Wyniki wyszukiwania - BazTech

1

BA-Net: Brightness prior guided attention network for colonic polyp segmentation

Xia Haiying, Qin Yilin, Tan Yumei, Song Shuixiang

Biocybernetics and Biomedical Engineering

|

2023

|

Vol. 43, no. 3

603--615

EN

Automatic polyp segmentation at colonoscopy plays an important role in the early diagnosis and surgery of colorectal cancer. However, the diversity of polyps in different images greatly increases the difficulty of accurately segmenting polyps. Manual segmentation of polyps in colonoscopic images is time-consuming and the rate of polyps missed remains high. In this paper, we propose a brightness prior guided attention network (BA-Net) for automatic polyp segmentation. Specifically, we first aggregate the high-level features of the last three layers of the encoder with an enhanced receptive field (ERF) module, which further fed to the decoder to obtain the initial prediction maps. Then, we introduce a brightness prior fusion (BF) module that fuses the brightness prior information into the multi-scale side-out high-level semantic features. The BF module aims to induce the network to localize salient regions, which may be potential polyps, to obtain better segmentation results. Finally, we propose a global reverse attention (GRA) module to combine the output of the BF module and the initial prediction map for obtaining long-range dependence and reverse refinement prediction results. With iterative refinement from higher-level semantics to lower-level semantics, our BA-Net can achieve more refined and accurate segmentation. Extensive experiments show that our BA-Net outperforms the stateof-the-art methods on six common polyp datasets.

2

Segmentation of the melanoma lesion and its border

Surówka Grzegorz, Ogorzałek Maciej

International Journal of Applied Mathematics and Computer Science

|

2022

|

Vol. 32, no. 4

683--699

EN

Segmentation of the border of the human pigmented lesions has a direct impact on the diagnosis of malignant melanoma. In this work, we examine performance of (i) morphological segmentation of a pigmented lesion by region growing with the adaptive threshold and density-based DBSCAN clustering algorithm, and (ii) morphological segmentation of the pigmented lesion border by region growing of the lesion and the background skin. Research tasks (i) and (ii) are evaluated by a human expert and tested on two data sets, A and B, of different origins, resolution, and image quality. The preprocessing step consists of removing the black frame around the lesion and reducing noise and artifacts. The halo is removed by cutting out the dark circular region and filling it with an average skin color. Noise is reduced by a family of Gaussian filters 3×3−7×7 to improve the contrast and smooth out possible distortions. Some other filters are also tested. Artifacts like dark thick hair or ruler/ink markers are removed from the images by using the DullRazor closing images for all RGB colors for a hair brightness threshold below a value of 25 or, alternatively, by the BTH transform. For the segmentation, JFIF luminance representation is used. In the analysis (i), out of each dermoscopy image, a lesion segmentation mask is produced. For the region growing we get a sensitivity of 0.92/0.85, a precision of 0.98/0.91, and a border error of 0.08/0.15 for data sets A/B, respectively. For the density-based DBSCAN algorithm, we get a sensitivity of 0.91/0.89, a precision of 0.95/0.93, and a border error of 0.09/0.12 for data sets A/B, respectively. In the analysis (ii), out of each dermoscopy image, a series of lesion, background, and border segmentation images are derived. We get a sensitivity of about 0.89, a specificity of 0.94 and an accuracy of 0.91 for data set A, and a sensitivity of about 0.85, specificity of 0.91 and an accuracy of 0.89 for data set B. Our analyses show that the improved methods of region growing and density-based clustering performed after proper preprocessing may be good tools for the computer-aided melanoma diagnosis.

3

Applications of machine-learning algorithms for prediction of benign and malignant breast lesions using ultrasound radiomics signatures: A multi-center study

Homayoun Hassan, Chan Wai Yee, Kuzan Taha Yusuf, Leong Wai Ling, Altintoprak Kübra Murzoglu, Mohammadi Afshin, Vijayananthan Anushya, Rahmat Kartini, Leong Sook Sam, Mirza-Aghazadeh-Attari Mohammad, Ejtehadifar Sajjad, Faeghi Fariborz, Acharya U. Rajendra, Ardakani Ali Abbasian

