AI4Health 2018 Abstracts


Short Papers
Paper Nr: 7
Title:

Deep Convolution Neural Network for Extreme Multi-label Text Classification

Authors:

Francesco Gargiulo, Stefano Silvestri and Mario Ciampi

Abstract: In this paper we present an analysis on the usage of Deep Neural Networks for extreme multi-label and multiclass text classification. We will consider two network models: the first one is formed by a word embeddings (WEs) stage followed by two dense layers, hereinafter Dense, and a second model with a convolution stage between the WEs and the dense layers, hereinafter CNN-Dense. We will take into account classification problems characterized by different number of labels, ranging from an order of 10 to an order of 30; 000, showing the different performances of the neural networks varying the total label number and the average number of labels for sample, exploiting the hierarchical structure of the label space of the dataset used for experimental assessment. It is worth noting that multi-label classification is an harder problem if compared to multi-class, due to the variable number of labels associated to each sample. We will even investigate on the behaviour of the neural networks as function of the training hyperparameters, analysing the link between them and the dataset complexity. All the result will be evaluated using the PubMed scientific articles collection as test case.

Paper Nr: 9
Title:

A Machine Learning Approach for Carotid Diseases using Heart Rate Variability Features

Authors:

Laura Verde and Giuseppe De Pietro

Abstract: In the last few years the incidence of carotid diseases has been increasing rapidly. Atherosclerosis constitutes a major cause of morbidities and mortalities worldwide. The early detection of these diseases is considered necessary to avoid tragic consequences and automatic systems and algorithms can be a valid support for their diagnosis. The main objective of this study is to investigate and compare the performances of different machine learning techniques capable of detecting the presence of a carotid disease by analysing the Heart Rate Variability (HRV) parameters of opportune electrocardiographic signals selected from an appropriate database available online on the Physionet website. All the analyses are evaluated in terms of accuracy, precision, recall and F-measure.

Paper Nr: 12
Title:

Determining Cardiopulmonary Resuscitation Parameters with Differential Evolution Optimization of Sinusoidal Curves

Authors:

Christian Lins, Andreas Klausen, Sebastian Fudickar, Sandra Hellmers, Myriam Lipprandt, Rainer Röhrig and Andreas Hein

Abstract: In this paper, we present a robust sinusoidal curve fitting method based on the Differential Evolution (DE) algorithm for determining cardiopulmonary resuscitation (CPR) parameters – naming chest compression frequency and depth – from skeletal motion data. Our implementation uses skeletal data from the RGB-D (RGB + Depth) Kinect v2 sensor and works without putting non-sensor related constraints such as specific view angles or distance to the system. Our approach is intended to be part of a robust and easy-to-use feedback system for CPR training, allowing its unsupervised training. We compare the sensitivity of our DE implementation with data recorded by a Laerdal Resusci Anne mannequin. Results show that the frequency of the DE-based CPR is recognized with a variance of 4:4 bpm (4.1%) in comparison to the reference of the Resusci Anne mannequin.

Paper Nr: 13
Title:

SENSdesc: Connect Sensor Queries and Context

Authors:

Dörthe Arndt, Pieter Bonte, Alexander Dejonghe, Ruben Verborgh, Filip De Turck and Femke Ongenae

Abstract: Modern developments confront us with an ever increasing amount of streaming data: different sensors in environments like hospitals or factories communicate their measurements to other applications. Having this data at disposal faces us with a new challenge: the data needs to be integrated to existing frameworks. As the availability of sensors can rapidly change, these need to be flexible enough to easily incorporate new systems without having to be explicitly configured. Semantic Web applications offer a solution for that enabling computers to ‘understand’ data. But for them the pure amount of data and different possible queries which can be performed on it can form an obstacle. This paper tackles this problem: we present a formalism to describe stream queries in the ontology context in which they might become relevant. These descriptions enable us to automatically decide based on the actual setting and the problem to be solved which and how sensors should be monitored further. This helps us to limit the streaming data taken into account for reasoning tasks and make stream reasoning more performant. We illustrate our approach on a health-care use case where different sensors are used to measure data on patients and their surrounding in a hospital.

