Wyniki wyszukiwania - BazTech

1

Enhanced computer-aided system for breast lesion classification and grading using novel radio frequency time series approach

Ghehi Elaheh Norouzi, Fallah Ali, Rashidi Saeid, Dastjerdi Maryam Mehdizadeh

Biocybernetics and Biomedical Engineering

|

2025

|

Vol. 45, iss. 4

655--668

EN

Objective: Breast cancer is the most common malignancy among women and leading cause of mortality. Accurate, non-invasive differentiation of benign and malignant lesions is a clinical priority to reduce unnecessary biopsies and enable timely treatment. Elastography and RF time series (RF TS) processing are effective ultrasound-based techniques for tissue characterization. To improve their accuracy, we introduced an innovative approach called RFTSDP (RF Time Series Dynamic Processing). In RFTSDP, data are recorded during mechanical stimulation, revealing tissue properties in RF echoes. Extracting relevant features enhances computer-aided methods and improves tissue classification and grading. Materials and methods: An implement was developed to induce vibrations at different frequencies. Data were collected from ex-vivo tissues embedded in normal mimicking phantoms. Raw focused, raw, and beamformed ultrafast data were recorded under no stimulation, constant force, and various vibrational stimulations using the Supersonic Imaging Aixplorer ultrasound system. Features were extracted from each RF TS across the time, time-frequency, spectral, and non-linear domains. Multiple classifiers were evaluated, among which support vector machines with different kernels achieved the best results. Results: Beyond the classification of cancerous versus non-cancerous tissue, we also classified different cancerous lesion types and graded invasive ductal carcinoma. The best results were achieved with beamformed ultrafast data under 65 Hz vibrational stimulation. The mean classification accuracies for 2-, 3-, and 5-class were 99.78 %, 99.06 % and 99.32 %, respectively. Conclusions: The outcomes affirm that applying vibration, particularly at an optimal frequency, enhances breast tissue classification. The proposed method demonstrated efficacy not only in distinguishing between cancerous and non-cancerous lesions but also in grading cancerous tissues.

2

Motor imagery EEG signal classification using optimized convolutional neural network

Thiyam Deepa Beeta, Raymond Shelishiyah, Avasarala Padmanabha Sarma

Przegląd Elektrotechniczny

|

2024

|

R. 100, nr 8

273--279

EN

Motor Imagery (MI) signals help the Brain-Computer Interface framework (BCI) to enable the binding of the human brain to external devices. Thus, both BCI and MI together are instrumental in enhancing the lives of patients affected by motor neuron disorders. A novel MI Electroencephalography (EEG) signal identification and classification approach is proposed in this work. An error-free extraction algorithm is required to extract and classify the temporal and spatial features successfully. This paper proposes the Hilbert Transform (HT) for band energy analysis and Gabor Filter for the selection of optimal frequency band. In this work, the Wavelet Packet Decomposition (WPD) algorithm is used for feature extraction and it decomposes the signal into high and low-frequency components before extracting band coefficients. Moreover, the Convolution Neural Network (CNN) classifier is employed for the classification of MI-EEG tasks. The classification accuracy of the CNN classifier is enhanced using Sea Lion Optimization (SLno) algorithm. The approach is verified using MATLAB and the results are substantially better than those found in the current research, with an average classification accuracy rate of 96.44% by employing a smaller number of criteria, lessening resource consumption, and eliminating the influence of individual differences. The recommended method minimizes classification computation time while enhancing classification accuracy.

XX

Sygnały obrazu motorycznego (MI) pomagają strukturze interfejsu mózg-komputer (BCI) umożliwić wiązanie ludzkiego mózgu z urządzeniami zewnętrznymi. Zatem zarówno BCI, jak i MI razem odgrywają zasadniczą rolę w poprawie życia pacjentów dotkniętych zaburzeniami neuronu ruchowego. W tej pracy zaproponowano nowatorskie podejście do identyfikacji i klasyfikacji sygnałów MI-Elektroencefalografii (EEG). Do pomyślnego wyodrębnienia i sklasyfikowania cech czasowych i przestrzennych wymagany jest bezbłędny algorytm ekstrakcji. W artykule zaproponowano transformatę Hilberta (HT) do analizy energii pasma oraz filtr Gabora do wyboru optymalnego pasma częstotliwości. W tej pracy do ekstrakcji cech wykorzystano algorytm Wavelet Packet Decomposition (WPD), który rozkłada sygnał na składowe o wysokiej i niskiej częstotliwości przed wyodrębnieniem współczynników pasma. Ponadto do klasyfikacji zadań MI-EEG wykorzystuje się klasyfikator Convolution Neural Network (CNN). Dokładność klasyfikacji klasyfikatora CNN jest zwiększona dzięki zastosowaniu algorytmu Sea Lion Optimization (SLno). Podejście to jest weryfikowane przy użyciu MATLAB-a, a wyniki są znacznie lepsze niż w bieżących badaniach, ze średnim współczynnikiem dokładności klasyfikacji wynoszącym 96.44% przy zastosowaniu mniejszej liczby kryteriów, mniejszym zużyciu zasobów i wyeliminowaniu wpływu różnic indywidualnych. Zalecana metoda minimalizuje czas obliczeń klasyfikacyjnych, jednocześnie zwiększając dokładność klasyfikacji.

