Deep Learning for Describing Breast Ultrasound Images with BI-RADS Terms.

Breast cancer is the most common cancer in women. Ultrasound is one of the most used techniques for diagnosis, but an expert in the field is necessary to interpret the test. Computer-aided diagnosis (CAD) systems aim to help physicians during this process. Experts use the Breast Imaging-Reporting and Data System (BI-RADS) to describe tumors according to several features (shape, margin, orientation...) and estimate their malignancy, with a common language. To aid in tumor diagnosis with BI-RADS explanations, this paper presents a deep neural network for tumor detection, description, and classification. An expert radiologist described with BI-RADS terms 749 nodules taken from public datasets. The YOLO detection algorithm is used to obtain Regions of Interest (ROIs), and then a model, based on a multi-class classification architecture, receives as input each ROI and outputs the BI-RADS descriptors, the BI-RADS classification (with 6 categories), and a Boolean classification of malignancy. Six hundred of the nodules were used for 10-fold cross-validation (CV) and 149 for testing. The accuracy of this model was compared with state-of-the-art CNNs for the same task. This model outperforms plain classifiers in the agreement with the expert (Cohen's kappa), with a mean over the descriptors of 0.58 in CV and 0.64 in testing, while the second best model yielded kappas of 0.55 and 0.59, respectively. Adding YOLO to the model significantly enhances the performance (0.16 in CV and 0.09 in testing). More importantly, training the model with BI-RADS descriptors enables the explainability of the Boolean malignancy classification without reducing accuracy.

Performance of students with different accessibility needs and preferences in "Design for All" MOOCs.

Recent research has shown that Massive Open Online Courses (MOOCs) create barriers for students with disabilities. Not taking into account their needs in the design, production or delivery of MOOCs may be one of the main causes behind this. It leads to poor compliance with suitable learning designs and web accessibility standards, as well as a lack of knowledge about the students' needs. The objective of our research is to analyze the learning performance of the students in MOOCs on topics related to Design for All, offered in an Open edX-based platform. Accessibility support was conceived from the outset, including compliance of both the platform and the learning resources with the WCAG 2.1 accessibility standard, and with a subset of the principles of Universal Design for Learning. Additionally, students were consulted on their accessibility needs and preferences, following publicly available modeling schemes and previous research. From a sample of 765 students, who completed at least one of the graded assessment activities of the course, a multilevel multiple logistic regression model was fitted. Based on that model, the results indicate that: a) users of screen readers and users of captions show a statistically significant positive association with a good performance when compared to students with no preferences, with an odds ratio of, respectively, OR = 13.482 and OR = 13.701; b) students who have low vision or very low vision show a significant negative association with a good performance when compared to users of screen readers and to users of captions, with OR = 26.817 and OR = 27.254, respectively.

Quality analysis of a breast thermal images database.

The study and early detection of breast cancer are key for its treatment. We carry out an exhaustive analysis of the most used database for mastology research with infrared images, analyzing the anomalies according to five quality dimensions: completeness, correctness, concordance, plausibility, and currency. We established control queries that looked for these anomalies and that can be used to ensure the quality of the database. Finally, we briefly review the more than 40 papers that use this database and that do not mention any of these anomalies. When analyzing the database, we found 365 anomalies related to personal and clinical data, and thermal images. The errors found in our research may lead to a modification of the results and conclusions made in the articles found in the literature, serve as a basis for improvements in the quality of the database, and help future researchers to work with it.

Cost-effectiveness analysis with unordered decisions.

INTRODUCTION: Cost-effectiveness analysis (CEA) is used increasingly in medicine to determine whether the health benefit of an intervention is worth the economic cost. Decision trees, the standard decision modeling technique for non-temporal domains, can only perform CEAs for very small problems. Influence diagrams can model much larger problems, but only when the decisions are totally ordered. OBJECTIVE: To develop a CEA method for problems with unordered or partially ordered decisions, such as finding the optimal sequence of tests for diagnosing a disease. METHODS: We explain how to model those problems using decision analysis networks (DANs), a new type of probabilistic graphical model, somewhat similar to Bayesian networks and influence diagrams. We present an algorithm for evaluating DANs with two criteria, cost and effectiveness, and perform some experiments to study its computational efficiency. We illustrate the representation framework and the algorithm using a hypothetical example involving two therapies and several tests and then present a DAN for a real-world problem, the mediastinal staging of non-small cell lung cancer. RESULTS: The evaluation of a DAN with two criteria, cost and effectiveness, returns a set of intervals for the willingness to pay, separated by incremental cost-effectiveness ratios (ICERs). The cost, the effectiveness, and the optimal intervention are specific for each interval, i.e., they depend on the willingness to pay. CONCLUSION: Problems involving several unordered decisions can be modeled with DANs and evaluated in a reasonable amount of time. OpenMarkov, an open-source software tool developed by our research group, can be used to build the models and evaluate them using a graphical user interface.

Multi-input convolutional neural network for breast cancer detection using thermal images and clinical data.