Biocybernetics and Biomedical Engineering

|

2022

|

Vol. 42, no. 3

921--933

EN

Artificial intelligence (AI) algorithms have an enormous potential to impact the field of radiology and diagnostic imaging, especially the field of cancer imaging. There have been efforts to use AI models to differentiate between benign and malignant breast lesions. However, most studies have been single-center studies without external validation. The present study examines the diagnostic efficacy of machine-learning algorithms in differentiating benign and malignant breast lesions using ultrasound images. Ultrasound images of 1259 solid non-cystic lesions from 3 different centers in 3 countries (Malaysia, Turkey, and Iran) were used for the machine-learning study. A total of 242 radiomics features were extracted from each breast lesion, and the robust features were considered for models’ development. Three machine-learning algorithms were used to carry out the classification task, namely, gradient boosting (XGBoost), random forest, and support vector machine. Sensitivity, specificity, accuracy, and area under the ROC curve (AUC) were determined to evaluate the models. Thirty-three robust features differed significantly between the two groups from all of the features. XGBoost, based on these robust features, showed the most favorable profile for all cohorts, as it achieved a sensitivity of 90.3%, specificity of 86.7%, the accuracy of 88.4%, and AUC of 0.890. The present study results show that incorporating selected robust radiomics features into well-curated machine-learning algorithms can generate high sensitivity, specificity, and accuracy in differentiating benign and malignant breast lesions. Furthermore, our results show that this optimal performance is preserved even in external validation datasets.

4

Two-stage multi-scale breast mass segmentation for full mammogram analysis without user intervention

Yan Yutong, Conze Pierre-Henri, Quellec Gwenolé, Lamard Mathieu, Cochener Beatrice, Coatrieux Gouenou

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 2

746--757

EN

Mammography is the primary imaging modality used for early detection and diagnosis of breast cancer. X-ray mammogram analysis mainly refers to the localization of suspicious regions of interest followed by segmentation, towards further lesion classification into benign versus malignant. Among diverse types of breast abnormalities, masses are the most important clinical findings of breast carcinomas. However, manually segmenting breast masses from native mammograms is time-consuming and error-prone. Therefore, an integrated computer-aided diagnosis system is required to assist clinicians for automatic and precise breast mass delineation. In this work, we present a two-stage multiscale pipeline that provides accurate mass contours from high-resolution full mammograms. First, we propose an extended deep detector integrating a multi-scale fusion strategy for automated mass localization. Second, a convolutional encoder-decoder network using nested and dense skip connections is employed to fine-delineate candidate masses. Unlike most previous studies based on segmentation from regions, our framework handles mass segmentation from native full mammograms without any user intervention. Trained on INbreast and DDSM-CBIS public datasets, the pipeline achieves an overall average Dice of 80.44% on INbreast test images, outperforming state-of-the-art. Our system shows promising accuracy as an automatic full-image mass segmentation system. Extensive experiments reveals robustness against the diversity of size, shape and appearance of breast masses, towards better interaction-free computer-aided diagnosis.

5

Microaneurysm detection in fundus images based on a novel end-to-end convolutional neural network

Liao Yinhan, Xia Haiying, Song Shuxiang, Li Haisheng

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 2

589--604

EN

Microaneurysms are the earliest symptom of diabetic retinopathy and play an important role in the screening of diabetic retinopathy. However, because of the complex background, automatic detection microaneurysm in fundus images is a challenging task. Firstly, motivated by the characteristics of microaneurysm, a novel deep convolutional encoder-decoder network for microaneurysm detection is designed to locate the MAs by the differences between the skip connection in the network. Then, a weighted dice loss, termed the smooth dice loss, is presented to put more focus on misclassified microaneurysms. Finally, an activation function with a long tail is used to produce an accurate probability map for MA detection. Plenty of experiments, conducted on the Retinopathy Online Challenge data-set and the e-ophtha-MA dataset, demonstrate that the proposed model achieves the comparable performance to the existing state-of-the-art methods on microaneurysm detection with only one-hundredth the running time compared with its counterparts. The proposed method is simple and effective, guarantees the performance while shortening the test time. It indicates the potential application in the auxiliary diagnosis of diabetic retinopathy screening.

6

Impact of noise on the performance of automatic systems for vocal fold lesions detection

Madruga Mario, Campos-Roca Yolanda, Pérez Carlos J.

