Government & Nonprofit

ALOE platform: an overview of a service-oriented architecture for research in breast cancer diagnosis supported by e-infrastructures

Description
This article presents an overview of the ALOE platform. ALOE provides a service-oriented architecture aimed at the research in the early detection of breast cancer diagnosis. The development of the ALOE platform is carried out by collaboration among
Published
of 16
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
     Int. J. Image Mining, Vol. 1, Nos. 2/3, 2015 245  Copyright © 2015 Inderscience Enterprises Ltd. ALOE platform: an overview of a service-oriented architecture for research in breast cancer diagnosis supported by e-infrastructures José M. Franco-Valiente* and César Suárez-Ortega Software Architecture Unit, Extremadura Research Center for Advanced Technologies Trujillo, Spain Fax: +34-927-323-237 Email: josemiguel.franco@ciemat.es Email: cesar.suarez@externos.ciemat.es *Corresponding author Abstract:  This article presents an overview of the ALOE platform. ALOE  provides a service-oriented architecture aimed at the research in the early detection of breast cancer diagnosis. The development of the ALOE platform is carried out by collaboration among CETA-CIEMAT, INEGI, FMUP-HSJ and UA. ALOE supports two research lines in breast cancer diagnosis: the development of well performing computer aided diagnosis (CAD) systems and the development of new tools-based on e-learning techniques to improve radiologists training. All ALOE modules are designed to work as a whole system but can be used individually in other systems and expose RESTful interfaces to be exploited by third party systems. ALOE components make use of e-Infrastructure resources to accomplish their tasks. The final objective of this work is to provide a reference platform for researchers, specialists, and students in breast cancer diagnosis. Keywords:  e-learning; e-infrastructure; breast-cancer; digital repositories; CAD systems. Reference  to this paper should be made as follows: Franco-Valiente, J.M. and Suárez-Ortega, C. (2015) ‘ALOE platform: an overview of a service-oriented architecture for research in breast cancer diagnosis supported by e-infrastructures’,  Int. J.    Image Mining  , Vol. 1, Nos. 2/3, pp.245–260. Biographical notes:  José M. Franco-Valiente is the Head of the Software Development Department at CETA-CIEMAT Supercomputing Centre (Trujillo, Spain). He has a MSc in Grid Computing and Parallelism at University of Extremadura (Spain), his main activity is focused on the exploitation of e-infrastructures to boost research in health, energy and environment. César Suárez-Ortega is a researcher at the Software Development Department at CETA-CIEMAT Supercomputing Centre (Trujillo, Spain). He has an MSc in Grid Computing and Parallelism at University of Extremadura (Spain), his main activity is focused on the design and development of scientific software applications that rely on distributed computer infrastructures.   246  J.M. Franco-Valiente andC. Suárez-Ortega   This paper is a revised and expanded version of a paper entitled ‘ALOE platform: an overview of a service-oriented architecture for research in breast cancer diagnosis supported by e-infrastructures’ presented at 8th Iberian Grid Infrastructure Conference (IBERGRID 2014), Porto (Portugal), 8–10 September 2014. 1 Introduction According to the World Health Organization, breast cancer is the second most common cancer worldwide, and first in women (Matheus and Schiabel, 2011). This disease caused more than half a million deaths in 2010 (Matheus and Schiabel, 2011), and in the European Union it is responsible of one every six deaths from cancer in women (Moreira et al., 2012). The detection of this kind of cancer in its early stages is a very effective method to reduce mortality (Lee, 2002) and mammography reading is the most common technique used by specialists (Lee, 2002) for early detection. Visual inspection of anomalies is a repetitive and fatiguing process which leads to a diagnosis error rate between 10% and 30% (Kerlikowske et al., 2000; Kolb et al., 2002). These errors are mainly false-negative results (overlooked malign lesions) and false-positive results (biopsies performed on  benign lesions). There are two main practises in order to improve the cancer detection rate: double reading of mammographies and single-reading of mammographies supported by computer-aided detection or computer-aided diagnosis systems (CADe and CADx respectively) (Gromet, 2008). In the first one, two radiologists check the same study to determine a diagnosis. In the second practise, a computer-aided diagnosis system (CAD) system assists the human expert to identify abnormalities and diagnose them. Though it is very effective (Brown et al., 1996), the double-reading method is expensive in time and economical cost because it needs two radiologists for each study. This problem is aggravated by the fact that there is a lack of specialists in the area of breast cancer diagnosis (Penhoet et al., 2004). Consequently, CAD systems have increased their  popularity in recent years (Kolb et al., 2002). Furthermore, both methods increase the sensitivity (rate of true-positive results) and the cancer detection rate (Gromet, 2008), but also increase the recall rate (positives that have to be confirmed with a biopsy) and the PPV 1  (rate of recalls that are true-positives (Taylor et al., 2008; Ramos-Pollan et al., 2011). Even having similar problems, both methods have relevant differences. In the one hand, CAD systems does not need extra radiologists to be performed, but in the other hand this method have bigger PPV 1  and recall rates than double reading (Gromet, 2008; Taylor et al., 2008; Kolb et al., 2002;  Noble et al., 2009). Since 2008, the Centre of Extremadura in Advanced Technologies (CETA-CIEMAT), the Institute of Mechanical Engineering and Industrial Management (INEGI) and the Faculty of Medicine of the University of Porto – Centro Hospitalar São João (FMUP-HSJ) collaborate to improve the diagnosis of breast cancer. In 2013, the Institute of Electronics and Telematics Engineering of Aveiro (IEETA) from the University of Aveiro joined the collaboration.     ALOE platform: an overview of a service-oriented architecture for research 247   During the first years of the collaboration, the research activities focused on the generation of well-performing CAD systems for breast cancer diagnosis. These systems were based on machine learning classifiers (MLCs) and were trained and evaluated by exploiting the computing resources of the GRID infrastructure (Foster and Kesselman, 2004) from IBERGRID/EGI. The best configurations of these classifiers have obtained results greater than 85% AUC (ROC), but there are some configurations under development with results that can be considered as ‘excellent’ (being more than 90% AUC) (Ramos-Pollan et al., 2011).  