Wyniki wyszukiwania - BazTech

1

Classification of lung nodule malignancy in computed tomography imaging utilising generative adversarial networks and semi-supervised transfer learning

Apostolopoulos Ioannis D., Papathanasiou Nikolaos D., Panayiotakis George S.

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 4

1243--1257

EN

The pulmonary nodules’ malignancy rating is commonly confined in patient follow-up; examining the nodule’s activity is estimated with the Positron Emission Tomography (PET) system or biopsy. However, these strategies are usually after the initial detection of the malignant nodules acquired from the Computed Tomography (CT) scan. In this study, a Deep Learning methodology to address the challenge of the automatic characterisation of Solitary Pulmonary Nodules (SPN) detected in CT scans is proposed. The research methodology is based on Convolutional Neural Networks, which have proven to be excellent automatic feature extractors for medical images. The publicly available CT dataset, called Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI), and a small CT scan dataset derived from a PET/CT system, is considered the classification target. New, realistic nodule representations are generated employing Deep Convolutional Generative Adversarial Networks to circumvent the shortage of large-scale data to train robust CNNs. Besides, a hierarchical CNN called Feature Fusion VGG19 (FF-VGG19) was developed to enhance feature extraction of the CNN proposed by the Visual Geometry Group (VGG). Moreover, the generated nodule images are separated into two classes by utilising a semi-supervised approach, called self-training, to tackle weak labelling due to DC-GAN inefficiencies. The DC-GAN can generate realistic SPNs, as the experts could only distinguish 23% of the synthetic nodule images. As a result, the classification accuracy of FF-VGG19 on the LIDCIDRI dataset increases by +7%, reaching 92.07%, while the classification accuracy on the CT dataset is increased by 5%, reaching 84,3%.

2

Semi-supervised learning with the clustering and Decision Trees classifier for the task of cognitive workload study

Wawrzyk Martyna

Journal of Computer Sciences Institute

|

2020

|

Vol. 15

214--218

EN

The paper is focused on application of the clustering algorithm and Decision Tress classifier (DTs) as a semi-supervised method for the task of cognitive workload level classification. The analyzed data were collected during examination of Digit Symbol Substitution Test (DSST) with use of eye-tracker device. 26 participants took part in examination as vol-unteers. There were conducted three parts of DSST test with different levels of difficulty. As a results three versions were obtained of data: low, middle and high level of cognitive workload. The case study covered clustering of collected data by using k-means algorithm to detect three clusters or more. The obtained clusters were evaluated by three internal indices to measure the quality of clustering. The David-Boudin index detected the best results in case of four clusters. Based on this information it is possible to formulate the hypothesis of the existence of four clusters. The obtained clus-ters were adopted as classes in supervised learning and have been subjected to classification. The DTs was applied in classification. There were obtained the 0.85 mean accuracy for three-class classification and 0.73 mean accuracy for four-class classification.

PL

Celem artykułu było zastosowanie klasteryzacji wraz z klasyfikatorem Drzew Decyzyjnych jako częściowo nadzoro-wanej metody klasyfikacji poziomu obciążenia poznawczego. Dane przeznaczone do analizy zostały zebrane podczas badania DSST (z ang. Digit Symbol Substitution Test) z użyciem urządzenia eye-tracker. 26 wolontariuszów wzięło udział w badaniu. Zostały przeprowadzone trzy części testu DSST o różnych poziomach trudności. W wyniku tego, otrzymano trzy wersje danych: z niskim, średnim i wysokim poziomem obciążenia poznawczego. Do analizy danych został użyty algorytm klasteryzacji k-means do wyznaczenia trzech lub większej liczby klastrów. Uzyskane klastry zostały poddane ocenie przy użyciu trzech wewnętrznych indeksów w celu zmierzenia jakości klasteryzacji. Indeks David-Boudin’a wykazał najlepsze rezultaty w przypadku istnienia czterech klastrów. Na podstawie tej informacji można sformułować hipotezę, iż dane są podzielone na 4 klastry, co oznaczałoby istnienie dodatkowego poziomu poznawczego. Uzyskane klastry zostały zaadoptowane jako klasy w uczeniu pod nadzorem. Do klasyfikacji danych został użyty klasyfikator Drzew Decyzyjnych . Otrzymano średnią dokładność równą 0.85 w przypadku 3-klasowej klasyfikacji oraz 0.73 średnią dokładność dla 4-klasowej klasyfikacji.

