Kemal Turhan

A novel automatic suspicious mass regions identification using Havrda & Charvat entropy and Otsu's N thresholding

Mass detection is a very important process for breast cancer diagnosis and computer aided systems. It can be very complex when the mass is small or invisible because of dense breast tissue. Therefore, the extraction of suspicious mass region can be very challenging. This paper proposes a novel segmentation algorithm to identify mass candidate regions in mammograms. The proposed system includes three parts: breast region and pectoral muscle segmentation, image enhancement and suspicious mass regions identification. The first two parts have been examined in previous studies. In this study, we focused on suspicious mass regions identification using a combination of Havrda & Charvat entropy method and Otsus N thresholding method. An open access Mammographic Image Analysis Society (MIAS) database, which contains 59 masses, was used for the study. The proposed system obtained a 93% sensitivity rate for suspicious mass regions identification in 56 abnormal and 40 normal images.

Medical images enhancement by using anisotropic filter and CLAHE

The purpose of image enhancement is to process an acquired image for better contrast and visibility of features of interest for visual examination as well as subsequent computer-aided analysis and diagnosis. Therefore, we have proposed an algorithm for medical images enhancement. In the study, we used top-hat transform, contrast limited histogram equalization and anisotropic diffusion filter methods. The system results are quite satisfactory for many different medical images like lung, breast, brain, knee and etc.

A Novel Algorithm for Segmentation of Suspicious Microcalcification Regions on Mammograms

Microcalcifications can be defined as the earliest sign of breast cancer and the early detection is very important. However, detection process is difficult because of their small size. Computer-based systems can assist the radiologist to increase the diagnostic accuracy. In this paper, we presented an automatic suspicious microcalcification regions segmentation system which can be used as a preprocessing step for microcalcifications detection. Our proposed system includes two main steps; preprocessing and segmentation. In the first step, we have implemented mammography image enhancement using top-hat transform and breast region segmentation using 3x3 median filtering, morphological opening and connected component labeling (CCL) methods. In the second step, a novel algorithm has been improved for segmentation of suspicious microcalcification regions. In the proposed segmentation algorithm, first Otsu’s N=3 thresholding, then dilation process and CCL methods have been applied on preprocessed mammography image. After this process, we took the upper region from the biggest two regions and if the pixels number of the taken region was greater than the limit value, that means the upper region was the pectoral muscle region and should be removed from the image. The limit value was determined according to the database results and prevented the false region segmentation for mammography images which have no pectoral muscle region. Successful results have been obtained on MIAS database.

Automatic Microcalcification Segmentation Using Rough Entropy and Fuzzy Approach

Microcalcifications have been mainly targeted as the earliest sign of breast cancer, thus their early detection is very important process. Since their size is very small and sometimes hidden by breast tissue, computer-based detection output can assist the radiologist to increase the diagnostic accuracy. This paper presents a research on mammography images using rough entropy and fuzzy approach. Our proposed method includes two main steps; preprocessing and segmentation. In the first step, we have implemented mammography image enhancement using wavelet transform, CLAHE and anisotropic diffusion filter then rough pectoral muscle extraction for false region reduction and better segmentation. In the second step, we have used Rough entropy to define a threshold and then, fuzzy based microcalcification enhancement, after these microcalcifications have been segmented using an iterative detection algorithm. By the combination of these methods, a novel hybrid algorithm has been developed and successful results have been obtained on MIAS database.

Analysis of Inter-rater Reliability of the Mammography Assessment after Image Processing

The aim of this study is to assess whether image processing causes information lost on mammography images. In the study, 50 mammogram from MIAS database (open database of mammograms) are selected: 20 images include mass, 20 images include calcification and 10 normal images. Selected images are read and marked by radiologists. The same radiologists read the enhanced version of images after three months later. In order to assess the consistency, inter-rater reliability statistics are used. Results indicate that image processing on mammography images especially images without calcification does not affect radiologists’ evaluation consistency. Also, it is indicated that images including calcifications reduce evaluation consistency of the radiologists and it is decided to use other image processing methods for images with calcifications.

Mortality Prediction with Lactate and Lactate Dehydrogenase

It has been proved in many studies that Lactate and Lactate dehydrogenase (LDH) are associated with mortality. In this study lactate test values of inpatients were analyzed with Support Vector Machines (SVM) to identify patients in high risk of death. In the data set containing 686 records with lactate results; 219 patients treated in the pediatric service and 467 of the patients are adults. Lactate levels of 331 patients are normal and levels of 355 patients are high. 89 patients with high lactate levels were recorded as dead. 97%, 96.6% and 92.3% accurate mortality classification rates were recorded with analyzes performed using different data sets and variables. Patient’s risk assessment can be assessed with such findings and treatments can be planned. Prediction of patients under high risk can provide opportunities for early intervention and mortality levels can be red.

A Model for Analyzing the Relation between Potassium (K) and Hemolysis Index (HI) with Clustering Method

This study was done to analyze the relation between potassium levels (K) and hemolysis index (HI). A different method for this kind of study -cluster analysis- was used to classify data, according to its hemolysis level. 5 clusters were obtained as a result of the cluster analysis using the absolute differences of first and last K and HI values. Regression analysis was used to understand impact of the hemolysis on K levels at each of 5 different clusters. Results which had HI difference values higher than 295 mg/dL were showed a very high correlation with K.