3

An automated framework to segment and classify gliomas using hybrid shuffled complex evolution with convolutional

Valarmathy G., Umapathi N., Suganthi Su, Jegadeesan R.

Przegląd Elektrotechniczny

|

2024

|

R. 100, nr 7

158--165

EN

The infiltrative nature and rapid progression of gliomas have made them the most prevalent as well as aggressive type of brain tumour. In the clinical routine, it is a difficult task to distinguish tumour boundaries from the healthy cells. For brain tumour diagnoses as well as evaluations of the intra operative treatment response, there is extensive utilisation of the Magnetic Resonance Imaging (MRI). With segmentation, infected regions of the brain tissue can be extracted from MRIs. The tumour region’s segmentation is a critical task for cancer diagnosis, treatment as well as treatment outcome assessment. The significant architecture named the Convolutional Neural Network (CNN) in deep learning is used. The CNN has shown outstanding improvement in the objects’ recognition as well as classification. It has much efficiency in a wide range of problems which deal with machine learning as well as computer vision. Akin to other techniques of deep learning, much difficulty is involved in training the CNN. In this work, proposed a novel meta-heuristic based algorithms have been used for optimizing CNN using Ant Colony Optimization (ACO), hybrid Shuffled Complex Evolution (SCE) with ACO and hybrid SCE with Particle Swarm Optimization (PSO) algorithm. The results show that the proposed method produces better results than existing methods.

PL

Naciekowy charakter i szybki postęp glejaków uczyniły je najczęstszym i najbardziej agresywnym rodzajem nowotworu mózgu. W praktyce klinicznej odróżnienie granic guza od zdrowych komórek jest trudnym zadaniem. W diagnostyce guza mózgu, a także ocenie śródoperacyjnej odpowiedzi na leczenie szeroko wykorzystuje się obrazowanie metodą rezonansu magnetycznego (MRI). Dzięki segmentacji zakażone obszary tkanki mózgowej można wyodrębnić z rezonansu magnetycznego. Segmentacja regionu nowotworowego jest kluczowym zadaniem w diagnostyce nowotworu, jego leczeniu, a także ocenie wyników leczenia. W głębokim uczeniu się wykorzystywana jest znacząca architektura zwana konwolucyjną siecią neuronową (CNN). CNN wykazało wyjątkową poprawę w zakresie rozpoznawania i klasyfikacji obiektów. Ma dużą skuteczność w szerokim zakresie problemów związanych z uczeniem maszynowym i wizją komputerową. Podobnie jak w przypadku innych technik głębokiego uczenia się, szkolenie CNN wiąże się z wieloma trudnościami. W tej pracy zaproponowane nowatorskie algorytmy oparte na metaheurystyce zostały wykorzystane do optymalizacji CNN przy użyciu algorytmu Ant Colony Optimization (ACO), hybrydowej Shuffled Complex Evolution (SCE) z ACO i hybrydowego SCE z algorytmem Particle Swarm Optimization (PSO). Wyniki pokazują, że proponowana metoda daje lepsze wyniki niż metody istniejące.

4

Ekstrakcja parametrów z próbek danych biometrycznych

Danek Paweł, Ćwirta Krzysztof, Kopniak Piotr

Journal of Computer Sciences Institute

|

2019

|

Vol. 13

323--331

PL

W artykule opisano możliwe sposoby ekstrakcji parametrów z próbek danych biometrycznych, takich jak odcisk palca czy nagranie głosu. Zweryfikowano wpływ konkretnych sposobów obróbki na skuteczność algorytmów obróbki próbek biometrycznych oraz ich porównania. Wykonano badania polegające na przetworzeniu dużej liczby próbek z użyciem wybranych algorytmów. W przypadku odcisku palca wykorzystano normalizację obrazu, filtr Gabora i porównanie z użyciem deskryptorów. Dla autoryzacji głosowej analizowano algorytmy LPC i MFCC. W przypadku obu rodzajów autoryzacji uzyskano zadowalającą skuteczność rzędu 60-80%.

EN

This article describes possible ways to extract parameters from biometric data samples, such as fingerprint or voice recording. Influence of particular approaches to biometric sample preparation and comparision algorithms accuracy was verified. Experiment involving processing big ammount of samples with usage of particular algorithms was performed. In fingerprint detection case the image normalization, Gabor filtering and comparision method based on descriptors were used. For voice authorization LPC and MFCC alghoritms were used. In both cases satisfying accuracy (60-80%) was the result of the surveys.

5

Automatic parameters selection of Gabor filters with the imperialism competitive algorithm with application to retinal vessel segmentation

Farokhian F., Yang C., Demirel H., Wu S., Beheshti I.

