Wyniki wyszukiwania - BazTech

1

Optimizing ultrasound image classification through transfer learning: fine-tuning strategies and classifier impact on pre-trained inner-layers

Bal-Ghaoui Mohamed, Alaoui Hachem El Yousfi, Jilbab Abdelilah, Bourouhou Abdennaser

Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska

|

2023

|

T. 13, nr 4

27--33

EN

Transfer Learning (TL) is a popular deep learning technique used in medical image analysis, especially when data is limited. It leverages pre-trained knowledge from State-Of-The-Art (SOTA) models and applies it to specific applications through Fine-Tuning (FT). However, fine-tuning large models can be time-consuming, and determining which layers to use can be challenging. This study explores different fine-tuning strategies for five SOTA models (VGG16, VGG19, ResNet50, ResNet101, and InceptionV3) pre-trained on ImageNet. It also investigates the impact of the classifier by usinga linear SVM for classification. The experiments are performed on four open-access ultrasound datasets related to breast cancer, thyroid nodules cancer, and salivary glands cancer. Results are evaluated using a five-fold stratified cross-validation technique, and metrics like accuracy, precision, and recall are computed. The findings show that fine-tuning 15% of the last layers in ResNet50 and InceptionV3 achieves good results. Using SVM for classification further improves overall performance by 6% for the two best-performing models. This research provides insights into fine-tuning strategiesandthe importance of the classifier in transfer learning for ultrasound image classification.

PL

Transfer Learning (TL) to popularna technika głębokiego uczenia stosowana w analizie obrazów medycznych, zwłaszcza gdy ilość danych jestograniczona. Wykorzystuje ona wstępnie wyszkoloną wiedzę z modeli State-Of-The-Art (SOTA) i zastosowanie ich do konkretnych aplikacji poprzez dostrajanie (Fine-Tuning –FT). Jednak dostrajanie dużych modeli może być czasochłonne, a określenie, których warstw użyć, może stanowić wyzwanie.W niniejszym badaniu przeanalizowano różne strategie dostrajania dla pięciu modeli SOTA (VGG16, VGG19, ResNet50, ResNet101 i InceptionV3) wstępnie wytrenowanych na ImageNet. Zbadano również wpływ klasyfikatora przy użyciu liniowej SVM do klasyfikacji. Eksperymenty przeprowadzonona czterech ogólnodostępnych zbiorach danych ultrasonograficznych związanych z rakiem piersi, rakiem guzków tarczycy i rakiemgruczołów ślinowych. Wyniki są oceniane przy użyciu techniki pięciowarstwowej walidacji krzyżowej, a wskaźniki takie jak dokładność, precyzja i odzyskiwanie są obliczane. Wyniki pokazują, że dostrojenie 15% ostatnich warstw w ResNet50 i InceptionV3 osiąga dobre wyniki. Użycie SVM do klasyfikacjidodatkowo poprawia ogólną wydajność o 6% dla dwóch najlepszych modeli. Badania te zapewniają informacje na temat strategii dostrajania i znaczenia klasyfikatoraw uczeniu transferowym dla klasyfikacji obrazów ultrasonograficznych.

2

Cancer gene recognition from microarray data with manta ray based enhanced ANFIS technique

