Browsing by Author "Tasdemir, Kasim"

Now showing 1 - 15 of 15

Citation - WoS: 21
Citation - Scopus: 28
Automated Quantification of Immunomagnetic Beads and Leukemia Cells from Optical Microscope Images
(Elsevier Sci Ltd, 2019) Uslu, Fatma; Icoz, Kutay; Tasdemir, Kasim; Yilmaz, Bulent
Quantification of tumor cells is crucial for early detection and monitoring the progress of cancer. Several methods have been developed for detecting tumor cells. However, automated quantification of cells in the presence of immunomagnetic beads has not been studied. In this study, we developed computer vision based algorithms to quantify the leukemia cells captured and separated by micron size immunomagnetic beads. Color, size based object identification and machine learning based methods were implemented to quantify targets in the images recorded by a bright field microscope. Images acquired by a 40x or a 20x objective were analyzed, the immunomagnetic beads were detected with an error rate of 0.0171 and 0.0384 respectively. Our results reveal that the proposed method attains 91.6% precision for the 40x objective and 79.7% for the 20x objective. This algorithm has the potential to be the signal readout mechanism of a biochip for cell detection. (C) 2019 Elsevier Ltd. All rights reserved.
Citation - WoS: 6
Citation - Scopus: 7
BLSTM Based Night-Time Wildfire Detection From Video
(Public Library Science, 2022) Agirman, Ahmet K.; Tasdemir, Kasim
Distinguishing fire from non-fire objects in night videos is problematic if only spatial features are to be used. Those features are highly disrupted under low-lit environments because of several factors, such as the dynamic range limitations of the cameras. This makes the analysis of temporal behavior of night-time fire indispensable for classification. To this end, a BLSTM based night-time wildfire event detection from a video algorithm is proposed. It is shown in the experiments that the proposed algorithm attains 95.15% of accuracy when tested against a wide variety of actual recordings of night-time wildfire incidents and 23.7 ms per frame detection time. Moreover, to pave the way for more targeted solutions to this challenging problem, experiment-based thorough investigations of possible sources of incorrect predictions and discussion of the unique nature of night-time wildfire videos are presented in the paper.
Citation - WoS: 4
Citation - Scopus: 4
Camera-Based Wildfire Smoke Detection for Foggy Environments
(SPIE - Society of Photo-Optical Instrumentation Engineers, 2022) Tas, Merve; Tas, Yusuf; Balki, Oguzhan; Aydin, Zafer; Tasdemir, Kasim
Smoke is the first visible sign of forest fires and the most commonly used feature for early forest fire detection using data from cameras. However, one of the natural challenges is the dense fog that appears in forests, which decreases the detection accuracy or triggers false alarms. In this study, we propose a system with a deep neural network-based image preprocessing approach that significantly improves the smoke segmentation and classification performance by dehazing the camera view. Our experimental results provide that the classification models reach 99% F1 score for the correct classification of smoke when the image dehazing method is used before the training process. The smoke localization system achieves 60% average precision when the mask region-based convolutional neural network is used with the ResNet101-FPN backbone. The proposed approach can be utilized for all smoke segmentation frameworks to increase fire detection performance. (c) 2022 SPIE and IS&T
Citation - WoS: 8
Citation - Scopus: 11
Deep Learning Based Semantic Segmentation and Quantification for MRD Biochip Images
(Elsevier Sci Ltd, 2022) Celebi, Fatma; Tasdemir, Kasim; Icoz, Kutay
Microfluidic platforms offer prominent advantages for the early detection of cancer and monitoring the patient response to therapy. Numerous microfluidic platforms have been developed for capturing and quantifying the tumor cells integrating several readout methods. Earlier, we have developed a microfluidic platform (MRD Biochip) to capture and quantify leukemia cells. This is the first study which employs a deep learning-based segmentation to the MRD Biochip images consisting of leukemic cells, immunomagnetic beads and micropads. Implementing deep learning algorithms has two main contributions; firstly, the quantification performance of the readout method is improved for the unbalanced dataset. Secondly, unlike the previous classical computer vision -based method, it does not require any manual tuning of the parameters which resulted in a more generalized model against variations of objects in the image in terms of size, color, and noise. As a result of these benefits, the proposed system is promising for providing real time analysis for microfluidic systems. Moreover, we compare different deep learning based semantic segmentation algorithms on the image dataset which are acquired from the real patient samples using a bright-field microscopy. Without cell staining, hyper-parameter optimized, and modified U-Net semantic segmentation algorithm yields 98.7% global accuracy, 86.1% mean IoU, 92.2% mean precision, 92.2% mean recall and 92.2% mean F-1 score measure on the patient dataset. After segmentation, quantification result yields 89% average precision, 97% average recall on test images. By applying the deep learning algorithms, we are able to improve our previous results that employed conventional computer vision methods.
Citation - WoS: 8
Energy Efficient Cosine Similarity Measures According to a Convex Cost Function
(Springer London Ltd, 2017) Akbas, Cem Emre; Gunay, Osman; Tasdemir, Kasim; Cetin, A. Enis
We propose a new family of vector similarity measures. Each measure is associated with a convex cost function. Given two vectors, we determine the surface normals of the convex function at the vectors. The angle between the two surface normals is the similarity measure. Convex cost function can be the negative entropy function, total variation (TV) function and filtered variation function constructed from wavelets. The convex cost functions need not to be differentiable everywhere. In general, we need to compute the gradient of the cost function to compute the surface normals. If the gradient does not exist at a given vector, it is possible to use the sub-gradients and the normal producing the smallest angle between the two vectors is used to compute the similarity measure. The proposed measures are compared experimentally to other nonlinear similarity measures and the ordinary cosine similarity measure. The TV-based vector product is more energy efficient than the ordinary inner product because it does not require any multiplications.
Citation - WoS: 13
Citation - Scopus: 14
Image-Analysis Based Readout Method for Biochip: Automated Quantification of Immunomagnetic Beads, Micropads and Patient Leukemia Cell
(Pergamon-Elsevier Science Ltd, 2020) Uslu, Fatma; Icoz, Kutay; Tasdemir, Kasim; Dogan, Refika S.; Yilmaz, Bulent
For diagnosing and monitoring the progress of cancer, detection and quantification of tumor cells is utmost important. Beside standard bench top instruments, several biochip-based methods have been developed for this purpose. Our biochip design incorporates micron size immunomagnetic beads together with micropad arrays, thus requires automated detection and quantification of not only cells but also the micropads and the immunomagnetic beads. The main purpose of the biochip is to capture target cells having different antigens simultaneously. In this proposed study, a digital image processing-based method to quantify the leukemia cells, immunomagnetic beads and micropads was developed as a readout method for the biochip. Color, size-based object detection and object segmentation methods were implemented to detect structures in the images acquired from the biochip by a bright field optical microscope. It has been shown that manual counting and flow cytometry results are in good agreement with the developed automated counting. Average precision is 85 % and average error rate is 13 % for all images of patient samples, average precision is 99 % and average error rate is 1% for cell culture images. With the optimized micropad size, proposed method can reach up to 95 % precision rate for patient samples with an execution time of 90 s per image.
Citation - WoS: 6
Citation - Scopus: 7
Improved Senescent Cell Segmentation on Bright-Field Microscopy Images Exploiting Representation Level Contrastive Learning
(Wiley, 2024) Celebi, Fatma; Boyvat, Dudu; Ayaz-Guner, Serife; Tasdemir, Kasim; Icoz, Kutay
Mesenchymal stem cells (MSCs) are stromal cells which have multi-lineage differentiation and self-renewal potentials. Accurate estimation of total number of senescent cells in MSCs is crucial for clinical applications. Traditional manual cell counting using an optical bright-field microscope is time-consuming and needs an expert operator. In this study, the senescence cells were segmented and counted automatically by deep learning algorithms. However, well-performing deep learning algorithms require large numbers of labeled datasets. The manual labeling is time consuming and needs an expert. This makes deep learning-based automated counting process impractically expensive. To address this challenge, self-supervised learning based approach was implemented. The approach incorporates representation level contrastive learning component into the instance segmentation algorithm for efficient senescent cell segmentation with limited labeled data. Test results showed that the proposed model improves mean average precision and mean average recall of downstream segmentation task by 8.3% and 3.4% compared to original segmentation model.
Citation - WoS: 14
Citation - Scopus: 21
Mitigation of H.264 and H.265 Video Compression for Reliable PRNU Estimation
(IEEE-Inst Electrical Electronics Engineers Inc, 2020) Altinisik, Enes; Tasdemir, Kasim; Sencar, Husrev Taha
The photo-response non-uniformity (PRNU) is a distinctive image sensor characteristic, and an imaging device inadvertently introduces its sensor's PRNU into all media it captures. Therefore, the PRNU can be regarded as a camera fingerprint and used for source attribution. The imaging pipeline in a camera, however, involves various processing steps that are detrimental to PRNU estimation. In the context of photographic images, these challenges are successfully addressed and the method for estimating a sensor's PRNU pattern is well established. However, various additional challenges related to generation of videos remain largely untackled. With this perspective, this work introduces methods to mitigate disruptive effects of widely deployed H.264 and H.265 video compression standards on PRNU estimation. Our approach involves an intervention in the decoding process to eliminate a filtering procedure applied at the decoder to reduce blockiness. It also utilizes decoding parameters to develop a weighting scheme and adjust the contribution of video frames at the macroblock level to PRNU estimation process. Results obtained on videos captured by 28 cameras show that our approach increases the PRNU matching metric up to more than five times over the conventional estimation method tailored for photos. Tests on a public dataset also verify that the proposed method improves the attribution performance by increasing the accuracy and allowing the use of smaller length videos to perform attribution.
Citation - Scopus: 2
Mixture of Learners for Cancer Stem Cell Detection Using Cd13 and H&E Stained Images
(SPIE - The International Society for Optics and Photonics, 2016) Oguz, Oguzhan; Akbas, Cem Emre; Mallah, Maen; Tasdemir, Kasim; Guzelcan, Ece Akhan; Muenzenmayer, Christian; Atalay, Rengul Cetin
In this article, algorithms for cancer stem cell (CSC) detection in liver cancer tissue images are developed. Conventionally, a pathologist examines of cancer cell morphologies under microscope. Computer aided diagnosis systems (CAD) aims to help pathologists in this tedious and repetitive work. The first algorithm locates CSCs in CD13 stained liver tissue images. The method has also an online learning algorithm to improve the accuracy of detection. The second family of algorithms classify the cancer tissues stained with H&E which is clinically routine and cost effective than immunohistochemistry (IHC) procedure. The algorithms utilize 1D-SIFT and eigen-analysis based feature sets as descriptors. Normal and cancerous tissues can be classified with 92.1% accuracy in H&E stained images. Classification accuracy of low and high-grade cancerous tissue images is 70.4%. Therefore, this study paves the way for diagnosing the cancerous tissue and grading the level of it using HSLE stained microscopic tissue images.
MRD Biyoçip İçin Görüntü İşleme Temelli Sinyal Elde Etme Metodu
(IEEE, 2019) Uslu, Fatma; Icoz, Kutay; Tasdemir, Kasim
The response of the cancer patients to chemotherapy treatment varies from person to person. For some patients cancer cells are resistant to treatment and these cells can relapse again which is known as minimal residual disease. A microfluidic-based biochip capable of monitoring minimal residual disease is under development by our research group. The role of the biochip is to capture the target cells, which were separated by immunomagnetic beads on micro square tiles. Then biochips are imaged using a bright field optical microscope and it is planned to perform image-processing methods to detect the target cells, immunomagnetic beads and micro tiles. In this work the current progress of image processing methods for differentiating the immunomagnetic beads and micro tiles is presented.
Citation - Scopus: 1
PCB Component Recognition With Semi-Supervised Image Clustering
(IEEE, 2021) Unal, Ahmet Emin; Tasdemir, Kasim; Bahcebasi, Akif
Classification of surface mounted devices plays an important role on automated inspection systems of printed component board production. Limited number of publicly available datasets which the components are labeled and high intraclass variance in these datasets causes the supervised approches to be inefficient. In this study a deep learning method, enhanced with an unsupervised clustering system, which uses a small set of labeled data is proposed. The method compared with the current studies and the supervised systems. Most optimized setting reached high accuracy results by outrunning current classification methods.
Citation - WoS: 15
Citation - Scopus: 14
A Review of Mammographic Region of Interest Classification
(Wiley Periodicals, inc, 2020) Yengec Tasdemir, Sena B.; Tasdemir, Kasim; Aydin, Zafer
Early detection of breast cancer is important and highly valuable in clinical practice. X-ray mammography is broadly used for prescreening the breast and is also attractive due to its noninvasive nature. However, experts can misdiagnose a significant proportion of the cases, which may either cause redundant examinations or cancer. In order to reduce false positive and negative rates of mammography screening, computer-aided breast cancer detection has been studied for more than 30 years and many methods have been proposed by the researchers. In this review, region of interest (ROI) classification methods, which operate on a predefined or segmented ROIs with a focus on mass classification are surveyed. A total of 72 high quality journal and conference papers are selected from the Web of Science (WOS) database that meet several inclusion criteria. A comparative analysis is provided based on ROI extraction methods, data sets and machine learning techniques employed, the prediction accuracies, and usage frequency statistics. Based on the performances obtained on publicly available data sets, the ROI classification problem from mammogram images can be considered as approaching to be solved. Nonetheless, it can still be used as complementary information in breast cancer detection from the whole mammograms, which has room for improvement. This article is categorized under: Application Areas > Science and Technology Technologies > Machine Learning Technologies > Classification
Citation - WoS: 11
Citation - Scopus: 20
ROI Detection in Mammogram Images Using Wavelet-Based Haralick and Hog Features
(IEEE, 2018) Tasdemir, Sena Busra Yengec; Tasdemir, Kasim; Aydin, Zafer
Digital mammography is a widespread medical imaging technique that is used for early detection and diagnosis of breast cancer. Detecting the region of interest (ROI) helps to locate the abnormal areas, which may be analyzed further by a radiologist or a CAD system. In this paper, a new classification method is proposed for ROI detection in mammography images. Features are extracted using Wavelet transform, Haralick and HOG descriptors. To reduce the number of dimensions and eliminate irrelevant features, a wrapper-based feature selection method is implemented. Several feature extraction methods and machine learning classifiers are compared by performing a leave-one-image-out cross-validation experiment on a difficult dataset. The proposed feature extraction method provides the best accuracy of 87.5% and the second-best area under curve (AUC) score of 84% when employed in a random forest classifier.
Citation - WoS: 23
Citation - Scopus: 30
Volume Fraction, Thickness, and Permeability of the Sealing Layer in Microbial Self-Healing Concrete Containing Biogranules
(Frontiers Media S.A., 2018) Ersan, Yusuf Cagatay; Palin, Damian; Tasdemir, Sena Busra Yengec; Tasdemir, Kasim; Jonkers, Henk M.; Boon, Nico; De Belie, Nele
Autonomous repair systems in construction materials have become a promising alternative to current unsustainable and labor-intensive maintenance methods. Biomineralization is a popular route that has been applied to enhance the self-healing capacity of concrete. Various axenic microbial cultures were coupled with protective carriers, and their combination appears to be useful for the development of healing agents for realizing self-healing concrete. The advantageous traits of non-axenic cultures, such as economic feasibility, self-protection, and high specific activity have been neglected so far, and thus the number of studies investigating their performance as healing agents is scarce. Here we present the self-healing performance of a mortar containing a healing agent consisting of non-axenic biogranules with a denitrifying core. Mortar specimens with a defined crack width of 400 mu m were used in the experiments and treated with tap water for 28 days. Self-healing was quantified in terms of the crack volume reduction, the thickness of the sealing layer along the crack depth and water permeability under 0.1 bar pressure. Complete visual crack closure was achieved in the bio-based specimens in 28 days, the thickness of the calcite layer was recorded as 10 mm and the healed crack volume was detected as 6%. Upon self-sealing of the specimens, the water permeability decreased by 83%. Overall, non-axenic biogranules with a denitrifying core shows great potential for development of self-healing bioconcrete.
YSA Kullanılarak Mamogramlardan Dokusal Öznitelik Tabanlı Meme Kanseri İlgi Bölgesi Sınıflandırılması
(2020) Taşdemir, Sena Büşra Yengeç; Tasdemir, Kasim; Aydin, Zafer
Radyoloji uzmanlarının mamografi görüntülerine bakarak yaptığı meme kanseriteşhislerinde tip bir hata oranı yüzde otuzlara kadar çıkmaktadır. Kanserin teşhisbaşarısını artırmak adına bu çalışmada uzmanlara yardımcı olacak yeni birBilgisayar Yardımlı Teşhis sistemi, kanserli ve normal dokuyu ayırt etmek içinönerilmektedir. Önerilen sistemde kontrast limitli histogram eşitleme (CLAHE)yöntemiyle iyileştirilen görüntülerin iki boyutlu parçacık dönüşümlerinden (2B–DWT) Haralick ve HOG öznitelikleri çıkarılmıştır. Özniteliklerin sayısını azaltmasıiçin temel bileşenler analizi (PCA) algoritması kullanılmıştır. Seçilen öznitelikler çokkatmanlı algılayıcı (MLP) mimari yapısına sahip yapay sinir ağına (YSA) girdi olarakverilmiştir. Çok katmanlı algılayıcı üzerinde Adam eniyileme yapıldığında %81tespit doğruluğu yakalanmıştır. Ayrıca, diğer bir çok temel makine öğrenmesi vederin öğrenme yöntemleri denenerek karşılaştırma sonuçları detaylı olaraksunulmuştur. Sınırlı sayıda veri kümesi kullanıldığında transfer öğrenim kullanılsadahi derin öğrenme yöntemlerinin tespit başarısı azalmıştır. Buna karşılık doğru önişleme, öznitelik seçilimi ve makine öğrenmesi yaklaşımları kullanıldığı zamangeleneksel bilgisayarlı görü yöntemleri daha başarılı sonuçlar vermiştir