidneys and Wearable Dialysis Devices

Background

Renal transplantation is the treatment of choice for chronic kidney disease (CKD) patients, but the shortage of kidneys and the disabling medical conditions these patients suffer from make dialysis essential for most of them. Since dialysis drastically affects the patients' lifestyle, there are great expectations for the development of wearable artificial kidneys, although their use is currently impeded by major concerns about safety. On the other hand, dialysis patients with hemodynamic instability do not usually tolerate intermittent dialysis therapy because of their inability to adapt to a changing scenario of unforeseen events. Thus, the development of novel wearable dialysis devices and the improvement of clinical tolerance will need contributions from new branches of engineering such as artificial intelligence (AI) and machine learning (ML) for the real-time analysis of equipment alarms, dialysis parameters, and patient-related data with a real-time feedback response. These technologies are endowed with abilities normally associated with human intelligence such as learning, problem solving, human speech understanding, or planning and decision-making. Examples of common applications of AI are visual perception (computer vision), speech recognition, and language translation. In this review, we discuss recent progresses in the area of dialysis and challenges for the use of AI in the development of artificial kidneys.

Summary and Key Messages

Emerging technologies derived from AI, ML, electronics, and robotics will offer great opportunities for dialysis therapy, but much innovation is needed before we achieve a smart dialysis machine able to analyze and understand changes in patient homeostasis and to respond appropriately in real time. Great efforts are being made in the fields of tissue engineering and regenerative medicine to provide alternative cell-based approaches for the treatment of renal failure, including bioartificial renal systems and the implantation of bioengineered kidney constructs.

Can nurses remain relevant in a technologically advanced future?

Technological breakthroughs occur at an ever-increasing rate thereby revolutionizing human health and wellness care. Technological advancements have drastically changed the structure and organization of the healthcare industry. McKinsey Global Institute estimates that 800 million workers worldwide could be replaced by robots by the year 2030. There is already a robotic revolution happening in healthcare wherein robots have made tasks and procedures more efficient and safer. Locsin and Ito has addressed the threat to nursing practice with human nurses being replaced by humanoid robots. Routine nursing care dictated solely by prescribed procedures and accomplishment of nursing tasks would be best performed by machines. With the future practice of nursing in a technologically advanced future transcending the implementation of nursing actions to achieve predictable outcomes, how can human nurses remain relevant as practitioners of nursing? Nurses should be involved in deciding which aspects of their practice can be delegated to technology. Nurses should oversee the introduction of automated technology and artificial intelligence ensuring their practice to be more about the universal aspects of human care continuing under a novel system. Nursing education and nursing research will change to encompass a differentiated demand for professional nursing practice with, and not for, robots in healthcare.

A facial expression controlled wheelchair for people with disabilities

BACKGROUND AND OBJECTIVES

In order to improve assistive technologies for people with reduced mobility, this paper develops a new intelligent real-time emotion detection system to control equipment, such as electric wheelchairs (EWC) or robotic assistance vehicles. Every year, degenerative diseases and traumas prohibit thousands of people to easily control the joystick of their wheelchairs with their hands. Most current technologies are considered invasive and uncomfortable such as those requiring the user to wear some body sensor to control the wheelchair.

METHODS

In this work, the proposed Human Machine Interface (HMI) provides an efficient hands-free option that does not require sensors or objects attached to the user's body. It allows the user to drive the wheelchair using its facial expressions which can be flexibly updated. This intelligent solution is based on a combination of neural networks (NN) and specific image preprocessing steps. First, the Viola-Jones combination is used to detect the face of the disability from a video. Subsequently, a neural network is used to classify the emotions displayed on the face. This solution called "The Mathematics Behind Emotion" is capable of classifying many facial expressions in real time, such as smiles and raised eyebrows, which are translated into signals for wheelchair control. On the hardware side, this solution only requires a smartphone and a Raspberry Pi card that can be easily mounted on the wheelchair.

RESULTS

Many experiments have been conducted to evaluate the efficiency of the control acquisition process and the user experience in driving a wheelchair through facial expressions. The classification accuracy can expect 98.6% and it can offer an average recall rate of 97.1%. Thus, all these experiments have proven that the proposed system is able of accurately recognizing user commands in real time. Indeed, the obtained results indicate that the suggested system is more comfortable and better adapted to severely disabled people in their daily lives, than conventional methods. Among the advantages of this system, we cite its real time ability to identify facial emotions from different angles.

CONCLUSIONS

The proposed system takes into account the patient's pathology. It is intuitive, modern, doesn't require physical effort and can be integrated into a smartphone or tablet. The results obtained highlight the efficiency and reliability of this system, which ensures safe navigation for the disabled patient.

Artificial intelligence outperforms experienced nephrologists to assess dry weight in pediatric patients on chronic hemodialysis

BACKGROUND

Dry weight is the lowest weight patients on hemodialysis can tolerate; correct dry weight estimation is necessary to minimize morbi-mortality, but is difficult to achieve. Here, we used artificial intelligence to improve the accuracy of dry weight assessment in hemodialysis patients.

METHODS/RESULTS

We designed a neural network which used bio-impedancemetry, blood volume monitoring, and blood pressure values as inputs; output was artificial intelligence dry weight. Fourteen pediatric patients were switched from nephrologist to artificial intelligence dry weight. Artificial intelligence dry weight was higher (28.6%), lower (50%), or identical to nephrologist dry weight. Mean difference between artificial intelligence and nephrologist dry weights was 0.497 kg (- 1.33 to + 1.29 kg). In patients for whom artificial intelligence dry weight was lower than nephrologist dry weight, systolic blood pressure significantly decreased after dry weight decrease to artificial intelligence dry weight (77th to 60th percentile, p = 0.022); anti-hypertensive treatments were successfully decreased or discontinued in 28.7% of cases. In patients for whom artificial intelligence dry weight was higher than nephrologist dry weight, no hypertension was observed after dry weight increase to artificial intelligence dry weight; when present, symptoms of dry weight underestimation receded.

CONCLUSIONS

Neural network predictions outperformed those of experienced nephrologists in most cases, proving artificial intelligence is a powerful tool for predicting dry weight in hemodialysis patients.

The Ugly Truth About Ourselves and Our Robot Creations: The Problem of Bias and Social Inequity

Recently, there has been an upsurge of attention focused on bias and its impact on specialized artificial intelligence (AI) applications. Allegations of racism and sexism have permeated the conversation as stories surface about search engines delivering job postings for well-paying technical jobs to men and not women, or providing arrest mugshots when keywords such as "black teenagers" are entered. Learning algorithms are evolving; they are often created from parsing through large datasets of online information while having truth labels bestowed on them by crowd-sourced masses. These specialized AI algorithms have been liberated from the minds of researchers and startups, and released onto the public. Yet intelligent though they may be, these algorithms maintain some of the same biases that permeate society. They find patterns within datasets that reflect implicit biases and, in so doing, emphasize and reinforce these biases as global truth. This paper describes specific examples of how bias has infused itself into current AI and robotic systems, and how it may affect the future design of such systems. More specifically, we draw attention to how bias may affect the functioning of (1) a robot peacekeeper, (2) a self-driving car, and (3) a medical robot. We conclude with an overview of measures that could be taken to mitigate or halt bias from permeating robotic technology.