Paper Nr: 17
Title:

Applying Artificial Intelligence in Healthcare Social Networks to Identity Critical Issues in Patients’ Posts

Authors:

Giacomo Fiumara, Antonio Celesti, Antonino Galletta, Lorenzo Carnevale and Massimo Villari

Abstract: Nowadays, the possibility of using social media in the healthcare field is attracting the attention of clinical professionals and of the whole healthcare industry. In this panorama, many Healthcare Social Networking (HSN) platforms are emerging with the purpose to enhance patient care and education. However, they also present potential risks for patients due to the possible distribution of poor-quality or wrong information. On one hand doctors want to promote the exchange of information among patients about a specific disease, but on the other hand they do not have the time to read patients’ posts and moderate them when required. In this paper, we propose an Artificial Intelligence (AI) approach based on a combination of stemming, lemmatization and Machine Learnign (ML) algorithms that allows to automatically analyse the patients’ posts of a HSN platform and identify possible critical issues so as to enable doctors to intervene when required. In particular, after a discussion of advantages and disadvantages of using a HSN platform, we discuss in detail an architecure that allows to analyse big data consisting of patients’ posts. In the end, real case studies are discussed highlighting future challenges.

Paper Nr: 18
Title:

Evaluation of the Adaptive Statistical Iterative Reconstruction Algorithm in Chest CT (Computed Tomography) - A Preliminary Study toward Its Employment in Low Dose Applications, Also in Conjunction with CAD (Computer Aided Detection)

Authors:

Patrizio Barca, Federica Palmas, Maria Evelina Fantacci and Davide Caramella

Abstract: Lung cancer is one of the leading cause of cancer death worldwide. Computed Tomography (CT) is the best imaging modality for the detection of small pulmonary nodules and for this reason its employment as a screening tool has been widely studied. However, radiation dose delivered in a chest CT examination must be considered, especially when potentially healthy people are examined in screening programs. In this context, iterative reconstruction (IR) algorithms have shown the potential to reduce image noise and radiation dose and computer aided detection (CAD) systems can be employed for supporting radiologists. Thus, the combined use of IR algorithms and CAD systems can be of practical interest. In this preliminary work we studied the potential improvements in the quality of phantom and clinical chest images reconstructed trough the Adaptive Statistical Iterative Reconstruction (ASIR, GE Healthcare, Waukesha, WI, USA) algorithm, in order to evaluate a possible employment of this algorithm in low dose chest CT imaging with CAD analysis. We analysed both clinical and phantom CT images. Noise, noise power spectrum (NPS) and modulation transfer function (MTF) were estimated for different inserts in the phantom images. Image contrast and contrast-to-noise ratio (CNR) of different nodules contained in clinical chest images were evaluated. Noise decreases non-linearly when increasing the ASIR blending level of reconstruction. ASIR modified the NPS. The MTF for ASIR-reconstructed images depended on tube load, contrast and blending level. Both image contrast and CNR increased with the ASIR blending level.

Paper Nr: 21
Title:

Investigating Random Forest Classification on Publicly Available Tuberculosis Data to Uncover Robust Transcriptional Biomarkers

Authors:

Carly A. Bobak, Alexander J. Titus and Jane E. Hill

Abstract: There has been increasing concern amongst the scientific community of a reproducibility crisis, particularly in the field of bioinformatics. Often, published research results do not correlate with clinical success. One theory explaining this phenomenon is that findings from homogeneous cohort studies are not generalizable to an inherently heterogeneous population. In this work, we integrate data from 4 distinct tuberculosis (TB) cohorts, for a total of 1164 samples, to find common differentially regulated genes which may be used to diagnose active TB from latent TB, treated TB, other diseases, and healthy controls. We selected 25 genes using random forest to get an AUC of 0.89 in our training data, and 0.86 in our test data. A total of 18 out of 25 genes had been previously associated with TB in independent studies, suggesting that integrating data may be an important tool for increasing micro-array research reproducibility.

