LEE- Biyomedikal Mühendisliği-Yüksek Lisans
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 3 / 3
-
ÖgeDetection of liver fibrosis on slide-level labeled unstained biopsies by quantitative phase imaging and multiple instance learning(Graduate School, 2023-02-14)In recent years, advances in artificial intelligence (AI) have had a great impact in medicine, particularly in diagnostic fields such as radiology and pathology. In this context, the recognition and detection of important patterns in medical data can be made with AI, which can improve diagnostic processes, automate and make them more accurate and efficient, as well as increase sensitivity and consistency. One area where AI has great potential is digital pathology (DP), which is increasingly replacing traditional histopathological examination. In DP, tissue slides are turned into high-quality digital images and it is very convenient to perform detection, segmentation and classification tasks using AI on these data. On the other hand, biopsy specimens are stained for imaging with contrast in conventional pathology and current digital pathology. However, these methods have disadvantages such as staining errors, color differences due to different stain reagent manufacturers and different staining conditions. Quantitative phase imaging (QPI) is a very suitable method to overcome these disadvantages. Biological samples used in classical microscope examination are very thin and poorly absorb and scatter light. Therefore, no contrast occurs during imaging without staining, ie they are transparent. Whereas conventional bright field imaging measures only the amplitude, OPI allows measurement of both amplitude and phase information of the sample. Thus, transparent samples can be imaged with high contrast via QPI without any staining. QPI has great potential for use in cell biology, imaging of red blood cells, and diagnosis of diseases by histopathological examination, such as cancer. Fourier Ptychography (FP) microscopy is a computational imaging technique that can perform QPI. FP combines two classical optical techniques, Phase Retrieval and Aperture Synthesis, to obtain a complex image (amplitude and phase) of the sample by iteratively combining low-resolution images illuminated from different angles in Fourier space. FP provides two advantages: 1- It provides high resolution images while maintaining a wide field of view (FOV), 2- It acquires quantitative phase information without using complex optical interference techniques. In this study, the aim was to detect liver fibrosis using Attention-based Deep Multiple Instance Learning (MIL) from phase images of liver biopsies obtained by FP microscopy. In this framework, very large liver biopsy phase images are divided into patches and fed to the model, and the model infers healthy or diseased states at the biopsy image level. Also, the attention layer used in this architecture ensures that different attention weights are assigned to each patch in each biopsy image. Thus, besides classification at the biopsy image level, the most significant patches on the slide leading to the positive label can be found. Thus, it can provide support to the pathologist in the detection of disease-related tissue regions. One of the aims of this study is to try to eliminate the subjectivity and inconsistency of traditional histopathological examination due to staining using QPI. Another aim of the study is to train a deep learning model to accurately diagnose liver fibrosis through the MIL framework with only biopsy-level labeling without any local labeling by the pathologist on phase images of liver biopsies. In addition, the attention mechanism ensures that areas of high diagnostic value are found. Thus, this approach has the potential to overcome some of the limitations of AI in medical diagnosis, such as interpretability and explainability. Phase images of liver biopsies with FP were obtained with high contrast and high resolution. As a result of deep learning studies on these data, an average of 88% accuracy was achieved and the most important patches were detected by the deep learning model. It was confirmed by the pathologist that these regions were associated with fibrosis. The results of this study show that FP and Attention-based MIL are very promising for digital pathology.
-
ÖgeNovel design of transducer for bone conduction use(Graduate School, 2023-06-22)Hearing losses are worldwide acknowledged health problems affecting overall life quality of suffering patients. Many different therapeutic prostheses were developed for these hearing traumas. One of the most preferred treatment process are hearing aids. Different types of hearing aids were developed to rehabilitate and treat people suffering from hearing traumas. One such type of hearing aid is the bone conduction (BC) hearing aid, which is helpful for indicated patients. BC has been used for many decades in hearing aids, and there are now several implantable versions. These hearing aids transmit the sound through the skull bones to the inner part of the ear, bypassing the outer and middle parts. BC hearing aids are essentially designed with a transducer principle. The transducers are used to convert energy from one form to another. This process is called as transduction as well. Bone conduction transducers transform incoming audio signals to the vibrations that is transmitted to the cochlea, which transmits converted signals to the brain through auditory nerves. We aim to find a magnetostrictive transducer design with optimized data. The obtained optimized values will be used the possible a transducer prototype design for bone conduction use. This prototype is properly able to perform together with an audio processor, microphones, receiver coil, supermagnet and a battery compartment. We investigated material options to replace piezoelectric ceramics that are frequently used in BC hearing implants. During investigations for material selection, the properties of performing effectiveness and biocompatibility of alloys were watched out. According to the literature review, Terfenol D, Galfenol, and Metglas 2714A® were identified as possible materials for bender in the transducer. The bender component of the simulated transducer is extremely critical for this study. It consists of both magnetostrictive and non-magnetostrictive alloys. Other parts of transducer are counterweight or core, coil, permanent magnet and tape. The material selections for components transducer apart from bender, inspired from other current transducer applications. ANSYS® Mechanical and Electronics environment was chosen for simulations and testing. Some mechanical parameters were determined before starting simulations by reviewing current transducer applications to create a transducer model in ANSYS®. Then, mechanical simulations were performed in the specific ranges of mechanical sizes by using this model. Mechanical dimensions were simulated and optimized regarding size and resonance frequency, similar to existing bone conduction transducers. We determined the reference resonance frequency value as inspired by previous bone conduction transducer studies. For resonance frequency optimization, modal and harmonic analysis were performed. Modal analysis was used to determine specific parameters to create a mathematical model that shows dynamic reaction of the vibrating structure. Meshing adjustments were arranged to have more efficient simulation results. We had specific boundary conditions and, they were applied in the mechanical simulations. The boundary condition adjustment was to show the effect of skull simulator on the connector. Point of mass and cylindirical support parameters were applied to the model as boundary conditions. On the other hand, 1N Force was utilized in the model as the load. As a result of modal and harmonic response analysis, it was found that 0.3 and 0.4 mm for Metglas 2714A®, and 0.4 and 0.5 mm for Galfenol and Terfenol D, were appropriate thicknesses. Furthermore, for all three materials, 20 mm length and 4.8 mm width were evaluated as appropriate in electromagnetic simulations. Electromagnetic simulations were performed by adding different types of super-strong neodymium permanent magnets and turns of the multilayer coil. Also, the optimized mechanical dimensions obtained from mechanical simulatios through ANSYS was utilized in electromagnetic simulations. Besides, mechanical dimesions of electromagnetic components were determined by studies carried out by previous similar studies. Furthermore, resistance, inductance, and the number of turns of the coil were calculated for each simulation. After evaluation, Metglas 2714A® Magnetic alloy and Terfenol D were deemed less suitable for this application because of their size and robustness. Optimized mechanical dimensions and electromagnetic parameters were suggested for ferromagnetic material to construct a bone conduction transducer prototype. It is suggested to use Galfenol alloy with 0.5 mm thickness, 20 mm length, and 4.8 mm for build a magnetostrictive transducer prototype for BC use by using a permanent neodymium magnet with 200-250 turns of the coil.
-
ÖgeDenetimsiz derin öğrenme kullanılarak dijital meme tomosentezi görüntülerinde bulanıklığın giderilmesi(Lisansüstü Eğitim Enstitüsü, 2023-09-25)Her yıl binlerce kişinin ölümüne neden olan kanserin teşhisinde, tümörün doğru lokalizasyonu tedavi süreci için hayati önem taşımaktadır. Hastalığın erken teşhisi, kanserli dokunun büyümesini ve daha fazla dokuya yayılmasını önlemek için çok önemlidir. Tıbbi görüntüleme teknikleri, kanserin erken teşhisine önemli katkı sağlamaktadır. Kadınlarda kansere bağlı ölüm oranlarında en yüksek paya sahip olan meme kanserinin erken teşhisinde de görüntüleme teknolojilerinden yararlanılmaktadır. Özellikle 40 yaş üstü kadınların kendilerini kontrol etmeleri ve ailelerindeki kanser öyküsüne bağlı olarak düzenli taramalardan geçmeleri önerilmektedir. Bu tarama işlemlerinde yaygın olarak mamografi kullanılmakla birlikte, dijital meme tomosentezi (DBT) kullanımı da giderek yaygınlaşmaktadır. Mamografi iki boyutlu (2B) görüntüler oluştururken, DBT üç boyutlu (3B) görüntüler oluşturur. DBT, mamografide üst üste binme nedeniyle gizli kalmış lezyonların teşhisini mümkün kılmaktadır. Ayrıca taramanın farklı açılarda yapılması, oluşturulan görüntülerle yanlış pozitif oranlarını azaltılmasını sağlamaktadır. Taramaların sınırlı açılarla gerçekleştirilmesi, önlenemeyen hasta hareketleri ve dedektör odaklama aşamasındaki aksaklıklar DBT görüntülerinde bulanıklık artefaktlarına neden olabilmektedir. Bulanıklık etkisi, görüntü kalitesini gözle görülür şekilde düşürerek uzmanın görüntüdeki anormallikleri fark etme olasılığının da düşmesine sebep olmaktadır. Bahsi geçen dezavantajların azaltılması için, farklı modalitelerde çeşitli görüntü restorasyon teknikleri uygulanmaktadır. Bu çalışmada, DBT görüntülerinde tanısal doğruluğu azaltan bulanıklık etkisinin önüne geçebilmek için bir derin öğrenme modeli önerilmektedir. Temel gerçek görüntülere erişimin mevcut olmaması veya zor bulunması nedeniyle, model denetimsiz olarak tasarlanmıştır. Önerilen modelin eğitimi için kanser görüntüleme arşivinden alınan ve 5060 hastaya ait normal, işlem yapılabilir, biyopsi ile kanıtlanmış iyi huylu ve biyopsi ile kanıtlanmış kanser olarak etiketlenmiş görüntüleri içeren Meme Kanseri Taraması-Dijital Meme Tomosentezi (BSC-DBT) veri seti kullanılmaktadır. İşlemci kapasitesinin az olması ve dicom görüntülerin saklanması için çok fazla bellek gerekmesinden ötürü görüntüler png uzantılı olarak kaydedilmektedir. Biçimleri değiştirilen görüntüler ilk olarak kontrast uyarlamalı adaptif histogram eşitleme (CLAHE) ve yeniden boyutlandırma işlemlerini içeren bir ön işleme sürecine tabi tutulmaktadır. Veri setinden seçilen 2600 görüntü eğitim verisi, 350 görüntü validasyon verisi ve 260 görüntü ise test verisi olarak ayrılmıştır. Seçilen verilere farklı seviyelerde kernel boyutları ve sigma değerleri ile Gaussian bulanıklık filtreleri uygulanmıştır. AE (Autoencoder) ve GANs (Generative Adversarial Networks) modellerinin art arda kullanılması ile tasarlanan bu model 2B konvolüsyon, batch normalizasyon, LeakyReLU aktivasyonu, MaxPool ve Dropout katmanlarını içermektedir. Tasarlanan model, bulanık görüntüyü temel alarak orijinal görüntüyü yeniden oluşturmayı amaçlar. Bulanık görüntü, gerçek dünya uygulamalarında yaygın olarak bulunan düşük kaliteli veya bozulmuş görüntülerin modellenmesi amacıyla kullanılmaktadır. Önerilen modelin başarısı kontrast-gürültü oranı (CNR), ortalama mutlak hata (MAE) ve yapısal benzerlik indeksi (SSIM) ve uzman radyolog tarafından yapılan yorumlar ile değerlendirilmiştir. Nicel değerlendirmeler sonucunda bulanıklıktan arındırılan görüntülerin, bulanık görüntülere kıyasla orijinal görüntülere daha benzer olduğunu ortaya koymaktadır. Hem nicel sonuçlar hem de nitel değerlendirmeler, önerilen modelin DBT görüntülerindeki bulanıklaştırma artefaktlarını ele almada oldukça umut verici olduğunu ve bunun da teşhis doğruluğunu artırmasının beklenebileceğini göstermektedir.