Diagnostic Performance of Deep Learning Algorithms Applied to Three Common Diagnoses in Dermatopathology

Background

Artificial intelligence is advancing at an accelerated pace into clinical applications, providing opportunities for increased efficiency, improved accuracy, and cost savings through computer-aided diagnostics. Dermatopathology, with emphasis on pattern recognition, offers a unique opportunity for testing deep learning algorithms.

Aims

This study aims to determine the accuracy of deep learning algorithms to diagnose three common dermatopathology diagnoses.

Methods

Whole slide images (WSI) of previously diagnosed nodular basal cell carcinomas (BCCs), dermal nevi, and seborrheic keratoses were annotated for areas of distinct morphology. Unannotated WSIs, consisting of five distractor diagnoses of common neoplastic and inflammatory diagnoses, were included in each training set. A proprietary fully convolutional neural network was developed to train algorithms to classify test images as positive or negative relative to ground truth diagnosis.

Results

Artificial intelligence system accurately classified 123/124 (99.45%) BCCs (nodular), 113/114 (99.4%) dermal nevi, and 123/123 (100%) seborrheic keratoses.

Conclusions

Artificial intelligence using deep learning algorithms is a potential adjunct to diagnosis and may result in improved workflow efficiencies for dermatopathologists and laboratories.

Automated microscopy for routine malaria diagnosis: a field comparison on Giemsa-stained blood films in Peru

Background

Microscopic examination of Giemsa-stained blood films remains a major form of diagnosis in malaria case management, and is a reference standard for research. However, as with other visualization-based diagnoses, accuracy depends on individual technician performance, making standardization difficult and reliability poor. Automated image recognition based on machine-learning, utilizing convolutional neural networks, offers potential to overcome these drawbacks. A prototype digital microscope device employing an algorithm based on machine-learning, the Autoscope, was assessed for its potential in malaria microscopy. Autoscope was tested in the Iquitos region of Peru in 2016 at two peripheral health facilities, with routine microscopy and PCR as reference standards. The main outcome measures include sensitivity and specificity of diagnosis of malaria from Giemsa-stained blood films, using PCR as reference.

Methods

A cross-sectional, observational trial was conducted at two peripheral primary health facilities in Peru. 700 participants were enrolled with the criteria: (1) age between 5 and 75 years, (2) history of fever in the last 3 days or elevated temperature on admission, (3) informed consent. The main outcome measures included sensitivity and specificity of diagnosis of malaria from Giemsa-stained blood films, using PCR as reference.

Results

At the San Juan clinic, sensitivity of Autoscope for diagnosing malaria was 72% (95% CI 64–80%), and specificity was 85% (95% CI 79–90%). Microscopy performance was similar to Autoscope, with sensitivity 68% (95% CI 59–76%) and specificity 100% (95% CI 98–100%). At San Juan, 85% of prepared slides had a minimum of 600 WBCs imaged, thus meeting Autoscope’s design assumptions. At the second clinic, Santa Clara, the sensitivity of Autoscope was 52% (95% CI 44–60%) and specificity was 70% (95% CI 64–76%). Microscopy performance at Santa Clara was 42% (95% CI 34–51) and specificity was 97% (95% CI 94–99). Only 39% of slides from Santa Clara met Autoscope’s design assumptions regarding WBCs imaged.

Conclusions

Autoscope’s diagnostic performance was on par with routine microscopy when slides had adequate blood volume to meet its design assumptions, as represented by results from the San Juan clinic. Autoscope’s diagnostic performance was poorer than routine microscopy on slides from the Santa Clara clinic, which generated slides with lower blood volumes. Results of the study reflect both the potential for artificial intelligence to perform tasks currently conducted by highly-trained experts, and the challenges of replicating the adaptiveness of human thought processes.

Electronic supplementary material

The online version of this article (10.1186/s12936-018-2493-0) contains supplementary material, which is available to authorized users.

An Analysis of the Interaction Between Intelligent Software Agents and Human Users

Interactions between an intelligent software agent (ISA) and a human user are ubiquitous in everyday situations such as access to information, entertainment, and purchases. In such interactions, the ISA mediates the user’s access to the content, or controls some other aspect of the user experience, and is not designed to be neutral about outcomes of user choices. Like human users, ISAs are driven by goals, make autonomous decisions, and can learn from experience. Using ideas from bounded rationality (and deploying concepts from artificial intelligence, behavioural economics, control theory, and game theory), we frame these interactions as instances of an ISA whose reward depends on actions performed by the user. Such agents benefit by steering the user’s behaviour towards outcomes that maximise the ISA’s utility, which may or may not be aligned with that of the user. Video games, news recommendation aggregation engines, and fitness trackers can all be instances of this general case. Our analysis facilitates distinguishing various subcases of interaction (i.e. deception, coercion, trading, and nudging), as well as second-order effects that might include the possibility for adaptive interfaces to induce behavioural addiction, and/or change in user belief. We present these types of interaction within a conceptual framework, and review current examples of persuasive technologies and the issues that arise from their use. We argue that the nature of the feedback commonly used by learning agents to update their models and subsequent decisions could steer the behaviour of human users away from what benefits them, and in a direction that can undermine autonomy and cause further disparity between actions and goals as exemplified by addictive and compulsive behaviour. We discuss some of the ethical, social and legal implications of this technology and argue that it can sometimes exploit and reinforce weaknesses in human beings.

Digital diabetes: Perspectives for diabetes prevention, management and research

Digital medicine, digital research and artificial intelligence (AI) have the power to transform the field of diabetes with continuous and no-burden remote monitoring of patients' symptoms, physiological data, behaviours, and social and environmental contexts through the use of wearables, sensors and smartphone technologies. Moreover, data generated online and by digital technologies - which the authors suggest be grouped under the term 'digitosome' - constitute, through the quantity and variety of information they represent, a powerful potential for identifying new digital markers and patterns of risk that, ultimately, when combined with clinical data, can improve diabetes management and quality of life, and also prevent diabetes-related complications. Moving from a world in which patients are characterized by only a few recent measurements of fasting glucose levels and glycated haemoglobin to a world where patients, healthcare professionals and research scientists can consider various key parameters at thousands of time points simultaneously will profoundly change the way diabetes is prevented, managed and characterized in patients living with diabetes, as well as how it is scientifically researched. Indeed, the present review looks at how the digitization of diabetes can impact all fields of diabetes - its prevention, management, technology and research - and how it can complement, but not replace, what is usually done in traditional clinical settings. Such a profound shift is a genuine game changer that should be embraced by all, as it can provide solid research results transferable to patients, improve general health literacy, and provide tools to facilitate the everyday decision-making process by both healthcare professionals and patients living with diabetes.

Artificial intelligence using neural network architecture for radiology (AINNAR): classification of MR imaging sequences

PURPOSE

The confusion of MRI sequence names could be solved if MR images were automatically identified after image data acquisition. We revealed the ability of deep learning to classify head MRI sequences.

MATERIALS AND METHODS

Seventy-eight patients with mild cognitive impairment (MCI) having apparently normal head MR images and 78 intracranial hemorrhage (ICH) patients with morphologically deformed head MR images were enrolled. Six imaging protocols were selected to be performed: T2-weighted imaging, fluid attenuated inversion recovery imaging, T2-star-weighted imaging, diffusion-weighted imaging, apparent diffusion coefficient mapping, and source images of time-of-flight magnetic resonance angiography. The proximal first image slices and middle image slices having ambiguous and distinctive contrast patterns, respectively, were classified by two deep learning imaging classifiers, AlexNet and GoogLeNet.