BACKGROUND AND OBJECTIVE: Breast cancer is the most common cancer in women. While mammography is the most widely used screening technique for the early detection of this disease, it has several disadvantages such as radiation exposure or high economic cost. Recently, multiple authors studied the ability of machine learning algorithms for early diagnosis of breast cancer using thermal images, showing that thermography can be considered as a complementary test to mammography, or even as a primary test under certain circumstances. Moreover, although some personal and clinical data are considered risk factors of breast cancer, none of these works considered that information jointly with thermal images. METHODS: We propose a novel approach for early detection of breast cancer combining thermal images of different views with personal and clinical data, building a multi-input classification model which exploits the benefits of convolutional neural networks for image analysis. First, we searched for structures using only thermal images. Next, we added the clinical data as a new branch of each of these structures, aiming to improve its performance. RESULTS: We applied our method to the most widely used public database of breast thermal images, the Database for Mastology Research with Infrared Image. The best model achieves a 97% accuracy and an area under the ROC curve of 0.99, with a specificity of 100% and a sensitivity of 83%. CONCLUSIONS: After studying the impact of thermal images and personal and clinical data on multi-input convolutional neural networks for breast cancer diagnosis, we conclude that: (1) adding the lateral views to the front view improves the performance of the classification model, and (2) including personal and clinical data helps the model to recognize sick patients.

Early intervention services, patterns of prescription and rates of discontinuation of antipsychotic treatment in first-episode psychosis.

AIMS: Non-compliance is still an important problem in psychotic patients. Although antipsychotic (AP) treatment leads to a decrease in psychotic relapses, there are no clear recommendations about how long treatment should be maintained after first-episode psychosis (FEP) and no indication of the rates and causes of treatment withdrawal in this group. METHODS: We evaluated a large sample of patients with FEP for 2 years to compare the time to all-cause treatment discontinuation of AP drugs and the time to the first relapse. We collected the sociodemographic and psychopathological characteristics of the sample. The number of relapses was also recorded. RESULTS: A total of 310 FEP patients were assessed across seven early intervention teams (mean age = 30.2 years; SD = 11.2). The most prevalent diagnosis at baseline was psychotic disorder not otherwise specified (36.1%), and the most commonly used APs were risperidone (26.5%) and olanzapine (18.7%). A lack of efficacy was the most frequent reason for the withdrawal of the first AP prescribed, followed by non-compliance. There were no differences in the relapse rates between different APs. Patients treated with long-acting injectable (LAI) APs presented less disengagement from services than patients treated with oral APs. CONCLUSIONS: Although there were no differences between the different APs in terms of relapse rates, LAIs had higher retention rates than oral APs in early intervention services. Compliance is still an important issue in Psychiatry, so clinicians should use different strategies to encourage it, such as the use of LAI treatments.

Differences in theory of mind between early and chronic stages in schizophrenia.

The evolution of social cognition throughout the course of schizophrenia is unclear not being possible to state whether it remains stable from early stages to chronicity, or it changes as the disease develops. For this purpose, 90 patients with schizophrenia and 139 healthy controls have been compared establishing 4 different groups paired by age and gender: first episode of psychosis patients (FEP), young healthy controls (YHC), chronic patients with schizophrenia (CS) and adult healthy controls (AHC). Performance in Theory of Mind (ToM) has been assessed using The Hinting Task and The Reading the Mind in the Eyes Test (RMET). In the Hinting Task, when comparing patients with their respective control group, differences found between CS patients and their corresponding controls (p < .001) are much bigger (almost twice) than differences between FEP patients and young controls (p = .001). In fact, young and adult healthy controls did not significantly differ in their scores, while the CS group showed significant worse performance than the FEP group. In the Reading the Mind in the Eyes test (RMET), patients globally performed worse than controls (p < .001). However, the Cohort × Diagnosis interaction was not significant (p = .27). In this task, there were no differences between CS and FEP scores. In conclusion, data suggest poor performance in all phases of the disease with a probable worsening related to chronicity especially in the aspects of social cognition measured by the Hinting Task.

Computerized cognitive remediation therapy, REHACOM, in first episode of schizophrenia: A randomized controlled trial.

Patients with schizophrenia show cognitive impairments that have been linked to poor social functioning. Computerized cognitive remediation therapy has shown to be effective in improving both cognition and functioning in chronic schizophrenia, but relatively little is known about how these approaches improve cognition and functioning when applied to early stages of psychosis. Eighty-six participants with a first episode of psychosis, undergoing a specific program for early stages of schizophrenia, undertook either the REHACOM computerized cognitive remediation intervention (n = 36), or an active control condition (n = 50) consisting in 24 one-hour sessions addressed twice a week. Clinical features, cognition and functioning were assessed at baseline, post-treatment and six months after finishing the intervention. A significant progressive improvement in neurocognition and functioning was globally shown with no differences observed between the experimental and control group at post training or follow up. All cognitive domains but Social Cognition improved between 0.5 and 1 S.D. through the study period. The computerized cognitive remediation therapy REHACOM has not proved to be effective on improving cognition nor functioning compared to controls in outpatients with a first episode of schizophrenia.