3

Metody wykrywania naruszeń prywatności w internetowych sieciach społecznych

Zabielski M., Kasprzyk R., Tarapata Z.

Symulacja w Badaniach i Rozwoju

|

2015

|

Vol. 6, nr 2

153--161

PL

Wraz z pojawieniem się internetowych sieci społecznych znaczenie aspektu prywatności w Internecie wzrosło drastycznie. Stąd ważne jest opracowanie mechanizmów, które uniemożliwią osobom niepowołanym dostęp do prywatnych danych osobowych. W pracy podjęta została próba określenia modeli naruszeń prywatności poprzez analizę wpływu struktury sieci oraz jej atrybutów na możliwości naruszenia prywatności w internetowej sieci społecznej. Wynikiem tych działań jest opracowanie koncepcji symulatora pozwalającego na weryfikację wniosków wypływających z utworzonych modeli.

EN

With the arrival of online social networks, the importance of privacy on the Internet has increased dramatically. Thus, it is important to develop mechanisms that will prevent our hidden personal data from unauthorized access. In this paper an attempt was made to present some set of privacy violation detection models defined from local – appropriate person personal data – and global point of view – online social network structure. The result of this activities, despite models, is conception of simulator, which will allow us to verify conclusions from the analysis of online social networks privacy violation.

4

Coupling as Strategy for Reducing Concept-Drift in Never-ending Learning Environments

Hruschka Jr E. R., Duarte M. C., Nicoletti M. C.

Fundamenta Informaticae

|

2013

|

Vol. 124, nr 1/2

47--61

EN

The project and implementation of autonomous computational systems that incrementally learn and use what has been learnt to, continually, refine its learning abilities throughout time is still a goal far from being achieved. Such dynamic systems would conform to the main ideas of the automatic learning model conventionally characterized as never-ending learning (NEL). The never-ending approach to learning exhibits similarities to the semi-supervised (SS) model which has been successfully implemented by bootstrap learning methods. Bootstrap learning has been one of the most successful among the SS-methods proposed to date and, as such, the natural candidate for implementing NEL systems. Bootstrap methods learn from an available labeled set of data, use the induced knowledge to label some unlabeled new data and, recurrently, learn again from both sets of data in a cyclic manner. However the use of SS methods, particularly bootstrapping methods, to implement NEL systems can give rise to a problem known as concept-drift. Errors that may occur when the system automatically labels new unlabeled data can, over time, cause the system to run off track. The development of new strategies to lessen the impact of concept-drift is an important issue that should be addressed if the goal is to increase the plausibility of developing such systems, employing bootstrap methods. Coupling techniques can play an important role in reducing concept-drift effects over machine learning systems, particularly those designed to perform tasks related to machine reading. This paper proposes and formalizes relevant coupling strategies for dealing with the concept-drift problem in a NEL environment implemented as the system RTWP (Read The Web in Portuguese); initial results have shown they are promising strategies for minimizing the problem taking into account a few system settings.

5

Change Detection of Remote Sensing Images with Semi-supervised Multilayer Perceptron

Patra S., Ghosh S., Ghosh A.

Fundamenta Informaticae

|

2008

|

Vol. 84, nr 3-4

429-442

EN

A context-sensitive change-detection technique based on semi-supervised learning with multilayer perceptron is proposed here. In order to take contextual information into account, input patterns are generated considering each pixel of the difference image along with its neighboring pixels. A heuristic technique is suggested to identify a few initial labeled patterns without using ground truth information. The network is initially trained using these labeled data. The unlabeled patterns are iteratively processed by the already trained perceptron to obtain a soft class label. Experimental results, carried out on two multispectral and multitemporal remote sensing images, confirm the effectiveness of the proposed approach.