RESULTS

AlexNet had accuracies of 73.3%, 73.6%, 73.1%, and 60.7% in the middle slices of MCI group, middle slices of ICH group, first slices of MCI group, and first slices of ICH group, while GoogLeNet had accuracies of 100%, 98.1%, 93.1%, and 94.8%, respectively. AlexNet significantly had lower classification ability than GoogLeNet for all datasets.

CONCLUSIONS

GoogLeNet could judge the types of head MRI sequences with a small amount of training data, irrespective of morphological or contrast conditions.

Fluorescence probes for real-time remote cyanobacteria monitoring: A review of challenges and opportunities

In recent years, there has been a widespread deployment of submersible fluorescence sensors by water utilities. They are used to measure diagnostic pigments and estimate algae and cyanobacteria abundance in near real-time. Despite being useful and promising tools, operators and decision-makers often rely on the data provided by these probes without a full understanding of their limitations. As a result, this may lead to wrong and misleading estimations which, in turn, means that researchers and technicians distrust these sensors. In this review paper, we list and discuss the main limitations of such probes, as well as identifying the effect of environmental factors on pigment production, and in turn, the conversion to cyanobacteria abundance estimation. We argue that a comprehensive calibration approach to obtain reliable readings goes well beyond manufacturers' recommendations, and should involve several context-specific experiments. We also believe that if such a comprehensive set of experiments is conducted, the data collected from fluorescence sensors could be used in artificial intelligence modelling approaches to reliably predict, in near real-time, the presence and abundance of different cyanobacteria species. This would have significant benefits for both drinking and recreational water management, given that cyanobacterial toxicity, and taste and odour compounds production, are species-dependent.

[The development of ophthalmology in artificial intelligence era]

Ophthalmology is a discipline that is highly dependent on the development of technology. Artificial intelligence is a new technological revolution, which will in all-round and fundamentally influence the progression of modern ophthalmology. Ophthalmologists should actively pay close attention to the development of artificial intelligence technology with great enthusiasm, and gradually realize the maximum utilization of artificial intelligence technology so as to facilitate the new development in ophthalmology. We should proactively seek for the opportunities of cooperating with the research organization and expert engineers in the area of artificial intelligence to promote the application of artificial intelligence in ophthalmology as soon as possible. Ophthalmology is likely to be radically changed by artificial intelligence technology. (Chin J Ophthalmol, 2018, 54: 645-648).

Machine learning: applications of artificial intelligence to imaging and diagnosis

Machine learning (ML) is a form of artificial intelligence which is placed to transform the twenty-first century. Rapid, recent progress in its underlying architecture and algorithms and growth in the size of datasets have led to increasing computer competence across a range of fields. These include driving a vehicle, language translation, chatbots and beyond human performance at complex board games such as Go. Here, we review the fundamentals and algorithms behind machine learning and highlight specific approaches to learning and optimisation. We then summarise the applications of ML to medicine. In particular, we showcase recent diagnostic performances, and caveats, in the fields of dermatology, radiology, pathology and general microscopy.

Hybrid algorithmic approach oriented to incipient rotor fault diagnosis on induction motors

This paper investigates the current monitoring for effective fault diagnosis in induction motor (IM) by using random forest (RF) algorithms. A rotor bar breakage of IM does not derive in a catastrophic fault but its timely detection can avoid catastrophic consequences in the stator or prevent malfunctioning of those applications in which this sort of fault is the primary concern. Current-based fault signatures depend enormously on the IM power source and in the load connected to the motor. Hence, homogeneous sets of current signals were acquired through multiple experiments at particular loading torques and IM feedings from an experimental test bench in which incipient rotor severities were considered. Understanding the importance of each fault signature in relation to its diagnosis performance is an interesting matter. To this end, we propose a hybrid approach based on Simulated Annealing algorithm to conduct a global search over the computed feature set for feature selection purposes, which reduce the computational requirements of the diagnosis tool. Then, a novel Oblique RF classifier is used to build multivariate trees, which explicitly learn optimal split directions at internal nodes through penalized Ridge regression. This algorithm has been compared with other state-of-the-art classifiers through careful evaluation of performance measures not encountered in this field.

Clear Cell Renal Cell Carcinoma: Machine Learning-Based Quantitative Computed Tomography Texture Analysis for Prediction of Fuhrman Nuclear Grade

OBJECTIVE

To evaluate the performance of quantitative computed tomography (CT) texture analysis using different machine learning (ML) classifiers for discriminating low and high nuclear grade clear cell renal cell carcinomas (cc-RCCs).

MATERIALS AND METHODS

This retrospective study included 53 patients with pathologically proven 54 cc-RCCs (31 low-grade [grade 1 or 2]; 23 high-grade [grade 3 or 4]). In one patient, two synchronous cc-RCCs were included in the analysis. Mean age was 57.5 years. Thirty-four (64.1%) patients were male and 19 were female (35.9%). Mean tumour size based on the maximum diameter was 57.4 mm (range, 16-145 mm). Forty patients underwent radical nephrectomy and 13 underwent partial nephrectomy. Following pre-processing steps, two-dimensional CT texture features were extracted using portal-phase contrast-enhanced CT. Reproducibility of texture features was assessed with the intra-class correlation coefficient (ICC). Nested cross-validation with a wrapper-based algorithm was used in feature selection and model optimisation. The ML classifiers were support vector machine (SVM), multilayer perceptron (MLP, a sort of neural network), naïve Bayes, k-nearest neighbours, and random forest. The performance of the classifiers was compared by certain metrics.

RESULTS

Among 279 texture features, 241 features with an ICC equal to or higher than 0.80 (excellent reproducibility) were included in the further feature selection process. The best model was created using SVM. The selected subset of features for SVM included five co-occurrence matrix (ICC range, 0.885-0.998), three run-length matrix (ICC range, 0.889-0.992), one gradient (ICC = 0.998), and four Haar wavelet features (ICC range, 0.941-0.997). The overall accuracy, sensitivity (for detecting high-grade cc-RCCs), specificity (for detecting high-grade cc-RCCs), and overall area under the curve of the best model were 85.1%, 91.3%, 80.6%, and 0.860, respectively.

CONCLUSIONS

The ML-based CT texture analysis can be a useful and promising non-invasive method for prediction of low and high Fuhrman nuclear grade cc-RCCs.

KEY POINTS

• Based on the percutaneous biopsy literature, ML-based CT texture analysis has a comparable predictive performance with percutaneous biopsy. • Highest predictive performance was obtained with use of the SVM. • SVM correctly classified 85.1% of cc-RCCs in terms of nuclear grade, with an AUC of 0.860.

From hype to reality: data science enabling personalized medicine

BACKGROUND

Personalized, precision, P4, or stratified medicine is understood as a medical approach in which patients are stratified based on their disease subtype, risk, prognosis, or treatment response using specialized diagnostic tests. The key idea is to base medical decisions on individual patient characteristics, including molecular and behavioral biomarkers, rather than on population averages. Personalized medicine is deeply connected to and dependent on data science, specifically machine learning (often named Artificial Intelligence in the mainstream media). While during recent years there has been a lot of enthusiasm about the potential of 'big data' and machine learning-based solutions, there exist only few examples that impact current clinical practice. The lack of impact on clinical practice can largely be attributed to insufficient performance of predictive models, difficulties to interpret complex model predictions, and lack of validation via prospective clinical trials that demonstrate a clear benefit compared to the standard of care. In this paper, we review the potential of state-of-the-art data science approaches for personalized medicine, discuss open challenges, and highlight directions that may help to overcome them in the future.

