Generating reference models for structurally complex data. Application to the stabilometry medical domain

A Novel Artificial-Intelligence-Based Approach for Classification of Parkinson's Disease Using Complex and Large Vocal Features

Parkinson's disease (PD) affects a large proportion of elderly people. Symptoms include tremors, slow movement, rigid muscles, and trouble speaking. With the aging of the developed world's population, this number is expected to rise. The early detection of PD and avoiding its severe consequences require a precise and efficient system. Our goal is to create an accurate AI model that can identify PD using human voices. We developed a transformer-based method for detecting PD by retrieving dysphonia measures from a subject's voice recording. It is uncommon to use a neural network (NN)-based solution for tabular vocal characteristics, but it has several advantages over a tree-based approach, including compatibility with continuous learning and the network's potential to be linked with an image/voice encoder for a more accurate multi modal solution, shifting SOTA approach from tree-based to a neural network (NN) is crucial for advancing research in multimodal solutions. Our method outperforms the state of the art (SOTA), namely Gradient-Boosted Decision Trees (GBDTs), by at least 1% AUC, and the precision and recall scores are also improved. We additionally offered an XgBoost-based feature-selection method and a fully connected NN layer technique for including continuous dysphonia measures, in addition to the solution network. We also discussed numerous important discoveries relating to our suggested solution and deep learning (DL) and its application to dysphonia measures, such as how a transformer-based network is more resilient to increased depth compared to a simple MLP network. The performance of the proposed approach and conventional machine learning techniques such as MLP, SVM, and Random Forest (RF) have also been compared. A detailed performance comparison matrix has been added to this article, along with the proposed solution's space and time complexity.

Applying data mining techniques to medical time series: an empirical case study in electroencephalography and stabilometry

One of the major challenges in the medical domain today is how to exploit the huge amount of data that this field generates. To do this, approaches are required that are capable of discovering knowledge that is useful for decision making in the medical field. Time series are data types that are common in the medical domain and require specialized analysis techniques and tools, especially if the information of interest to specialists is concentrated within particular time series regions, known as events. This research followed the steps specified by the so-called knowledge discovery in databases (KDD) process to discover knowledge from medical time series derived from stabilometric (396 series) and electroencephalographic (200) patient electronic health records (EHR). The view offered in the paper is based on the experience gathered as part of the VIIP project. Knowledge discovery in medical time series has a number of difficulties and implications that are highlighted by illustrating the application of several techniques that cover the entire KDD process through two case studies. This paper illustrates the application of different knowledge discovery techniques for the purposes of classification within the above domains. The accuracy of this application for the two classes considered in each case is 99.86% and 98.11% for epilepsy diagnosis in the electroencephalography (EEG) domain and 99.4% and 99.1% for early-age sports talent classification in the stabilometry domain. The KDD techniques achieve better results than other traditional neural network-based classification techniques.

Clinical Complexity in Medicine: A Measurement Model of Task and Patient Complexity

BACKGROUND

Complexity in medicine needs to be reduced to simple components in a way that is comprehensible to researchers and clinicians. Few studies in the current literature propose a measurement model that addresses both task and patient complexity in medicine.

OBJECTIVE

The objective of this paper is to develop an integrated approach to understand and measure clinical complexity by incorporating both task and patient complexity components focusing on the infectious disease domain. The measurement model was adapted and modified for the healthcare domain.

METHODS

Three clinical infectious disease teams were observed, audio-recorded and transcribed. Each team included an infectious diseases expert, one infectious diseases fellow, one physician assistant and one pharmacy resident fellow. The transcripts were parsed and the authors independently coded complexity attributes. This baseline measurement model of clinical complexity was modified in an initial set of coding processes and further validated in a consensus-based iterative process that included several meetings and email discussions by three clinical experts from diverse backgrounds from the Department of Biomedical Informatics at the University of Utah. Inter-rater reliability was calculated using Cohen's kappa.

RESULTS

The proposed clinical complexity model consists of two separate components. The first is a clinical task complexity model with 13 clinical complexity-contributing factors and 7 dimensions. The second is the patient complexity model with 11 complexity-contributing factors and 5 dimensions.

CONCLUSION

The measurement model for complexity encompassing both task and patient complexity will be a valuable resource for future researchers and industry to measure and understand complexity in healthcare.

Connected health and integrated care: Toward new models for chronic disease management

The increasingly aging population in Europe and worldwide brings up the need for the restructuring of healthcare. Technological advancements in electronic health can be a driving force for new health management models, especially in chronic care. In a patient-centered e-health management model, communication and coordination between patient, healthcare professionals in primary care and hospitals can be facilitated, and medical decisions can be made timely and easily communicated. Bringing the right information to the right person at the right time is what connected health aims at, and this may set the basis for the investigation and deployment of the integrated care models. In this framework, an overview of the main technological axes and challenges around connected health technologies in chronic disease management are presented and discussed. A central concept is personal health system for the patient/citizen and three main application areas are identified. The connected health ecosystem is making progress, already shows benefits in (a) new biosensors, (b) data management, (c) data analytics, integration and feedback. Examples are illustrated in each case, while open issues and challenges for further research and development are pinpointed.

Big data - smart health strategies. Findings from the yearbook 2014 special theme

OBJECTIVES

To select best papers published in 2013 in the field of big data and smart health strategies, and summarize outstanding research efforts.

METHODS

A systematic search was performed using two major bibliographic databases for relevant journal papers. The references obtained were reviewed in a two-stage process, starting with a blinded review performed by the two section editors, and followed by a peer review process operated by external reviewers recognized as experts in the field.

RESULTS

The complete review process selected four best papers, illustrating various aspects of the special theme, among them: (a) using large volumes of unstructured data and, specifically, clinical notes from Electronic Health Records (EHRs) for pharmacovigilance; (b) knowledge discovery via querying large volumes of complex (both structured and unstructured) biological data using big data technologies and relevant tools; (c) methodologies for applying cloud computing and big data technologies in the field of genomics, and (d) system architectures enabling high-performance access to and processing of large datasets extracted from EHRs.

CONCLUSIONS

The potential of big data in biomedicine has been pinpointed in various viewpoint papers and editorials. The review of current scientific literature illustrated a variety of interesting methods and applications in the field, but still the promises exceed the current outcomes. As we are getting closer towards a solid foundation with respect to common understanding of relevant concepts and technical aspects, and the use of standardized technologies and tools, we can anticipate to reach the potential that big data offer for personalized medicine and smart health strategies in the near future.