Paper Nr: 24
Title:

Design and Implementation of a Data Acquisition System for R Peak Detection in Electrocardiograms

Authors:

Gabriela Idali Ibarra Fierro, Ricardo Rodriguez-Jorge, J. Mizera-Pietraszko and Edgar A. Martinez-Garcia

Abstract: This paper presents a data acquisition system for the R peak detection in electrocardiograms. R wave is one of the most important sections of the QRS complex, which has an essential role in diagnosis of heart rhythm irregularities. This paper employs Hilbert transform and adaptive threshold technique for the detection of R-peak. The performance of the system was tested using standard ECG waveform records from the MIT-BIH arrhythmia database. In addition, tests were carried out using the sensors Single Lead Heart Rate Monitor for real-time data processing.

Paper Nr: 26
Title:

Clinical Ontology Mapping - Toward Automatic Care Plan Recommendation

Authors:

Khai Nguyen, Kaisei Reio and Ryutaro Ichise

Abstract: In this paper, we share a sketch of an automatic care plan recommendation system in Japan. After that, we describe our proposed method and experience in the first step: clinical ontology mapping. We discuss the difficulties, method, preliminary results of a case study, which is to find corresponding mappings between two ontologies, the Minimum Data Set 3.0 (MDS) and the International Classification of Functioning, Disability and Health (ICF)

Paper Nr: 28
Title:

Towards a Decision Support System for Disorders of the Cardiovascular System - Diagnosing and Evaluation of the Treatment Efficiency

Authors:

Anton Dolganov and Vladimir Kublanov

Abstract: The study describes a preliminary stage of the decision support system development for cardiovascular system disorders. As the clinical model of the disorders, the arterial hypertension was used. The study consisted of two steps: diagnosing of the arterial hypertension and an evaluation of the treatment efficiency during the neuro-electrostimulation application. For the diagnosing part, a clinical study was conducted involving heart rate variability signals recording while performing tilt-test functional load. Performance of different machine learning techniques and feature selection strategies in task of binary classification (healthy volunteers and patients suffering from arterial hypertension) were compared. The genetic programming feature selection and quadratic discriminant analysis classifier reached the highest classification accuracy. Best feature combinations were used to evaluate a treatment efficiency. The results indicate the potential of the proposed decision support system.

Posters
Paper Nr: 5
Title:

An Evolutionary Approach for Estimating the Blood Glucose by Exploiting Interstitial Glucose Measurements

Authors:

Ivanoe De Falco, Antonio Della Cioppa, Tomas Koutny, Michal Krcma, Umberto Scafuri and Ernesto Tarantino

Abstract: The diabetes is correlated to a malfunction of the pancreas that produces very little or no insulin. A way to improve the quality of life of people with diabetes is to implement an artificial pancreas able to inject an insulin bolus when necessary. The aim of this paper is to devise a possibly step in constructing the fundamental element of such an artificial pancreas - estimation of the blood glucose (BG) through interstitial glucose (IG) measurements. In particular, a new methodology is presented to derive a mathematical relationship between BG and IG by exploiting the ability of the evolutionary techniques in solving this regression task. An automatic procedure is used to estimate the missing BG values within this database. To validate the discovered model a comparison with other models is carried out during the experimental phase.