CONCLUSIONS

There is a need for an interdisciplinary effort, including data scientists, physicians, patient advocates, regulatory agencies, and health insurance organizations. Partially unrealistic expectations and concerns about data science-based solutions need to be better managed. In parallel, computational methods must advance more to provide direct benefit to clinical practice.

Can the artificial intelligence technique of reinforcement learning use continuously-monitored digital data to optimize treatment for weight loss?

Behavioral weight loss (WL) trials show that, on average, participants regain lost weight unless provided long-term, intensive-and thus costly-intervention. Optimization solutions have shown mixed success. The artificial intelligence principle of "reinforcement learning" (RL) offers a new and more sophisticated form of optimization in which the intensity of each individual's intervention is continuously adjusted depending on patterns of response. In this pilot, we evaluated the feasibility and acceptability of a RL-based WL intervention, and whether optimization would achieve equivalent benefit at a reduced cost compared to a non-optimized intensive intervention. Participants (n = 52) completed a 1-month, group-based in-person behavioral WL intervention and then (in Phase II) were randomly assigned to receive 3 months of twice-weekly remote interventions that were non-optimized (NO; 10-min phone calls) or optimized (a combination of phone calls, text exchanges, and automated messages selected by an algorithm). The Individually-Optimized (IO) and Group-Optimized (GO) algorithms selected interventions based on past performance of each intervention for each participant, and for each group member that fit into a fixed amount of time (e.g., 1 h), respectively. Results indicated that the system was feasible to deploy and acceptable to participants and coaches. As hypothesized, we were able to achieve equivalent Phase II weight losses (NO = 4.42%, IO = 4.56%, GO = 4.39%) at roughly one-third the cost (1.73 and 1.77 coaching hours/participant for IO and GO, versus 4.38 for NO), indicating strong promise for a RL system approach to weight loss and maintenance.

[Robotics and augmented reality : Current state of development and future perspectives]

BACKGROUND

Digitalization in surgery is gaining attention in the surgical community, with robotics and augmented reality as key issues.

ROBOTICS

The term surgical robot is basically not adequate to describe currently available telesupport and manipulation systems. These are passive tools which have to be activated by the surgeon and only provide relatively low levels of active support. Accordingly, justification of use is currently difficult with respect to the cost-benefit relationship. A real breakthrough will be achieved by upgrading them into genuine intelligent and collaborative support systems and justify the term as the true meaning of robotics.

AUGMENTED REALITY (AR)

Augmented or enriched reality improves or facilitates normal sensory perception by the integration of additional information of a different nature. Intuitive perception of the surgical site would have the potential to revolutionize surgery, but prior to clinical use, the matching of the real and the virtual world still has to be optimized (referencing); however, AR is now already a valuable tool for training and simulation as well as workflow support in the operating room (OR).

CRITICAL COMMENT AND PERSPECTIVES

The promising new technological development towards the future cooperative surgical OR environment, including both robotic and AR modules, will have a significant impact on surgery, even in the mid-term. Decisive for this is that surgeons actively take part in the evaluation of this process to ensure that future "intelligent" tools will remain mere assistant or supporting systems.

Computer-aided detection in chest radiography based on artificial intelligence: a survey

As the most common examination tool in medical practice, chest radiography has important clinical value in the diagnosis of disease. Thus, the automatic detection of chest disease based on chest radiography has become one of the hot topics in medical imaging research. Based on the clinical applications, the study conducts a comprehensive survey on computer-aided detection (CAD) systems, and especially focuses on the artificial intelligence technology applied in chest radiography. The paper presents several common chest X-ray datasets and briefly introduces general image preprocessing procedures, such as contrast enhancement and segmentation, and bone suppression techniques that are applied to chest radiography. Then, the CAD system in the detection of specific disease (pulmonary nodules, tuberculosis, and interstitial lung diseases) and multiple diseases is described, focusing on the basic principles of the algorithm, the data used in the study, the evaluation measures, and the results. Finally, the paper summarizes the CAD system in chest radiography based on artificial intelligence and discusses the existing problems and trends.

The "Use" of Sex Robots: A Bioethical Issue

The manufacture of humanoid robots with embedded artificial intelligence and for sexual purposes has generated some debates within bioethics, in which diverse competing views have been presented. Themes such as sexuality and its deviations, the objectification of women, the relational problems of contemporary life, loneliness, and even the reproductive future of the species constitute the arguments which have emerged in relation to this subject. Based on these themes, this article presents the current state of the use of female sex robots, the bioethical problems that arise, and how bioethics could serve as a medium for both thinking about and resolving some of these challenges.

Artificial intelligence as a medical device in radiology: ethical and regulatory issues in Europe and the United States

Abstract

Worldwide interest in artificial intelligence (AI) applications is growing rapidly. In medicine, devices based on machine/deep learning have proliferated, especially for image analysis, presaging new significant challenges for the utility of AI in healthcare. This inevitably raises numerous legal and ethical questions. In this paper we analyse the state of AI regulation in the context of medical device development, and strategies to make AI applications safe and useful in the future. We analyse the legal framework regulating medical devices and data protection in Europe and in the United States, assessing developments that are currently taking place. The European Union (EU) is reforming these fields with new legislation (General Data Protection Regulation [GDPR], Cybersecurity Directive, Medical Devices Regulation, In Vitro Diagnostic Medical Device Regulation). This reform is gradual, but it has now made its first impact, with the GDPR and the Cybersecurity Directive having taken effect in May, 2018. As regards the United States (U.S.), the regulatory scene is predominantly controlled by the Food and Drug Administration. This paper considers issues of accountability, both legal and ethical. The processes of medical device decision-making are largely unpredictable, therefore holding the creators accountable for it clearly raises concerns. There is a lot that can be done in order to regulate AI applications. If this is done properly and timely, the potentiality of AI based technology, in radiology as well as in other fields, will be invaluable.

Teaching Points

• AI applications are medical devices supporting detection/diagnosis, work-flow, cost-effectiveness.

• Regulations for safety, privacy protection, and ethical use of sensitive information are needed.

• EU and U.S. have different approaches for approving and regulating new medical devices.

• EU laws consider cyberattacks, incidents (notification and minimisation), and service continuity.

• U.S. laws ask for opt-in data processing and use as well as for clear consumer consent.

Improving accuracy of corneal power measurement with partial coherence interferometry after corneal refractive surgery using a multivariate polynomial approach

Background

To improve accuracy of IOLMaster (Carl Zeiss, Jena, Germany) in corneal power measurement after myopic excimer corneal refractive surgery (MECRS) using multivariate polynomial analysis (MPA).

Methods

One eye of each of 403 patients (mean age 31.53 ± 8.47 years) was subjected to MECRS for a myopic defect, measured as spherical equivalent, ranging from − 9.50 to − 1 D (mean − 4.55 ± 2.20 D). Each patient underwent a complete eye examination and IOLMaster scan before surgery and at 1, 3 and 6 months follow up. Axial length (AL), flatter keratometry value (K1), steeper keratometry value (K2), mean keratometry value (KM) and anterior chamber depth measured from the corneal endothelium to the anterior surface of the lens (ACD) were used in a MPA to devise a method to improve accuracy of KM measurements.