Mishra Purnendu, Bhoi Nilamani

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 3

916--932

EN

Recognizing the cancer genes from the microarray dataset is considered as the most essential research topic in bioinformatics and computational biology domain. Microarray dataset represents the state of each cell at the molecular level which is identified as the important diagnostic tool in medical field. Analyzing the microarray data may provide a huge support for cancer gene classification. Therefore recently a number of artificial intelligence and machine learning techniques are developed which utilize the microarray data for distinguishing the cancer and non-cancer cells. But still now these techniques does not achieved a satisfactory performance. Therefore, an efficient technique that provides a crisp output for cancer classification is required. To overcome such defect, an enhanced ANFIS (EANFIS) method is used in this proposed architecture for classifying the cancer genes. The convergence time of ANFIS gets increased during learning process, therefore to avoid such issue the Manta ray foraging optimization (MaFO) algorithm is hybrid along with ANFIS which improves the overall classification performance. The data given as an input to the classification process is pre-processed at the initial phase using the Ensemble Kalman Filter (EnKF) technique. After pre-processing, the genes having similar properties are clustered using an adaptive density-based spatial clustering with noise (ADBSCAN) clustering technique. Finally, the performance of proposed enhanced ANFIS is evaluated using the precision, accuracy, f-measure, recall, sensitivity, and specificity metrics. Further, the clustering based performance evaluation is also carried out using the cluster index metrics. Finally, the comparison with the state-of-the-art techniques is also performed to show the effectiveness of proposed approach.

3

Role of image processing in the cancer diagnosis

Detyna J., Jeleń Ł., Jeleń M.

Bio-Algorithms and Med-Systems

|

2011

|

Vol. 7, no. 4

5--9

EN

Cancer is still one of the most deadly diseases. It is a well known fact that the early diagnosis is crucial and allows for the successful treatment while cancers diagnosed in their late stage are almost impossible to treat. For precise and objective diagnosis there is a need for a computerized method for cytological image processing, which is an integral part of a diagnosis process. In this work we present a classification system for grading cancer malignancy. In particular, issues of image processing in the aspect of medical diagnosis presented by prof. R. Tadeusiewicz and Dr. J. Śmietański in [1].

4

Grading breast cancer malignancy with neural networks

Jeleń Ł., Lipiński A., Detyna J., Jeleń M.

Bio-Algorithms and Med-Systems

|

2011

|

Vol. 7, no. 2

47--53

EN

Breast cancer is one of the most often diagnosed cancers among middle-aged women. It is a well known fact that the early diagnosis is crucial and allows for the successful treatment while cancers diagnosed in their late stage are almost impossible to treat. For precise and objective diagnosis there is a need for a computerized method for malignancy grading, which is an integral part of a diagnosis process. In this work we present a classification system for grading cancer malignancy based on the Bloom – Richardson grading scheme. This is a well known grading scheme among the pathologist used during the diagnosis process. To achieve such a classification we extracted 16 features that were then used to classify the malignancy into two classes. Each class represents the malignancy of the cancer according to Bloom – Richardson grading scheme. According to that scheme two types of features are considered, where each type is extracted from images recorded at two different magnifications. Three structural features were calculated from low magnification images and thirteen polymorphic features were derived from high magnification images. To classify the malignancy grades, the multilayer perceptron was used. The described system was able to classify the malignancy with the error rate of 13.5%. In this paper we also present first clinical trials that allow for the verification of the obtained classification rate. The clinical trial showed that the depicted system has a high performance achieving an accuracy of 93.08% .

5

A hybrid classifier based on SVM method for cancer clssification

Piątkowska W., Martyna J.

Studia Informatica

|

2009

|

Vol. 30, nr 2A

229-307

EN

n this paper, we proposed a new method of applying Support Vector Machines (SVMs) for cancer classification. We proposed a hybrid classifier that considers the degree of a membership function of each class with the help of Fuzzy Naive Bayes (FNB) and then organizes one-versus-rest (OVR) SVMs as the architecture classifying into the corresponding class. In this method, we used a novel system of ordering the recognized expression profiles by means of using FNB and genering SVMs with the OVR scheme. The results show that our hybrid classifier is comparable to the conventional methods.

PL

W artykule zaproponowano nową metodę klasyfikacji chorób onkologicznych. Użyto w niej m.in. naiwnego, rozmytego klasyfikatora bayesowskiego (ang. Fuzzy Naive Bayes) oraz maszyny z wektorami wspierającymi (ang. Support Vector Machines) jako systemu klasyfikującego. Tak powstały hybrydowy klasyfikator klasyfikuje choroby onkologiczne porównywalnie z konwencjonalnymi metodami.