Paper Nr: 6
Title:

Preliminary Steps towards Efficient Classification in Large Medical Datasets: Structure Optimization for Deep Learning Networks through Parallelized Differential Evolution

Authors:

Ivanoe De Falco, Giuseppe De Pietro, Antonio Della Cioppa, Giovanna Sannino, Umberto Scafuri and Ernesto Tarantino

Abstract: Deep Neural Networks are being more and more widely used to perform several tasks over highly-sized datasets, one of them being classification. Finding good configurations for Deep Neural Network structures is a very important problem in general, and particularly in the medical domain. Currently, either trial-and-error methodologies or sampling-based ones are considered. This paper describes some preliminary steps towards effectively facing this task. The first step consists in the use of Differential Evolution, a kind of an Evolutionary Algorithm. The second lies in using a parallelized version in order to reduce the turnaround time. The preliminary results obtained here show that this approach can be useful in easily obtaining structures that allow increases in the network accuracy with respect to those provided by humans.

Paper Nr: 8
Title:

Optimizing Social Interaction - A Computational Approach to Support Patient Engagement

Authors:

Italo Zoppis, Riccardo Dondi, Eugenio Santoro, Gianluca Castelnuovo, Francesco Sicurello and Giancarlo Mauri

Abstract: Social media can directly support disease management by creating online spaces where patients can interact with clinicians, and share experiences with other patients. Nevertheless, much more work remains to be carried out for providing and sharing an optimized information content. In this paper we formulate, from a theoretical perspective, an optimization problem aimed to encourage the creation of a sub-network of patients which, being recently diagnosed, wish to deepen their knowledge about their pathologies with some other patients, whose clinical profile turn to be similar, and have already been followed up within specific, even alternative, care centers. We will focus on the hardness of the proposed problem and provide a Genetic Algorithm (GA-based) approach to seek faster approximated solutions.

Paper Nr: 22
Title:

Supporting Palliative Care Services - An IS System to Monitor the Patients and Manage the Mobile Support Team

Authors:

Arsénio Reis, Eliza Bento da Guia, Vitor Rodrigues and João Barroso

Abstract: The health centers group (ACES) “Douro Sul II” is creating a Community Support Team for Palliative Care (ECSCP) in order to provide palliative care services for the ACES’s population of 73,713 registered users. The team comes as a follow-up from a strategic plan, recently issued by the health ministry, in order to serve the patients in their own homes, providing the necessary support to them, to their families, and their caregivers. This approach has several benefits for the patients and their families, as well as for the healthcare system itself. To further promote the effectiveness of the ECSCP team, it was planned to develop an information system (IS), comprising several application modules, with the main objective to monitor the patients in their homes and deliver information to support the planning and execution of the ECSCP team activities. The system is based on an electronic services platform and several mobile and web applications, to be used by the patients, team’s staff and coordination. This way, we expect to overcome the geographic issues of the ACES territory, as well as the team’s human resources constraints, while remotely monitoring the patients and providing the necessary support, if and when needed, contributing to maintaining the conditions for the patients to live with dignity and quality in the comfort of their own homes.

Paper Nr: 23
Title:

eHealth Context Inference - A Review of Open Source Frameworks Initiatives

Authors:

Arsénio Reis, Dennis Paulino, Paulo Martins, Hugo Paredes and João Barroso

Abstract: The collection of health and fitness longitudinal data can be used to model disease progression and shape new algorithms to diagnose and predict health hazards. Continuously tracking vital signs, in particular heart rate and skin temperature, can be very informative by using models and algorithms to predict and notify the user about when he might be falling ill. With the current wearable devices and the proper algorithms, the individual can be permanently monitored, which might be much more interesting than a one-off reading comparison with the population average, made by a doctor. It would be possible to intervene earlier and to prevent somebody from becoming seriously ill. From a broader perspective, the knowledge about a user’s health can be considered as an element of that user’s context and be used by context aware applications to provide higher value to the user. After the trivialization of the data acquisition sensors, wearable devices, and raw data, the next logical step is the development of contained software components that can infer and produce knowledge from the raw data. These components and the knowledge they produce can be used by all sorts of applications in order to further customize their usage by a specific user. Customization and context awareness, in regard to health, is a wide field for research and there are a multitude of proposals for models and algorithms. In this review work we searched for software components (frameworks, software libraries, etc.), freely available and that can be used as building blocks for other computer systems by software developers.