Results

Using AL, K1, K2 and ACD measured after surgery in polynomial degree 2 analysis, mean error of corneal power evaluation after MECRS was + 0.16 ± 0.19 D.

Conclusions

MPA was found to be an effective tool in devising a method to improve precision in corneal power evaluation in eyes previously subjected to MECRS, according to our results.

[Big data and artificial intelligence for diagnostic decision support in atypical dementia]

The differential diagnosis of atypical dementia remains difficult. The use of positron emission tomography (PET) still represents the gold standard for imaging diagnostics. According to the current evidence, however, magnetic resonance imaging (MRI) is almost equal to fluorodeoxyglucose (FDG)-PET, but only when using new big data and machine learning methods. In cases of atypical dementia, especially in younger patients and for follow-up, MRI is preferable to computed tomography (CT). In the clinical routine, promising MRI procedures are e. g. the automated volumetry of anatomical 3D images, as well as a non-contrast-enhanced MRI perfusion method, called arterial spin labeling (ASL). Because of the rapidly growing amount of biomarker data, there is a need for computer-aided big data analyses and artificial intelligence. Based on fast analyses of the diverse and rapidly increasing amount of clinical, imaging, epidemiological, molecular genetic and economic data, new knowledge on the pathogenesis, prevention and treatment can be generated. Technical availability, homogenization of the underlying data and the availability of large reference data are the basis for the widespread establishment of promising analytical methods.

Artificial Intelligence Approaches in Hematopoietic Cell Transplantation: A Review of the Current Status and Future Directions

The evidence-based literature on healthcare is currently expanding exponentially. The opportunities provided by the advancement in artificial intelligence (AI) tools such as machine learning are appealing in tackling many of the current healthcare challenges. Thus, AI integration is expanding in most fields of healthcare, including the field of hematology. This study aims to review the current applications of AI in the field of hematopoietic cell transplantation (HCT). A literature search was done involving the following databases: Ovid MEDLINE, including In-Process and other non-indexed citations, and Google Scholar. The abstracts of the following professional societies were also screened: American Society of Hematology, American Society for Blood and Marrow Transplantation, and European Society for Blood and Marrow Transplantation. The literature review showed that the integration of AI in the field of HCT has grown remarkably in the last decade and offers promising avenues in diagnosis and prognosis in HCT populations targeting both pre- and post-transplant challenges. Studies of AI integration in HCT have many limitations that include poorly tested algorithms, lack of generalizability, and limited use of different AI tools. Machine learning techniques in HCT are an intense area of research that needs much development and extensive support from hematology and HCT societies and organizations globally as we believe that this will be the future practice paradigm.

Big data and machine learning in critical care: Opportunities for collaborative research

The introduction of clinical information systems (CIS) in Intensive Care Units (ICUs) offers the possibility of storing a huge amount of machine-ready clinical data that can be used to improve patient outcomes and the allocation of resources, as well as suggest topics for randomized clinical trials. Clinicians, however, usually lack the necessary training for the analysis of large databases. In addition, there are issues referred to patient privacy and consent, and data quality. Multidisciplinary collaboration among clinicians, data engineers, machine-learning experts, statisticians, epidemiologists and other information scientists may overcome these problems. A multidisciplinary event (Critical Care Datathon) was held in Madrid (Spain) from 1 to 3 December 2017. Under the auspices of the Spanish Critical Care Society (SEMICYUC), the event was organized by the Massachusetts Institute of Technology (MIT) Critical Data Group (Cambridge, MA, USA), the Innovation Unit and Critical Care Department of San Carlos Clinic Hospital, and the Life Supporting Technologies group of Madrid Polytechnic University. After presentations referred to big data in the critical care environment, clinicians, data scientists and other health data science enthusiasts and lawyers worked in collaboration using an anonymized database (MIMIC III). Eight groups were formed to answer different clinical research questions elaborated prior to the meeting. The event produced analyses for the questions posed and outlined several future clinical research opportunities. Foundations were laid to enable future use of ICU databases in Spain, and a timeline was established for future meetings, as an example of how big data analysis tools have tremendous potential in our field.

Machine learning in neurology: what neurologists can learn from machines and vice versa

Artificial intelligence is increasingly becoming a part of everyday life. This raises the question whether clinical neurology can benefit from these novel methods to increase diagnostic accuracy. Several recent studies have used machine learning classifiers to predict whether subjects suffer from a neurological disorder. This article discusses whether these methods are ready to make their entrance into clinical practice. The underlying principles of classification will be explored, as well as the potential pitfalls. Strengths of machine learning methods are that they are unbiased and very sensitive to patterns emerging from small changes spread across a large number of variables. Potential pitfalls are that building reliable classifiers requires large amounts of well-selected data and extensive validation. Currently, machine learning classifiers offer neurologists a new diagnostic tool which can aid in the diagnosis of cases with a high degree of uncertainty.

Futuristic biosensors for cardiac health care: an artificial intelligence approach

Biosensor-based devices are pioneering in the modern biomedical applications and will be the future of cardiac health care. The coupling of artificial intelligence (AI) for cardiac monitoring-based biosensors for the point of care (POC) diagnostics is prominently reviewed here. This review deciphers the most significant machine-learning algorithms for the futuristic biosensors along with the internet of things, computational techniques and microchip-based essential cardiac biomarkers for real-time health monitoring and improving patient compliance. The present review also discusses the recently developed cardiac biosensors along with technical strategies involved in their mechanism of working and their applications in healthcare. Additionally, it provides a key for the ontogeny of an effective and supportive hierarchical protocol for clinical decision-making about personalized medicine through combinatory information analysis, and integrated multidisciplinary AI approaches.

Ethical Design of Intelligent Assistive Technologies for Dementia: A Descriptive Review

The use of Intelligent Assistive Technology (IAT) in dementia care opens the prospects of reducing the global burden of dementia and enabling novel opportunities to improve the lives of dementia patients. However, with current adoption rates being reportedly low, the potential of IATs might remain under-expressed as long as the reasons for suboptimal adoption remain unaddressed. Among these, ethical and social considerations are critical. This article reviews the spectrum of IATs for dementia and investigates the prevalence of ethical considerations in the design of current IATs. Our screening shows that a significant portion of current IATs is designed in the absence of explicit ethical considerations. These results suggest that the lack of ethical consideration might be a codeterminant of current structural limitations in the translation of IATs from designing labs to bedside. Based on these data, we call for a coordinated effort to proactively incorporate ethical considerations early in the design and development of new products.