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 3

1039--1056

EN

Automatic voice condition analysis systems have been developed to automatically discriminate pathological voices from healthy ones in the context of two disorders related to exudative lesions of Reinke’s space: nodules and Reinke’s edema. The systems are based on acoustic features, extracted from sustained vowel recordings. Reduced subsets of features have been obtained from a larger set by a feature selection algorithm based on Whale Optimization in combination with Support Vector Machine classification. Robustness of the proposed systems is assessed by adding noise of two different types (synthetic white noise and actual noise recorded in a clinical environment) to corrupt the speech signals. Two speech databases were used for this investigation: the Massachusetts Eye and Ear Infirmary (MEEI) database and a second one specifically collected in Hospital San Pedro de Alcántara (Cáceres, Spain) for the scope of this work (UEX-Voice database). The results show that the prediction performance of the detection systems appreciably decrease when moving from MEEI to a database recorded in more realistic conditions. For both pathologies, the prediction performance declines under noisy conditions, being the effect of white noise more pronounced than the effect of noise recorded in the clinical environment.

7

Deep learning on ultrasound images of thyroid nodules

Sharifi Yasaman, Bakhshali Mohamad Amin, Dehghani Toktam, DanaiAshgzari Morteza, Sargolzaei Mahdi, Eslami Saeid

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 2

636--655

EN

Due to safety, easy accessibility, noninvasively and cost-effectiveness of ultrasound imaging, this technology becomes one of the main contributors for analyzing thyroid nodules. However, interpretation of ultrasound images is a challenging task that subjects to the radiologist’s prior medical knowledge and observational skills. There is a significant need for reliable, objective, and automated approaches for the meaningful assessment of ultrasound images. Many areas of machine learning including computer vision and image processing have been revolutionized by the recent advances in the field of deep learning. The current study systematically reviews the existing literatures and evaluates technical characteristics of the deep learning applications on the ultrasound images of thyroid nodules. In this review, all of the included studies have been published from 2017 to 2020 indicating the recent growing interest in the utilization of deep learning-based techniques for assessment of ultrasound images of thyroid nodules. Although deep learning has demonstrated potential for analyzing thyroid nodules’ ultrasound images, this review highlights several existing barriers that need to be addressed in future works such as dealing with data limitation, generating public and valid datasets, and determining standard evaluation metrics. This survey outlines several methods (e.g., data augmentation and transfer learning) recently proposed to address similar challenges in other fields. Furthermore, to improve the diagnostic accuracy of the deep learning models, utilization of complementary information with multi-modal images are suggested.

8

Automatic detection of coronavirus disease (COVID-19) in X-ray and CT images: A machine learning based approach

Kassani Sara Hosseinzadeh, Kassani Peyman Hosseinzadeh, Wesolowski Michal J., Schneider Kevin A., Deters Ralph

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 3

867--879

EN

The newly identified Coronavirus pneumonia, subsequently termed COVID-19, is highly transmittable and pathogenic with no clinically approved antiviral drug or vaccine available for treatment. The most common symptoms of COVID-19 are dry cough, sore throat, and fever. Symptoms can progress to a severe form of pneumonia with critical complications, including septic shock, pulmonary edema, acute respiratory distress syndrome and multi-organ failure. While medical imaging is not currently recommended in Canada for primary diagnosis of COVID-19, computer-aided diagnosis systems could assist in the early detection of COVID-19 abnormalities and help to monitor the progression of the disease, potentially reduce mortality rates. In this study, we compare popular deep learningbased feature extraction frameworks for automatic COVID-19 classification. To obtain the most accurate feature, which is an essential component of learning, MobileNet, DenseNet, Xception, ResNet, InceptionV3, InceptionResNetV2, VGGNet, NASNet were chosen amongst a pool of deep convolutional neural networks. The extracted features were then fed into several machine learning classifiers to classify subjects as either a case of COVID-19 or a control. This approach avoided task-specific data pre-processing methods to support a better generalization ability for unseen data. The performance of the proposed method was validated on a publicly available COVID-19 dataset of chest X-ray and CT images. The DenseNet121 feature extractor with Bagging tree classifier achieved the best performance with 99% classification accuracy. The second-best learner was a hybrid of the a ResNet50 feature extractor trained by LightGBM with an accuracy of 98%.

