Agnieszka Stankiewicz

The effectiveness of the classification process using deep neural networks depends on the correct preparation of data sets. In this article, the correctness of seven versions of the LOCT B-scan OCT dataset (which can be found in public data repositories) was verified using SHA encoding. Furthermore, the correctness of the classification process for three most representative versions of these datasets was analyzed. We have observed that in some repositories, the validation and test parts of the data included some data from the training part. Therefore, the experimental results in many articles are not correct, and the effectiveness results presented in various publications are surprisingly high. Classification experiments were performed on the original version with duplicates, its extended variation, and on a cleaned version. Deep neural network training was performed using well-established neural network architectures used in medical image classification. The conducted research indicates that the results of many scientific publications should be re-evaluated using data that does not contain duplications between the training and test sets.

The aim of this study was to develop a mobile application for Android devices dedicated to the classification of pathological changes in human eye optical coherence tomography (OCT) B-scans. The classification process is conducted using convolutional neural networks (CNNs). Six models were trained during the study: a simple convolutional neural network with three convolutional layers, VGG16, InceptionV3, Xception, Joint Attention Network + MobileNetV2 and OpticNet-71. All of these models were converted to TensorFlow Lite format to implement them into a mobile application. For this purpose, three models with the best parameters were chosen, taking accuracy, precision, recall, F1-score and confusion matrix into consideration. The Android application designed for the classification of OCT images was developed using the Kotlin programming language within the Android Studio integrated development environment. With the application, classification can be performed on an image chosen from the user’s files or an image acquired using the photo-taking function. The results of the classification are displayed for three neural networks, along with the respective classification times for each neural network and the associated image undergoing the classification task. The mobile application has been tested using various smartphones. The testing phase included an evaluation of image classification times and score accuracy, considering factors such as image acquisition method, i.e., camera or gallery.

Deep learning solutions can be used to classify pathological changes of the human retina visualized in OCT images. Available datasets that can be used to train neural network models include OCT images (B-scans) of classes with selected pathological changes and images of the healthy retina. These images often require correction due to improper acquisition or intensity variations related to the type of OCT device. This article provides a detailed assessment of the impact of preprocessing on classification efficiency. The histograms of OCT images were examined and, depending on the histogram distribution, incorrect image fragments were removed. At the same time, the impact of histogram equalization using the standard method and the Contrast-Limited Adaptive Histogram Equalization (CLAHE) method was analyzed. The most extensive dataset of Labeled Optical Coherence Tomography (LOCT) images was used for the experimental studies. The impact of changes was assessed for different neural network architectures and various learning parameters, assuming classes of equal size. Comprehensive studies have shown that removing unnecessary white parts from the input image combined with CLAHE improves classification accuracy up to as much as 4.75% depending on the used network architecture and optimizer type.

The use of neural networks for retinal vessel segmentation has gained significant attention in recent years. Most of the research related to the segmentation of retinal blood vessels is based on fundus images. In this study, we examine five neural network architectures to accurately segment vessels in fundus images reconstructed from 3D OCT scan data. OCT-based fundus reconstructions are of much lower quality compared to color fundus photographs due to noise and lower and disproportionate resolutions. The fundus image reconstruction process was performed based on the segmentation of the retinal layers in B-scans. Three reconstruction variants were proposed, which were then used in the process of detecting blood vessels using neural networks. We evaluated performance using a custom dataset of 24 3D OCT scans (with manual annotations performed by an ophthalmologist) using 6-fold cross-validation and demonstrated segmentation accuracy up to 98%. Our results indicate that the use of neural networks is a promising approach to segmenting the retinal vessel from a properly reconstructed fundus.

W artykule zaprezentowano kurs Podstawy Elektroniki opracowany dla kierunku Automatyka i Robotyka oraz sposób jego realizacji w trybach nauczania zdalnym i hybrydowym. Opisano zmiany wprowadzone przez autorów w ramach wykładów i laboratoriów, a także wykorzystane narzędzia wspomagające prac zdalną. Dodatkowo zestawiono kilkuletnie wyniki ocen przedmiotu ze strony studentów i rezultaty egzaminacyjne, porównując trzy tryby nauczania: stacjonarny, zdalny i hybrydowy. Wyniki wskazują, że odpowiednia adaptacja i wykorzystanie nowoczesnych możliwości narzędzi wspomagających nauk samodzielną przynoszą pozytywne efekty, a nauczanie zdalne nie musi oznacza pogorszenia efektów uczenia się.