Integration of artificial intelligence methods and life cycle assessment to predict energy output and environmental impacts of paddy production

Prediction of agricultural energy output and environmental impacts play important role in energy management and conservation of environment as it can help us to evaluate agricultural energy efficiency, conduct crops production system commissioning, and detect and diagnose faults of crop production system. Agricultural energy output and environmental impacts can be readily predicted by artificial intelligence (AI), owing to the ease of use and adaptability to seek optimal solutions in a rapid manner as well as the use of historical data to predict future agricultural energy use pattern under constraints. This paper conducts energy output and environmental impact prediction of paddy production in Guilan province, Iran based on two AI methods, artificial neural networks (ANNs), and adaptive neuro fuzzy inference system (ANFIS). The amounts of energy input and output are 51,585.61MJkg^-1 and 66,112.94MJkg^-1, respectively, in paddy production. Life Cycle Assessment (LCA) is used to evaluate environmental impacts of paddy production. Results show that, in paddy production, in-farm emission is a hotspot in global warming, acidification and eutrophication impact categories. ANN model with 12-6-8-1 structure is selected as the best one for predicting energy output. The correlation coefficient (R) varies from 0.524 to 0.999 in training for energy input and environmental impacts in ANN models. ANFIS model is developed based on a hybrid learning algorithm, with R for predicting output energy being 0.860 and, for environmental impacts, varying from 0.944 to 0.997. Results indicate that the multi-level ANFIS is a useful tool to managers for large-scale planning in forecasting energy output and environmental indices of agricultural production systems owing to its higher speed of computation processes compared to ANN model, despite ANN's higher accuracy.

Machine Learning Interface for Medical Image Analysis

TensorFlow is a second-generation open-source machine learning software library with a built-in framework for implementing neural networks in wide variety of perceptual tasks. Although TensorFlow usage is well established with computer vision datasets, the TensorFlow interface with DICOM formats for medical imaging remains to be established. Our goal is to extend the TensorFlow API to accept raw DICOM images as input; 1513 DaTscan DICOM images were obtained from the Parkinson's Progression Markers Initiative (PPMI) database. DICOM pixel intensities were extracted and shaped into tensors, or n-dimensional arrays, to populate the training, validation, and test input datasets for machine learning. A simple neural network was constructed in TensorFlow to classify images into normal or Parkinson's disease groups. Training was executed over 1000 iterations for each cross-validation set. The gradient descent optimization and Adagrad optimization algorithms were used to minimize cross-entropy between the predicted and ground-truth labels. Cross-validation was performed ten times to produce a mean accuracy of 0.938 ± 0.047 (95 % CI 0.908-0.967). The mean sensitivity was 0.974 ± 0.043 (95 % CI 0.947-1.00) and mean specificity was 0.822 ± 0.207 (95 % CI 0.694-0.950). We extended the TensorFlow API to enable DICOM compatibility in the context of DaTscan image analysis. We implemented a neural network classifier that produces diagnostic accuracies on par with excellent results from previous machine learning models. These results indicate the potential role of TensorFlow as a useful adjunct diagnostic tool in the clinical setting.

Detection and diagnosis of dental caries using a deep learning-based convolutional neural network algorithm

OBJECTIVES

Deep convolutional neural networks (CNNs) are a rapidly emerging new area of medical research, and have yielded impressive results in diagnosis and prediction in the fields of radiology and pathology. The aim of the current study was to evaluate the efficacy of deep CNN algorithms for detection and diagnosis of dental caries on periapical radiographs.

MATERIALS AND METHODS

A total of 3000 periapical radiographic images were divided into a training and validation dataset (n = 2400 [80%]) and a test dataset (n = 600 [20%]). A pre-trained GoogLeNet Inception v3 CNN network was used for preprocessing and transfer learning. The diagnostic accuracy, sensitivity, specificity, positive predictive value, negative predictive value, receiver operating characteristic (ROC) curve, and area under the curve (AUC) were calculated for detection and diagnostic performance of the deep CNN algorithm.

RESULTS

The diagnostic accuracies of premolar, molar, and both premolar and molar models were 89.0% (80.4-93.3), 88.0% (79.2-93.1), and 82.0% (75.5-87.1), respectively. The deep CNN algorithm achieved an AUC of 0.917 (95% CI 0.860-0.975) on premolar, an AUC of 0.890 (95% CI 0.819-0.961) on molar, and an AUC of 0.845 (95% CI 0.790-0.901) on both premolar and molar models. The premolar model provided the best AUC, which was significantly greater than those for other models (P < 0.001).

CONCLUSIONS

This study highlighted the potential utility of deep CNN architecture for the detection and diagnosis of dental caries. A deep CNN algorithm provided considerably good performance in detecting dental caries in periapical radiographs. CLINICAL SIGNIfiCANCE: Deep CNN algorithms are expected to be among the most effective and efficient methods for diagnosing dental caries.

[Big data and deep learning in preventive and rehabilitation medicine]

The digitalization in medicine has led to almost universal availability of information to different healthcare professionals and accelerated clinical pathways. Fast-track concepts and short hospital stays require intelligent and practicable systems in preventive and rehabilitation medicine. This includes optimization of movement analysis by innovative tools such as detectors sensing skin movements, portable feedback systems for monitoring, robot-assisted devices, and prevention programs based on reliable data. Finally, clinical structures are needed to exploit the maximal potential of artificial intelligence (AI) and deep learning. One example is the establishment of inter- and transdisciplinary professional teams such as a RehaBoard. In contrast to other cost-intensive disciplines such as oncology, the introduction of AI into rehabilitation orthopedics and trauma surgery with the support of cross-sectoral cooperation has great potential for performing well in patient benefit-orientated competition (value-based competition).

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 2: Validation

ESA defines as Earth Observation (EO) Level 2 information product a multi-spectral (MS) image corrected for atmospheric, adjacency, and topographic effects, stacked with its data-derived scene classification map (SCM), whose legend includes quality layers cloud and cloud-shadow. No ESA EO Level 2 product has ever been systematically generated at the ground segment. To fill the information gap from EO big data to ESA EO Level 2 product in compliance with the GEO-CEOS stage 4 validation (Val) guidelines, an off-the-shelf Satellite Image Automatic Mapper (SIAM) lightweight computer program was selected to be validated by independent means on an annual 30 m resolution Web-Enabled Landsat Data (WELD) image composite time-series of the conterminous U.S. (CONUS) for the years 2006 to 2009. The SIAM core is a prior knowledge-based decision tree for MS reflectance space hyperpolyhedralization into static (non-adaptive to data) color names. For the sake of readability, this paper was split into two. The present Part 2-Validation-accomplishes a GEO-CEOS stage 4 Val of the test SIAM-WELD annual map time-series in comparison with a reference 30 m resolution 16-class USGS National Land Cover Data (NLCD) 2006 map. These test and reference map pairs feature the same spatial resolution and spatial extent, but their legends differ and must be harmonized, in agreement with the previous Part 1 - Theory. Conclusions are that SIAM systematically delivers an ESA EO Level 2 SCM product instantiation whose legend complies with the standard 2-level 4-class FAO Land Cover Classification System (LCCS) Dichotomous Phase (DP) taxonomy.