9

Application of Artificial Intelligence techniques for the detection of Alzheimer’s disease using structural MRI images

Zhao Xinxing, Ang Candice Ke En, Acharya U. Rajendra, Cheong Kang Hao

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 2

456--473

EN

Alzheimer’s disease (AD) is an irreversible, progressive brain disorder that slowly destroys memory and thinking skills. It is one of the leading types of dementia for persons aged above 65 worldwide. In order to achieve accurate and timely diagnosis, and for detection of AD in its early stages, numerous Artificial Intelligence (AI) based Computer-aided Diagnostic (CAD) approaches have been proposed using data from brain imaging. In this paper, we review the recent application of AI based CAD systems on AD and its stages, with a particular focus on the use of structural MRI due to its cost effectiveness and lack of ionizing radiation. We will review important factors of different AI techniques pertinent to AD, summarize contributions from different research groups, critically discuss challenges involved and propose directions for future research. Ultimately, it would be ideal for development of a diagnostic framework that could be applicable to not only AD, but to different types of dementia as well in the future.

10

A comparative analysis of automatic classification and grading methods for knee osteoarthritis focussing on X-ray images

Saini Deepak, Chand Trilok, Chouhan Devendra K., Prakash Mahesh

Biocybernetics and Biomedical Engineering

|

2021

|

Vol. 41, no. 2

419--444

EN

Objective: The purpose of present review paper is to introduce the reader to key directions of manual, semi-automatic and automatic knee osteoarthritis (OA) severity classification from plain radiographs. This is a narrative review article in which we have described recent developments in severity evaluation of knee OA from X-ray images. We have primarily focussed on automatic analysis and have reviewed articles in which machine learning, transfer learning, active learning, etc. have been employed on X-ray images to access and classify the severity of knee OA. Methods: All original research articles on OA detection and classification using X-ray images published in English were searched on PubMed database, Google Scholar, RSNA radiology databases in year 2019. The search terms of ‘‘knee Osteoarthritis” were combined with search terms ‘‘Machine Learning”, ‘severity” and ‘‘X-ray”. Results: The initial search on various publication databases revealed a total of 743 results, out of which only 26 articles were considered relevant to radiographic knee OA severity analysis. The majority of the articles were based on automatic analysis. Manual segmentation based articles were least in numbers. Conclusion: Computer aided methods to diagnose knee OA are great tools to detect OA at ealry stages. Advancements in Human Computer Interface systems have led the researchers to bridge the gap between machine learning algorithms and expert healthcare professionals to provide better and timely treatment options to the knee OA affected patients.

11

Identification of HEp-2 specimen images with mitotic cell patterns

Gupta Krati, Bhavsar Arnav, Sao Anil K.

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 3

1233--1249

EN

In this paper, we propose and analyze a novel framework to identify the HEp-2 specimen images, consisting of mitotic spindle (MS) pattern cells. It is based on the fact that the cells showing MS patterns will always be present with other interphase type pattern cells, in a whole slide image (WSI) or specimen image, but the number of MS cells will be very small, unlike interphase patterns. Considering this fact, the work contributes in presenting a framework, using different strategies such as cells-based, region-based and complete image-based approaches. For cell-based approach, the distinctive characteristic of MS patterns is represented using morphology and texture-based features, followed by a traditional Support Vector Machine (SVM) based classifier, whereas the region-based approach uses a Convolutional Neural Network (CNN) as feature extractor and baseline classifier. Finally, for the image-based approach, the Faster Region-CNN (Faster-RCNN) based object detection framework has been applied, considering MS patterns as distinct objects. The region and image-based approaches also contribute in avoiding the requirement of DAPI based segmentation masks. Another contribution of this work is to use a novel and clearly specified threshold-based decision-making criteria for patterns declaration of the specimens with MS pattern cells. All the proposed strategies, integrated with decision-making criteria are validated on a publicly available dataset and across various experiments, we demonstrate good performance, i.e., max. MCC 0.92 in one case. Hence, the proposed framework proves to be an effective solution for the problem statement.