Evaluation of Markov Models with Discontinuities.

Background. Several methods, such as the half-cycle correction and the life-table method, were developed to attenuate the error introduced in Markov models by the discretization of time. Elbasha and Chhatwal have proposed alternative "corrections" based on numerical integration techniques. They present an example whose results suggest that the trapezoidal rule, which is equivalent to the half-cycle correction, is not as accurate as Simpson's 1/3 and 3/8 rules. However, they did not take into consideration the impact of discontinuities. Objective. To propose a method for evaluating Markov models with discontinuities. Design. Applying the trapezoidal rule, we derive a method that consists of adjusting the model by setting the cost at each point of discontinuity to the mean of the left and right limits of the cost function. We then take from the literature a model with a cycle length of 1 year and a discontinuity on the cost function and compare our method with other "corrections" using as the gold standard an equivalent model with a cycle length of 1 day. Results. As expected, for this model, the life-table method is more accurate than assuming that transitions occur at the beginning or the end of cycles. The application of numerical integration techniques without taking into account the discontinuity causes large errors. The model with averaged cost values yields very small errors, especially for the trapezoidal and the 1/3 Simpson rules. Conclusion. In the case of discontinuities, we recommend applying the trapezoidal rule on an averaged model because this method has a mathematical justification, and in our empirical evaluation, it was more accurate than the sophisticated 3/8 Simpson rule.

Cost-effectiveness of pediatric bilateral cochlear implantation in Spain.

OBJECTIVES/HYPOTHESIS: To determine the incremental cost-effectiveness of bilateral versus unilateral cochlear implantation for 1-year-old children suffering from bilateral sensorineural severe to profound hearing loss from the perspective of the Spanish public health system. STUDY DESIGN: Cost-utility analysis. METHODS: We conducted a general-population survey to estimate the quality-of-life increase contributed by the second implant. We built a Markov influence diagram and evaluated it for a life-long time horizon with a 3% discount rate in the base case. RESULTS: The incremental cost-effectiveness ratio of simultaneous bilateral implantation with respect to unilateral implantation for 1-year-old children with severe to profound deafness is €10,323 per quality-adjusted life year (QALY). For sequential bilateral implantation, it rises to €11,733/QALY. Both options are cost-effective for the Spanish health system, whose willingness to pay is estimated at around €30,000/QALY. The probabilistic sensitivity analysis shows that the probability of bilateral implantation being cost-effective reaches 100% for that cost-effectiveness threshold. CONCLUSIONS: Bilateral implantation is clearly cost-effective for the population considered. If possible, it should be done simultaneously (i.e., in one surgical operation), because it is as safe and effective as sequential implantation, and saves costs for the system and for users and their families. Sequential implantation is also cost-effective for children who have received the first implant recently, but it is difficult to determine when it ceases to be so because of the lack of detailed data. These results are specific for Spain, but the model can easily be adapted to other countries. LEVEL OF EVIDENCE: 2C. Laryngoscope, 127:2866-2872, 2017.

Markov Influence Diagrams.

Markov influence diagrams (MIDs) are a new type of probabilistic graphical model that extends influence diagrams in the same way that Markov decision trees extend decision trees. They have been designed to build state-transition models, mainly in medicine, and perform cost-effectiveness analyses. Using a causal graph that may contain several variables per cycle, MIDs can model various patient characteristics without multiplying the number of states; in particular, they can represent the history of the patient without using tunnel states. OpenMarkov, an open-source tool, allows the decision analyst to build and evaluate MIDs-including cost-effectiveness analysis and several types of deterministic and probabilistic sensitivity analysis-with a graphical user interface, without writing any code. This way, MIDs can be used to easily build and evaluate complex models whose implementation as spreadsheets or decision trees would be cumbersome or unfeasible in practice. Furthermore, many problems that previously required discrete event simulation can be solved with MIDs; i.e., within the paradigm of state-transition models, in which many health economists feel more comfortable.

Estudio descriptivo de los pacientes ingresados en el Hospital Psiquiátrico Penitenciario de Alicante / Descriptive assessment of the patients admitted to the Alicante Penitentiary Psychiatric Hospital

Desde la entrada en vigor del nuevo Código Penal en el año 1995, se ha producido un cambio significativo en lo que se refiere al tratamiento del enfermo mental en el ámbito penal. El presente trabajo es un estudio retro y prospectivo en el que valoramos aspectos sociodemográficos, clínicos y legales de los pacientes ingresados en el Hospital Psiquiátrico Penitenciario de Alicante desde junio de 1996 hasta abril de 2000. Las edades están comprendidas entre los 20 y los 85 años. Los diagnósticos más frecuentes (siguiendo los criterios ICD-10) son: esquizofrenia, retraso mental, trastorno por ideas delirantes persistentes y trastorno orgánico cerebral. La mayoría de los pacientes son varones (AU)