Estimating risk of severe neonatal morbidity in preterm births under 32 weeks of gestation

Background: A large recent study analyzed the relationship between multiple factors and neonatal outcome and in preterm births. Study variables included the reason for admission, indication for delivery, optimal steroid use, gestational age, and other potential prognostic factors. Using stepwise multivariable analysis, the only two variables independently associated with serious neonatal morbidity were gestational age and the presence of suspected intrauterine growth restriction as a reason for admission. This finding was surprising given the beneficial effects of antenatal steroids and hazards associated with some causes of preterm birth. Multivariable logistic regression techniques have limitations. Without testing for multiple interactions, linear regression will identify only individual factors with the strongest independent relationship to the outcome for the entire study group. There may not be a single "best set" of risk factors or one set that applies equally well to all subgroups. In contrast, machine learning techniques find the most predictive groupings of factors based on their frequency and strength of association, with no attempt to identify independence and no assumptions about linear relationships.Objective: To determine if machine learning techniques would identify specific clusters of conditions with different probability estimates for severe neonatal morbidity and to compare these findings to those based on the original multivariable analysis.Materials and methods: This was a secondary analysis of data collected in a multicenter, prospective study on all admissions to the neonatal intensive care unit between 2013 and 2015 in 10 hospitals. We included all patients with a singleton, stillborn, or live newborns, with a gestational age between 23 0/7 and 31 6/7 week. The composite endpoint, severe neonatal morbidity, defined by the presence of any of five outcomes: death, grade 3 or 4 intraventricular hemorrhage (IVH), and ≥28 days on ventilator, periventricular leukomalacia (PVL), or stage III necrotizing enterocolitis (NEC), was present in 238 of the 1039 study patients. We studied five explanatory variables: maternal age, parity, gestational age, admission reason, and status with respect to antenatal steroid administration. We concentrated on Classification and Regression Trees because the resulting structure defines clusters of risk factors that often bear resemblance to clinical reasoning. Model performance was measured using area under the receiver-operator characteristic curves (AUC) based on 10 repetitions of 10-fold cross-validation.Results: A hybrid technique using a combination of logistic regression and Classification and Regression Trees had a mean cross-validated AUC of 0.853. A selected point on its receiver-operator characteristic (ROC) curve corresponding to a sensitivity of 81% was associated with a specificity of 76%. Rather than a single curve representing the general relationship between gestational age and severe morbidity, this technique found seven clusters with distinct curves. Abnormal fetal testing as a reason for admission with or without growth restriction and incomplete steroid administration would place a 20-year-old patient on the highest risk curve.Conclusions: Using a relatively small database and a few simple factors known before birth it is possible to produce a more tailored estimate of the risk for severe neonatal morbidity on which clinicians can superimpose their medical judgment, experience, and intuition.

Will artificial intelligence solve the human resource crisis in healthcare?

Artificial intelligence (AI) has the potential to ease the human resources crisis in healthcare by facilitating diagnostics, decision-making, big data analytics and administration, among others. For this we must first tackle the technological, ethical and legal obstacles.The human resource crisis is widening worldwide, and it is obvious that it is not possible to provide care without workforce. How can disruptive technologies in healthcare help solve the variety of human resource problems? Will technology empower physicians or replace them? How can the medical curriculum, including post-graduate education prepare professionals for the meaningful use of technology? These questions have been growing for decades, and the promise of disruptive technologies filling them is imminent with digital health becoming widespread. Authors of this essay argue that AI might not only fill the human resources gap, but also raises ethical questions we need to deal with today.While there are even more questions to address, our stand is that AI is not meant to replace caregivers, but those who use AI will probably replace those who don't. And it is possible to prepare for that.

A new self-learning computational method for footprints of early human migration processes

We present a new self-learning computational method searching for footprints of early migration processes determining the genetic compositions of recent human populations. The data being analysed are 26- and 18-dimensional mitochondrial and Y-chromosomal haplogroup distributions representing 50 recent and 34 ancient populations in Eurasia and America. The algorithms search for associations of haplogroups jointly propagating in a significant subset of these populations. Joint propagations of Hgs are detected directly by similar ranking lists of populations derived from Hg frequencies of the 50 Hg distributions. The method provides us the most characteristic associations of mitochondrial and Y-chromosomal haplogroups, and the set of populations where these associations propagate jointly. In addition, the typical ranking lists characterizing these Hg associations show the geographical distribution, the probable place of origin and the paths of their protection. Comparison to ancient data verifies that these recent geographical distributions refer to the most important prehistoric migrations supported by archaeological evidences.

Future Direction for Using Artificial Intelligence to Predict and Manage Hypertension

PURPOSE OF REVIEW

Evidence that artificial intelligence (AI) is useful for predicting risk factors for hypertension and its management is emerging. However, we are far from harnessing the innovative AI tools to predict these risk factors for hypertension and applying them to personalized management. This review summarizes recent advances in the computer science and medical field, illustrating the innovative AI approach for potential prediction of early stages of hypertension. Additionally, we review ongoing research and future implications of AI in hypertension management and clinical trials, with an eye towards personalized medicine.

RECENT FINDINGS

Although recent studies demonstrate that AI in hypertension research is feasible and possibly useful, AI-informed care has yet to transform blood pressure (BP) control. This is due, in part, to lack of data on AI's consistency, accuracy, and reliability in the BP sphere. However, many factors contribute to poorly controlled BP, including biological, environmental, and lifestyle issues. AI allows insight into extrapolating data analytics to inform prescribers and patients about specific factors that may impact their BP control. To date, AI has been mainly used to investigate risk factors for hypertension, but has not yet been utilized for hypertension management due to the limitations of study design and of physician's engagement in computer science literature. The future of AI with more robust architecture using multi-omics approaches and wearable technology will likely be an important tool allowing to incorporate biological, lifestyle, and environmental factors into decision-making of appropriate drug use for BP control.

Medical students' attitude towards artificial intelligence: a multicentre survey

OBJECTIVES

To assess undergraduate medical students' attitudes towards artificial intelligence (AI) in radiology and medicine.

MATERIALS AND METHODS

A web-based questionnaire was designed using SurveyMonkey, and was sent out to students at three major medical schools. It consisted of various sections aiming to evaluate the students' prior knowledge of AI in radiology and beyond, as well as their attitude towards AI in radiology specifically and in medicine in general. Respondents' anonymity was ensured.

RESULTS

A total of 263 students (166 female, 94 male, median age 23 years) responded to the questionnaire. Around 52% were aware of the ongoing discussion about AI in radiology and 68% stated that they were unaware of the technologies involved. Respondents agreed that AI could potentially detect pathologies in radiological examinations (83%) but felt that AI would not be able to establish a definite diagnosis (56%). The majority agreed that AI will revolutionise and improve radiology (77% and 86%), while disagreeing with statements that human radiologists will be replaced (83%). Over two-thirds agreed on the need for AI to be included in medical training (71%). In sub-group analyses male and tech-savvy respondents were more confident on the benefits of AI and less fearful of these technologies.

CONCLUSION

Contrary to anecdotes published in the media, undergraduate medical students do not worry that AI will replace human radiologists, and are aware of the potential applications and implications of AI on radiology and medicine. Radiology should take the lead in educating students about these emerging technologies.

KEY POINTS

• Medical students are aware of the potential applications and implications of AI in radiology and medicine in general. • Medical students do not worry that the human radiologist or physician will be replaced. • Artificial intelligence should be included in medical training.

Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic images

BACKGROUND

Image recognition using artificial intelligence with deep learning through convolutional neural networks (CNNs) has dramatically improved and been increasingly applied to medical fields for diagnostic imaging. We developed a CNN that can automatically detect gastric cancer in endoscopic images.

METHODS

A CNN-based diagnostic system was constructed based on Single Shot MultiBox Detector architecture and trained using 13,584 endoscopic images of gastric cancer. To evaluate the diagnostic accuracy, an independent test set of 2296 stomach images collected from 69 consecutive patients with 77 gastric cancer lesions was applied to the constructed CNN.