12

Classification of plantar foot alterations by fuzzy cognitive maps against multi-layer perceptron neural network

Ramirez-Bautista Julian Andres, Huerta-Ruelas Jorge Adalberto, Kóczy László T., Hatwágner Miklós F., Chaparro-Cárdenas Silvia L., Hernández-Zavala Antonio

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 1

404--414

EN

Load distribution analysis on foot surface allows knowing human mechanical behavior and aids the doctor in the detection of gait disorders like, the risk of foot ulcerations, leg discrepancy, and footprint alterations. Plantar pressure data combined with techniques that use integral reasoning produce easy understanding medical tools for assisting in treatment, early detection, and the development of preventive strategies. The present research compares the classification of human plantar foot alterations using Fuzzy Cognitive Maps (FCM) trained by Genetic Algorithm (GA) against a Multi-Layer Perceptron Neural Network (MLPNN). One hundred and fifty-one subject volunteers (aged 7–77) were classified previously with the flat foot (n = 70) and cavus foot (n = 81) by specialized physicians of the Piédica diagnostic center. The trial walking was conducted using plantar pressure platforms FreeMed®. The foot surface was divided into 14 areas that included toe 1 st to 5th, metatarsal joint 1st to 5th, lateral midfoot, medial midfoot, lateral heel, and medial heel. Pressure data were normalized for each area. Better performance in the classification using small amounts of data were found by using Fuzzy rather than non-Fuzzy approach.

13

An automated computer-aided diagnosis system for classification of MR images using texture features and gbest-guided gravitational search algorithm

Shanker Ravi, Bhattacharya Mahua

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 2

815--835

EN

The segmentation and classification of brain magnetic resonance (MR) images are the crucial and challenging task for radiologists. The conventional methods for analyzing brain images are time-consuming and ineffective in decision-making. Thus, to overcome these limita-tions, this work proposes an automated and robust computer-aided diagnosis (CAD) system for accurate classification of normal and abnormal brain MR images. The proposed CAD system has the ability to assist the radiologists for diagnosis of brain MR images at an early stage of abnormality. Here, to improve the quality of images before their segmentation, contrast limited adaptive histogram equalization (CLAHE) is employed. The segmentation of the region of interest is obtained using the multilevel Otsu's thresholding algorithm. In addition, the proposed system selects the most significant and relevant features from the texture and multiresolution features. The multiresolution features are extracted using discrete wavelet transform (DWT), stationary wavelet transform (SWT), and fast discrete curvelet transform (FDCT). Moreover, the Tamura and local binary pattern (LBP) are used to extract the texture features from the images. These features are used to classify the brain MR images using feedforward neural network (FNN) classifier, where different meta-heuristic optimization algorithms, e.g., genetic algorithm (GA), particle swarm optimization (PSO), gravitational search algorithm (GSA), and gbest-guided gravitational search algorithm (GG-GSA) are employed for optimizing the weights and biases of FNN. The extensive experimen-tal results on DS-195, DS-180, and three standard datasets show that the classification accuracy of GG-GSA based FNN classifier outperforms all mentioned meta-heuristic-based classifiers and several state-of-the-art methods.

14

A novel privacy-supporting 2-class classification technique for brain MRI images

Devi Swagatika, Sahoo Manmath Narayan, Bakshi Sambit

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 3

1022--1035

EN

Developing automated Computer Aided Diagnosis (CAD) framework for assisting radiologists in a fast and effective classification of brain Magnetic Resonance (MR) images is of great importance, given plausible usage of Electronic Health Records (EHR) in healthcare system. This work aims at proposing two novel privacy supporting classifiers for automatic segregation of brain MR images. To ensure privacy, our article employs a spatial steganographic approach to hide patients sensitive health information in brain images itself. Proposed methods employ Discrete Wavelet Transform (DWT) for extracting relevant features from original and stego images. Subsequently, Symmetrical Uncertainty Ranking (SUR) and Probabilistic Principal Components Analysis (PPCA) are used to obtain a reduced feature vector for Support Vector Machine (SVM) and Filtered Classifier (FC) respectively. The experiments are carried out on two benchmark datasets DS-75 and DS-160 collected from Harvard Medical School website and one larger input pool of self-collected dataset NITR-DHH. To validate this work, the proposed schemes are experimented on both original and stego brain MR images and are compared against eight state-of-the-art classification techniques with respect to six standard parameters. The results reveal that the proposed techniques are robust and scalable with respect to the size of the datasets. Moreover, it is concluded that applying steganographic algorithm on brain MR images yield equally satisfactory classification rate.

15

A deep learning approach to detect Covid-19 coronavirus with X-Ray images

Jain Govardhan, Mittal Deepti, Thakur Daksh, Mittal Madhup K.

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 4

1391--1405

EN

Rapid and accurate detection of COVID-19 coronavirus is necessity of time to prevent and control of this pandemic by timely quarantine and medical treatment in absence of any vaccine. Daily increase in cases of COVID-19 patients worldwide and limited number of available detection kits pose difficulty in identifying the presence of disease. Therefore, at this point of time, necessity arises to look for other alternatives. Among already existing, widely available and low-cost resources, X-ray is frequently used imaging modality and on the other hand, deep learning techniques have achieved state-of-the-art performances in computer-aided medical diagnosis. Therefore, an alternative diagnostic tool to detect COVID-19 cases utilizing available resources and advanced deep learning techniques is proposed in this work. The proposed method is implemented in four phases, viz., data augmentation, preprocessing, stage-I and stage-II deep network model designing. This study is performed with online available resources of 1215 images and further strengthen by utilizing data augmentation techniques to provide better generalization of the model and to prevent the model overfitting by increasing the overall length of dataset to 1832 images. Deep network implementation in two stages is designed to differentiate COVID-19 induced pneumonia from healthy cases, bacterial and other virus induced pneumonia on X-ray images of chest. Comprehensive evaluations have been performed to demonstrate the effectiveness of the proposed method with both (i) training-validation-testing and (ii) 5- fold cross validation procedures. High classification accuracy as 97.77%, recall as 97.14% and precision as 97.14% in case of COVID-19 detection shows the efficacy of proposed method in present need of time. Further, the deep network architecture showing averaged accuracy/ sensitivity/specificity/precision/F1-score of 98.93/98.93/98.66/96.39/98.15 with 5-fold cross validation makes a promising outcome in COVID-19 detection using X-ray images.

16

Multi-channel acoustic analysis of phoneme /s/ mispronunciation for lateral sigmatism detection

Krecichwost Michal, Miodonska Zuzanna, Badura Pawel, Trzaskalik Joanna, Mocko Natalia

Biocybernetics and Biomedical Engineering

|

2019

|

Vol. 39, no. 1

246--255

EN

The paper presents a method for computer-aided detection of lateral sigmatism. The aim of the study is to design an automated sigmatism diagnosis tool. For that purpose, a reference speech corpus has been collected. It contains 438 recordings of a phoneme /s/ surrounded by certain vowels with normative and simulated pathological pronunciation. The acoustic signal is recorded with an acoustic mask, which is a set of microphones organised in a semi-cylindrical surface around the subject's face. Frames containing /s/ phoneme are subjected to beamforming and feature extraction. Two different feature vectors containing, e.g., Mel-frequency cepstral coefficients and fricative formants, are defined and evaluated in terms of binary classification involving support vector machines. A single-channel analysis is con- fronted with multi-channel processing. The experimental results show that the multi-channel speech signal processing supported by beamforming is able to increase the pathology detection capabilities in general.

17

Automatic method for assessment of proliferation index in digital images of DLBCL tissue section

Gomolka Ryszard S., Korzynska Anna, Siemion Krzysztof, Gabor-Siatkowska Karolina, Klonowski Wlodzimierz

Biocybernetics and Biomedical Engineering

|

2019

|

Vol. 39, no. 1

30--37

EN

Diffuse large B-cell lymphoma (DLBCL) is a fast-growing and aggressive neoplasm originating from B lymphocytes. Evaluation of proliferation index (PI) based on Ki67 immunohistochemical nuclear staining is used to distinguish proliferating (immunopositive) from nonproliferating (immunonegative) lymphoma cells. Human interpretation of PI varies and is time-consuming, therefore automatic computer-assisted approach may facilitate the performance. Herein we propose a new fully automatic proliferation index estimation (FLAPIE) algorithm, dedicated to detection of immunopositive and immunonegative nuclei, and evaluation of PI in digital microscopy images of DAB&H-stained samples from patients with high-grade DLBCL. FLAPIE performs nuclei detection in original RGB colour space and is independent of image brightness due to its textural-statistical approach. Validation of FLAPIE was performed in 61 non-overlapping whole-slide imagefragments and compared to the results of PI estimation by QuPath open-source software, MetPiKi algorithm and manual evaluation by two independent observers. Interobserver agreement was calculated between the nuclei count and PIs by two observers. High concordance was found between both DAB and H-stained nuclei count, and PIs by two observers. Compared to MetPiKi, FLAPIE presented improved results of DAB and H-stained nuclei detection. In contrary to MetPiKi and QuPath, FLAPIE performed nuclei detection in all images and its results closely matched the number of DAB-stained nuclei evaluated by two observers. No significant difference was found between PIs by all computational methods and observers. FLAPIE achieved good results in PI estimation and prospectively aims to serve as a tool for clinical application in support of patients selection and decision to treatment.