Nowadays, a new research line has been launched by the project team. It is focused on reducing the rate of missed cancers (not detected cancers) by improving the training  process of new specialists in the breast cancer detection. To achieve the objective set, new software applications based on e-learning principles (Suárez-Ortega and Franco-Valiente, 2013) are being developed. This recently opened research line is motivated by the importance acquired by e-learning at the moment in education (Ellet et al. 2003). E-learning is the use of internet related technologies to deliver solutions that enhance knowledge and performance (Rosenberg, 2001). It is natural that the medical community has shown interest in this topic, because of the importance of medical informatics (MI) in the modern medical  practice (Fain, 1997). The main challenge of the introduction of e-learning in medical education lies on the need of specific tools for medicine students in most of the cases (Branstetter et al., 2007). This fact is very important in the case of radiologists training, not only because of the need of new specialists, but also for the high rate of students that abandon their training (Kolb et al., 2002). The early exposure of training in radiology to medical students has shown a high improvement in the impression of this speciality among them, leading to higher rates of individuals considering radiology as a career to study (Branstetter et al., 2007). The typical way to train new radiology students consists of the readings of radiology images (in our case mammograms) and consulting sessions based on real cases. This methodology perfectly fits with some e-learning methods, especially with simulated cases with self-assessment. Consequently, the consortium considered that the know-how acquired during the first years of the collaboration and the direct contact with target users will lead to develop new tools based on e-learning techniques that improve the training of new specialists in  breast cancer diagnosis. Moreover, these tools aim at being also used by experienced radiologists that want to improve their skill in diagnosis, so that they can practice at their own pace with real cases out of their daily work. Both research lines coordinate their activities to develop the ALOE platform, a service-oriented architecture based on advanced computing and storage resources from e-infrastructures. The following sections in this article present the ALOE platform, its main components, current status and future work. 2 Description of the ALOE platform This section provides a general description of the ALOE platform. As mentioned in the  previous section, ALOE implements the first outcomes of the research activities and coordinates them to provide a full-stack solution to researchers, students and specialists in breast cancer diagnosis. The driving idea behind ALOE is assisting the learning   248  J.M. Franco-Valiente andC. Suárez-Ortega   process of breast cancer CADx methods through an integrated intelligent tutoring system, in which it will be possible: 1 accessing to a wide-ranging annotated breast cancer digital repository (BCDR) that includes representative cases of breast cancer 2 exploiting a set of provided tools for exploring and testing CADx methods 3 consulting a state-of-the-art ontology knowledge base on breast cancer 4 executing self-evaluation facilities. Figure 1  ALOE platform architecture (see online version for colours) As shown in Figure 1, ALOE is formed by a group of services, data repositories and user interfaces (both web and desktop applications) interconnected among them and supported  by advanced computing and storage infrastructures. As shown in the architectural model exposed in Figure 1, the system can be described mainly in three main modules that cooperate but which can also work autonomously. This autonomy is given thanks to the degree of specialisation of each of those modules since they have been designed with the aim of enabling their reuse in future systems, either as separate individual modules, or in collaboration with others. These three modules implement services that expose RESTful (Fielding and Taylor, 2000) interfaces. These interfaces enable their connection via the internet when using the hypertext transfer protocol secure protocol (HTTPS). This feature allows a federated deployment of the components in different physical locations (i.e., in different computing centres from an infrastructure such as that of IBERGRID/EGI (EGI Webpage, https://www.egi.eu/), in a cloud computing provider such as Amazon AWS or Microsoft Azure, or in both of them simultaneously). Besides, it also provokes that the overall system performance depends on the available bandwidth and the underlying network latency.     ALOE platform: an overview of a service-oriented architecture for research 249   The following sub-sections try to describe in detail the three main modules that shape the ALOE platform and the profiles involved. 2.1 Profiles Figure 2 shows the profiles supported by the ALOE platform. The defined profiles allow different levels of access described as follow: 1 General public: this is the profile in ALOE with the lowest level of access. It will be granted to people without any prior knowledge about breast cancer CADx methods. Users included in this profile will have access to: •   the ontology of concepts on breast cancer •   a reduced set of selected use-cases from BCDR •   the classification results of validated machine learning classifiers. 2 Medical students: in addition to the privileges of the general public profile, users included in this profile will have full access to: •   the content of BCDR (to see full patients’ records and information) •   the execution of all the developed machine learning classifiers •   all the self-auto-evaluation facilities. 3 Internal medical doctors: in addition to the privileges of the medical students profile, users included in this profile will have full access to: •   the validation and enhancement of the breast cancer ontology •   the edition of the clinical information of new patients’ cases and the identification and classification of new lesions or abnormalities •   the validation of the self-evaluation facilities. 4 Specialised radiologists: this is the profile with the highest set of privileges. In addition to the previously mentioned privileges, users included in this profile will have full access to: •   the addition of conclusive diagnoses to new patients on BCDR •   the evaluation of the performance of all supported breast cancer CADx methods on clinical practice •   the edition of the patient cases that profiles with lower privileges will be able to view. 5 Allied science researchers: user included on this profile will take charge of: •   Selecting best performing subsets of features introduced by the specialised radiologists for training new machine learning classifiers. •   Training breast cancer CADx methods to be included in ALOE. These CADx methods are the ones used by the previous profile.
Search
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks