Ömer Kayaaltı

Staging of the liver fibrosis from CT images using texture features

Even though liver biopsy is critical for evaluating chronic hepatitis and fibrosis, it is an invasive, costly, and difficult to standardize approach. The developments in medical image processing and artificial intelligence methods have advanced the potential of using computer-aided diagnosis techniques in the classification of liver tissues. The aim of this study was to develop a non-invasive, cost-effective, and fast approach to specify fibrosis stage using the texture properties of computed tomography images of liver. Gray level co-occurrence matrix, discrete wavelet transform, and discrete Fourier transform were the image analysis tools in the feature extraction phase. Following dimension reduction of the texture features support vector machines and k-nearest neighbor methods were used in the classification phase of this study. Our results showed that our approach is feasible in fibrosis staging especially in pairwise stage comparisons with success rate of approximately 90%.

Texture analysis of liver cirrhosis

Liver with cirrhosis emerges when the cells of liver begin to die and the tissues become a functional knot from these. In the diagnosis of fibrosis, the needle biopsy is a golden standard. Although this technique is a good techique in reaching accurate diagnosis, its being an invasive method arises disadvantage. The developments in medical image processing and artificial intelligence techniques have advanced the potential of using diagnosis system in classification of liver tissues. In this study, we have aimed at producing some objective measures using image analysis, which will be of assistance in the diagnosis of cirrhosis. In order to differentiate between regions of liver with cirrhosis and healthy parenchymal tissues, we have used first order statistical texture features and second order texture features computed from gray level cooccurrence matrix of liver computerized tomography (CT) images. Then liver CT images of healthy people and people with cirrhosis have been classified with support vector machines (SVM) by using all these acquired features. The most successful classification has been calculated as 85.19% with the method of 10 fold cross-validation.

Comparison of lung tumor segmentation methods on PET images

Lung cancer is the most common cause of cancer-related deaths that occur all over the world. Recently, various image processing approaches have been used on PET images in order to characterize the uniformity, density, coarseness, roughness, and regularity (i.e., texture properties) of the intratumoral 18F-fluorodeoxyglucose (FDG) uptake. The first and important step of this kind of analysis is to differentiate tumor region from other structures and background, which is called segmentation. In this study, k-means, active contour (snake), and Otsus tresholding methods were applied on PET images obtained from 36 patients and the performances were compared by the nuclear medicine expert in our team. The results show that Otsu tresholding approach is more selective.

Registration and fusion of lung tumor PET/CT images

Image fusion attracts attention in medical field due to complementary behavior and application such as diagnosis and treatment planning. In this study, first positron emission tomography (PET) and computed tomography (CT) images coming from 8 nonsmall cell lung cancer were registered then wavelet and principal component analysis methods were applied to fuse images. According to mutual information metric and nuclear medicine expert wavelet method gave better results when compared to PCA.

Comparison of first order statistical and autoregressive model features for activity prediction

Activity recognition is an important subject with many applications in health care, emergency care, and assisted living. Nowadays, activity information can be acquired using small accelerometers connected to the body, including the ones available in smartphones. In this study, we assessed the influence of autoregressive model parameters or features on activity detection or classification. Our results indicate that, compared to relatively simple features such as first order statistics, autoregressive model features have rather low impact in determining or improving performance of automatic activity detection using machine intelligence.

Texture analysis of liver hydatid cyst

Images which are obtained in clinical radiology are generally evaluated visually. Some information which is available in the images, but not possible to be seen visually can be useful for diagnosis of some diseases. Cyst hydatid which is a parasitic liver disease is still an important health problem in countries where animal breeding is widespread. In this study, we aimed at producing some objective measures using image analysis, which will be of assistance in the diagnosis of cyst hydatid. The prevalence rate of this condition is relatively high in Turkey. In order to differentiate between regions of liver with cyst hydatid and healthy parenchymal tissues, we have used second order texture features computed from gray level cooccurrence matrix of liver CT images. We have then used these features from the two groups in designing a classifier using probabilistic neural network. Our results indicate that the texture features computed from the gray level cooccurrence matrix do not constitute a good candidate to be used in classification and/or diagnosis of liver tissue as normal or cystic. This must be due to homogeneity of these two tissue types within themselves.