18

Representation learning-based unsupervised domain adaptation for classification of breast cancer histopathology images

Alirezazadeh P., Hejrati B., Monsef-Esfahani A., Fathi A.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 3

671--683

EN

Breast cancer has high incidence rate compared to the other cancers among women. This disease leads to die if it does not diagnosis early. Fortunately, by means of modern imaging procedure such as MRI, mammography, thermography, etc., and computer systems, it is possible to diagnose all kind of breast cancers in a short time. One type of BC images is histology images. They are obtained from the entire cut-off texture by use of digital cameras and contain invaluable information to diagnose malignant and benign lesions. Recently by requesting to use the digital workflow in surgical pathology, the diagnosis based on whole slide microscopy image analysis has attracted the attention of many researchers in medical image processing. Computer aided diagnosis (CAD) systems are developed to help pathologist make a better decision. There are some weaknesses in histology images based CAD systems in compared with radiology images based CAD systems. As these images are collected in different laboratory stages and from different samples, they have different distributions leading to mismatch of training (source) domain and test (target) domain. On the other hand, there is the great similarity between images of benign tumors with those of malignant. So if these images are analyzed undiscriminating, this leads to decrease classifier performance and recognition rate. In this research, a new representation learning-based unsupervised domain adaptation method is proposed to overcome these problems. This method attempts to distinguish benign extracted feature vectors from those of malignant ones by learning a domain invariant space as much as possible. This method achieved the average classification rate of 88.5% on BreaKHis dataset and increased 5.1% classification rate compared with basic methods and 1.25% with state-of-art methods.

19

Automated quantification of ultrasonic fatty liver texture based on curvelet transform and SVD

Bharath R., Mishra P. K., Rajalakshmi P.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 1

145--157

EN

Fatty liver is a prevalent disease and is the major cause for the dysfunction of the liver. If fatty liver is untreated, it may progress into chronic diseases like cirrhosis, hepatocellular carcinoma, liver cancer, etc. Early and accurate detection of fatty liver is crucial to prevent the fatty liver progressing into chronic diseases. Based on the severity of fat, the liver is categorized into four classes, namely Normal, Grade I, Grade II and Grade III respectively. Ultrasound scanning is the widely used imaging modality for diagnosing the fatty liver. The ultrasonic texture of liver parenchyma is specific to the severity of fat present in the liver and hence we formulated the quantification of fatty liver as a texture discrimination problem. In this paper, we propose a novel algorithm to discriminate the texture of fatty liver based on curvelet transform and SVD. Initially, the texture image is decomposed into sub-band images with curvelet transform enhancing gradients and curves in the texture, then an absolute mean of the singular values are extracted from each curvelet decomposed image, and used it as a feature representation for the texture. Finally, a cubic SVM classifier is used to classify the texture based on the extracted features. Tested on a database of 1000 image textures with 250 image textures belonging to each class, the proposed algorithm gave an accuracy of 96.9% in classifying the four grades of fat in the liver.

20

Multi-step process in computer assisted diagnosis of posterior cruciate ligaments

Zarychta P.

Biocybernetics and Biomedical Engineering

|

2016

|

Vol. 36, no. 4

657--669

EN

A multi-step methodology resulting in a three-dimensional visualization and construction of feature vector of posterior cruciate ligament is presented. In the first step the location of the posterior cruciate ligament is established using the fuzzy image concept. The fuzzy image concept is based on the entropy measure of fuzziness extended to two dimensions. In order to reduce the area of analysis, the region of interest including the ligament structures is detected. In this case, the fuzzy C-means algorithm with median modification helping to reduce blurred edges was implemented. After finding the region of interest, the fuzzy connectedness procedure was performed. This procedure permitted to extract the ligament structures. On the basis of the extracted posterior cruciate ligament structures, the three-dimensional visualization of this ligament was built and, with the support of experts' knowledge, an appropriate feature vector was constructed and its values assigned for normal and pathological cases. Correct results were obtained for over 88% of 97 cases.