Biocybernetics and Biomedical Engineering

|

2017

|

Vol. 37, no. 1

246--254

EN

Retinal images play an important role in the early diagnosis of diseases such as diabetes. In the present study, an automatic image processing technique is proposed to segment retinal blood vessels in fundus images. The technique includes the design of a bank of 180 Gabor filters with varying scale and elongation parameters. Furthermore, an optimization method, namely, the imperialism competitive algorithm (ICA), is adopted for automatic parameter selection of the Gabor filter. In addition, a systematic method is proposed to determine the threshold value for reliable performance. Finally, the performance of the proposed approach is analyzed and compared with that of other approaches on the basis of the publicly available DRIVE database. The proposed method achieves an area under the receiver operating characteristic curve of 0.953 and an average accuracy of up to 0.9392. Thus, the results show that the proposed method is well comparable with alternative methods in the literature.

6

Hierarchical classification of normal, fatty and heterogeneous liver diseases from ultrasound images using serial and parallel feature fusion

Alivar A., Danyali H., Helfroush M. S.

Biocybernetics and Biomedical Engineering

|

2016

|

Vol. 36, no. 4

697--707

EN

This study presents a computer-aided diagnostic system for hierarchical classification of normal, fatty, and heterogeneous liver ultrasound images using feature fusion techniques. Both spatial and transform domain based features are used in the classification, since they have positive effects on the classification accuracy. After extracting gray level co-occurrence matrix and completed local binary pattern features as spatial domain features and a number of statistical features of 2-D wavelet packet transform sub-images and 2-D Gabor filter banks transformed images as transform domain features, particle swarm optimization algorithm is used to select dominant features of the parallel and serial fused feature spaces. Classification is performed in two steps: First, focal livers are classified from the diffused ones and second, normal livers are distinguished from the fatty ones. For the used database, the maximum classification accuracy of 100% and 98.86% is achieved by serial and parallel feature fusion modes, respectively, using leave-one-out cross validation (LOOCV) method and support vector machine (SVM) classifier.

7

Real-Time Denim Fabric Inspection Using Image Analysis

Çelik H. I., Topalbekitoglu M., Dülger L. C.

Fibres & Textiles in Eastern Europe

|

2015

|

Vol. 23, no. 3 (111)

85--90

EN

In spite of using modern weaving technology, many types of fabric defects occur during production. Most defects arising in the production process of a fabric are still detected by human inspection. A machine vision system that can be adapted to different types of fabric inspection machines is proposed in this study. Image frames of denim fabric were acquired using a CCD line-scan camera. An algorithm was developed by using the Gabor filter and double thresholding methods. The performance of the algorithm was tested real-time by analysing a denim fabric sample which contained six types of defects: hole, warp lacking, weft lacking, soiled yarn, water soil and yarn flow (knot). The defective regions of the denim fabric sample were detected and labelled successfully.

PL

Mimo stosowania współczesnych technologii tkackich, wiele typów defektów tkaniny powstaje podczas ich produkcji. Większość defektów podczas procesu produkcji tkaniny w dalszym ciągu jest wykrywana poprzez kontrolę pracowników. W pracy zaproponowano automatyczny system inspekcji wizualnej, który można adoptować do różnych typów maszyn, realizujących inspekcje tkaniny. Wykonano ramki ilustrujące tkaninę typu Denim przy zastosowaniu liniowej kamery skanującej typu CCD. Opracowano algorytm stosując filtr Gabor A i metodę podwójnego progu dyskryminującego. Skuteczność algorytmu testowano analizując próbki tkaniny Denim zawierające 6 typów defektów - dziura, brak osnowy, brak wątku, zabrudzona przędza, zmoczona przędza i pęczki. Sukcesywnie określano zaznaczone uszkodzenia.

8

Amplitude and Frequency Modulation in Speaker Recognition Systems

Ciota Z.

International Journal of Microelectronics and Computer Science

|

2012

|

Vol. 3, nr 2

41-44

EN

The paper presents a review of the nowadays methods of voice vector extraction, applied in such speech processing, like person identification and emotion recognition. A special attention was held on mixed time-frequency analysis based on temporary frequency approach. The methods of calculation of time - frequency voice characterization were also described. The most important building blocks of identification and recognition of speakers have been presented. The characterization of feature vectors suitable for identification and verification in microcomputer systems was described. Components and appropriate method of speech identification based on the long-term spectra vectors were discussed.

9

Image retrieval based on hierarchical Gabor filters

Andrysiak T., Choraś M.

International Journal of Applied Mathematics and Computer Science

|

2005

|

Vol. 15, no. 4

471-480

EN

Content Based Image Retrieval (CBIR) is now a widely investigated issue that aims at allowing users of multimedia information systems to automatically retrieve images coherent with a sample image. A way to achieve this goal is the computation of image features such as the color, texture, shape, and position of objects within images, and the use of those features as query terms. We propose to use Gabor filtration properties in order to find such appropriate features. The article presents multichannel Gabor filtering and a hierarchical image representation. Then a salient (characteristic) point detection algorithm is presented so that texture parameters are computed only in a neighborhood of salient points. We use Gabor texture features as image content descriptors and efficiently emply them to retrieve images.