RESULTS

The CNN required 47 s to analyze 2296 test images. The CNN correctly diagnosed 71 of 77 gastric cancer lesions with an overall sensitivity of 92.2%, and 161 non-cancerous lesions were detected as gastric cancer, resulting in a positive predictive value of 30.6%. Seventy of the 71 lesions (98.6%) with a diameter of 6 mm or more as well as all invasive cancers were correctly detected. All missed lesions were superficially depressed and differentiated-type intramucosal cancers that were difficult to distinguish from gastritis even for experienced endoscopists. Nearly half of the false-positive lesions were gastritis with changes in color tone or an irregular mucosal surface.

CONCLUSION

The constructed CNN system for detecting gastric cancer could process numerous stored endoscopic images in a very short time with a clinically relevant diagnostic ability. It may be well applicable to daily clinical practice to reduce the burden of endoscopists.

Digital Radiographic Image Processing and Analysis

This article describes digital radiographic imaging and analysis from the basics of image capture to examples of some of the most advanced digital technologies currently available. The principles underlying the imaging technologies are described to provide a better understanding of their strengths and limitations.

Grid-like units help deep learning agent to navigate

An artificial-intelligence model based on deep learning developed units in a hidden layer that resembled mammalian grid cells in the hippocampus when the agent was taught to integrate paths. The full model performed sophisticated navigational tasks-in some cases even better than a human.

Fully automated detection and segmentation of meningiomas using deep learning on routine multiparametric MRI

Objectives

Magnetic resonance imaging (MRI) is the method of choice for imaging meningiomas. Volumetric assessment of meningiomas is highly relevant for therapy planning and monitoring. We used a multiparametric deep-learning model (DLM) on routine MRI data including images from diverse referring institutions to investigate DLM performance in automated detection and segmentation of meningiomas in comparison to manual segmentations.

Methods

We included 56 of 136 consecutive preoperative MRI datasets [T1/T2-weighted, T1-weighted contrast-enhanced (T1CE), FLAIR] of meningiomas that were treated surgically at the University Hospital Cologne and graded histologically as tumour grade I (n = 38) or grade II (n = 18). The DLM was trained on an independent dataset of 249 glioma cases and segmented different tumour classes as defined in the brain tumour image segmentation benchmark (BRATS benchmark). The DLM was based on the DeepMedic architecture. Results were compared to manual segmentations by two radiologists in a consensus reading in FLAIR and T1CE.

Results

The DLM detected meningiomas in 55 of 56 cases. Further, automated segmentations correlated strongly with manual segmentations: average Dice coefficients were 0.81 ± 0.10 (range, 0.46-0.93) for the total tumour volume (union of tumour volume in FLAIR and T1CE) and 0.78 ± 0.19 (range, 0.27-0.95) for contrast-enhancing tumour volume in T1CE.

Conclusions

The DLM yielded accurate automated detection and segmentation of meningioma tissue despite diverse scanner data and thereby may improve and facilitate therapy planning as well as monitoring of this highly frequent tumour entity.

Key Points

• Deep learning allows for accurate meningioma detection and segmentation

• Deep learning helps clinicians to assess patients with meningiomas

• Meningioma monitoring and treatment planning can be improved

Electronic supplementary material

The online version of this article (10.1007/s00330-018-5595-8) contains supplementary material, which is available to authorized users.

Computer-Aided Diagnosis of Thyroid Nodules via Ultrasonography: Initial Clinical Experience

Objective

To prospectively evaluate the diagnostic performance of computer-aided diagnosis (CAD) for detection of thyroid cancers via ultrasonography (US).

Materials and Methods

This study included 50 consecutive patients with 117 thyroid nodules on US during the period between June 2016 and July 2016. A radiologist performed US examinations using real-time CAD integrated into a US scanner. We compared the diagnostic performance of radiologist, the CAD system, and the CAD-assisted radiologist for the detection of thyroid cancers.

Results

The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of the CAD system were 80.0, 88.1, 83.3, 85.5, and 84.6%, respectively, and were not significantly different from those of the radiologist (p > 0.05). The CAD-assisted radiologist showed improved diagnostic sensitivity compared with the radiologist alone (92.0% vs. 84.0%, p = 0.037), while the specificity and PPV were reduced (85.1% vs. 95.5%, p = 0.005 and 82.1% vs. 93.3%, p = 0.008). The radiologist assisted by the CAD system exhibited better diagnostic sensitivity and NPV than the CAD system alone (92.0% vs. 80.0%, p = 0.009 and 93.4% vs. 88.9%, p = 0.013), while the specificities and PPVs were not significantly different (88.1% vs. 85.1%, p = 0.151 and 83.3% vs. 82.1%, p = 0.613, respectively).

Conclusion

The CAD system may be an adjunct to radiological intervention in the diagnosis of thyroid cancer.

Artificial intelligence in radiation oncology: A specialty-wide disruptive transformation?

Artificial intelligence (AI) is emerging as a technology with the power to transform established industries, and with applications from automated manufacturing to advertising and facial recognition to fully autonomous transportation. Advances in each of these domains have led some to call AI the "fourth" industrial revolution [1]. In healthcare, AI is emerging as both a productive and disruptive force across many disciplines. This is perhaps most evident in Diagnostic Radiology and Pathology, specialties largely built around the processing and complex interpretation of medical images, where the role of AI is increasingly seen as both a boon and a threat. In Radiation Oncology as well, AI seems poised to reshape the specialty in significant ways, though the impact of AI has been relatively limited at present, and may rightly seem more distant to many, given the predominantly interpersonal and complex interventional nature of the specialty. In this overview, we will explore the current state and anticipated future impact of AI on Radiation Oncology, in detail, focusing on key topics from multiple stakeholder perspectives, as well as the role our specialty may play in helping to shape the future of AI within the larger spectrum of medicine.

GEO-CEOS stage 4 validation of the Satellite Image Automatic Mapper lightweight computer program for ESA Earth observation level 2 product generation - Part 1: Theory

ESA defines as Earth Observation (EO) Level 2 information product a single-date multi-spectral (MS) image corrected for atmospheric, adjacency and topographic effects, stacked with its data-derived scene classification map (SCM), whose legend includes quality layers cloud and cloud-shadow. No ESA EO Level 2 product has ever been systematically generated at the ground segment. To fill the information gap from EO big data to ESA EO Level 2 product in compliance with the GEO-CEOS stage 4 validation (Val) guidelines, an off-the-shelf Satellite Image Automatic Mapper (SIAM) lightweight computer program was validated by independent means on an annual 30 m resolution Web-Enabled Landsat Data (WELD) image composite time-series of the conterminous U.S. (CONUS) for the years 2006-2009. The SIAM core is a prior knowledge-based decision tree for MS reflectance space hyperpolyhedralization into static color names. Typically, a vocabulary of MS color names in a MS data (hyper)cube and a dictionary of land cover (LC) class names in the scene-domain do not coincide and must be harmonized (reconciled). The present Part 1-Theory provides the multidisciplinary background of a priori color naming. The subsequent Part 2-Validation accomplishes a GEO-CEOS stage 4 Val of the test SIAM-WELD annual map time-series in comparison with a reference 30 m resolution 16-class USGS National Land Cover Data 2006 map, based on an original protocol for wall-to-wall thematic map quality assessment without sampling, where the test and reference maps feature the same spatial resolution and spatial extent, but whose legends differ and must be harmonized.