Deep Learning Based Egg Fertility Detection.

This study investigates the implementation of deep learning (DL) approaches to the fertile egg-recognition problem, based on incubator images. In this study, we aimed to classify chicken eggs according to both segmentation and fertility status with a Mask R-CNN-based approach. In this manner, images can be handled by a single DL model to successfully perform detection, classification and segmentation of fertile and infertile eggs. Two different test processes were used in this study. In the first test application, a data set containing five fertile eggs was used. In the second, testing was carried out on the data set containing 18 fertile eggs. For evaluating this study, we used AP, one of the most important metrics for evaluating object detection and segmentation models in computer vision. When the results obtained were examined, the optimum threshold value (IoU) value was determined as 0.7. According to the IoU of 0.7, it was observed that all fertile eggs in the incubator were determined correctly on the third day of both test periods. Considering the methods used and the ease of the designed system, it can be said that a very successful system has been designed according to the studies in the literature. In order to increase the segmentation performance, it is necessary to carry out an experimental study to improve the camera and lighting setup prepared for taking the images.

Automatic phase reversal detection in routine EEG.

Electroencephalograph (EEG), a valuable tool in the clinical evaluation, is readily available, safe and provides information about brain function. EEG interpretation is important for the diagnosis of neurological disorders. The long-term EEG data may be required to document and study neurosciences that include many epileptic activities and phase reversal (PR) etc. However, analyze of the long-term EEG done by an expert neurologist is much time consuming and quite difficult. Therefore, an automatic PR determination method for analyzing of long-term EEG is described in this study. The presented technique was applied to the pathological EEG recordings that were obtained from two different datasets gathered as a retrospective in Selcuk University Hospital (SUH) and Boston Children's Hospital (BCH). With this method, PR in the dataset was determined and then compared with the ones detected by the specialist doctor. Two tests were carried out in the SUH dataset and the classification success of the method was 83.22% for test 1 and 85.19% for test 2. On the other hand, three tests were carried out for two different position values for BCH dataset. The highest classification success of the six tests was 75% for test 5, while the lowest classification success appeared as 58.33% for test 6. As a result, the overall success in the detection of PR with the conducted method is 84.20% for SUH and 66.7% for BCH. According to these results, the determination of PR that is known to be indicative of neurological disorders and presenting them to expert information will accelerate the interpretation of long-term EEG recordings.

Measuring the Effect of Filters on Segmentation of Developmental Dysplasia of the Hip.

BACKGROUND: Developmental dysplasia of the hip (DDH) can be detected with ultrasonography (USG) images. However, the accuracy of this method is dependent on the skill of the radiologist. Radiologists measure the hip joint angles without computer-based diagnostic systems. This causes mistakes in the diagnosis of DDH. OBJECTIVES: In this study, we aimed to automate segmentation of DDH ultrasound images in order to make it convenient for radiologic diagnosis by this recommended system. MATERIALS AND METHODS: This experiment consisted of several steps, in which pure DDH and various noise-added images were formed. Then, seven different filters (mean, median, Gaussian, Wiener, Perona and Malik, Lee, and Frost) were applied to the images, and the output images were evaluated. The study initially evaluated the filter implementations on the pure DDH images. Then, three different noise functions, speckle, salt and pepper, and Gaussian, were applied to the images and the noisy images were filtered. In the last part, the peak signal to noise ratio (PSNR) and mean square error (MSE) values of the filtered images were evaluated. PSNR and MSE distortion measurements were applied to determine the image qualities of the original image and the output image. As a result, the differences in the results of different noise removal filters were observed. RESULTS: The best results of PSNR values obtained in filtering were: Wiener (43.49), Perona and Malik (27.68), median (40.60) and Lee (35.35) for the noise functions of raw images, Gaussian noise added, salt and pepper noise added and speckle noise added images, respectively. After the segmentation process, it was seen that applying filtering to DDH USG images had low influence. We correctly segmented the ilium zone with the active contour model. CONCLUSION: Various filters are needed to improve the image quality. In this study, seven different filters were implemented and investigated on both noisy and noise-free images.