Clinical keratoconus (KCN) detection is a challenging and time-consuming task. In the diagnosis process, ophthalmologists must revise demographic and clinical ophthalmic examinations. The latter include slit-lamb, corneal topographic maps, and Pentacam indices (PI). We propose an Ensemble of Deep Transfer Learning (EDTL) based on corneal topographic maps. We consider four pretrained networks, SqueezeNet (SqN), AlexNet (AN), ShuffleNet (SfN), and MobileNet-v2 (MN), and fine-tune them on a dataset of KCN and normal cases, each including four topographic maps. We also consider a PI classifier. Then, our EDTL method combines the output probabilities of each of the five classifiers to obtain a decision based on the fusion of probabilities. Individually, the classifier based on PI achieved 93.1% accuracy, whereas the deep classifiers reached classification accuracies over 90% only in isolated cases. Overall, the average accuracy of the deep networks over the four corneal maps ranged from 86% (SfN) to 89.9% (AN). The classifier ensemble increased the accuracy of the deep classifiers based on corneal maps to values ranging (92.2% to 93.1%) for SqN and (93.1% to 94.8%) for AN. Including in the ensemble-specific combinations of corneal maps’ classifiers and PI increased the accuracy to 98.3%. Moreover, visualization of first learner filters in the networks and Grad-CAMs confirmed that the networks had learned relevant clinical features. This study shows the potential of creating ensembles of deep classifiers fine-tuned with a transfer learning strategy as it resulted in an improved accuracy while showing learnable filters and Grad-CAMs that agree with clinical knowledge. This is a step further towards the potential clinical deployment of an improved computer-assisted diagnosis system for KCN detection to help ophthalmologists to confirm the clinical decision and to perform fast and accurate KCN treatment.
The feature extraction step plays major role for proper object classification and recognition, this step depends mainly on correct object detection in the given scene, the object detection algorithms may result with some noises that affect the final object shape, a novel approach is introduced in this paper for filling the holes in that object for better object detection and for correct feature extraction, this method is based on the hole definition which is the black pixel surrounded by a connected boundary region, and hence trying to find a connected contour region that surrounds the background pixel using roadmap racing algorithm, the method shows a good results in 2D space objects.
Keywords: object filling, object detection, objec
This paper is concerned with introducing and studying the first new approximation operators using mixed degree system and second new approximation operators using mixed degree system which are the core concept in this paper. In addition, the approximations of graphs using the operators first lower and first upper are accurate then the approximations obtained by using the operators second lower and second upper sincefirst accuracy less then second accuracy. For this reason, we study in detail the properties of second lower and second upper in this paper. Furthermore, we summarize the results for the properties of approximation operators second lower and second upper when the graph G is arbitrary, serial 1, serial 2, reflexive, symmetric, tra
... Show MoreSteganography can be defined as the art and science of hiding information in the data that could be read by computer. This science cannot recognize stego-cover and the original one whether by eye or by computer when seeing the statistical samples. This paper presents a new method to hide text in text characters. The systematic method uses the structure of invisible character to hide and extract secret texts. The creation of secret message comprises four main stages such using the letter from the original message, selecting the suitable cover text, dividing the cover text into blocks, hiding the secret text using the invisible character and comparing the cover-text and stego-object. This study uses an invisible character (white space
... Show MoreRecognizing speech emotions is an important subject in pattern recognition. This work is about studying the effect of extracting the minimum possible number of features on the speech emotion recognition (SER) system. In this paper, three experiments performed to reach the best way that gives good accuracy. The first one extracting only three features: zero crossing rate (ZCR), mean, and standard deviation (SD) from emotional speech samples, the second one extracting only the first 12 Mel frequency cepstral coefficient (MFCC) features, and the last experiment applying feature fusion between the mentioned features. In all experiments, the features are classified using five types of classification techniques, which are the Random Forest (RF),
... Show MoreMany tools and techniques have been recently adopted to develop construction materials that are less harmful and friendlier to the environment. New products can be achieved through the recycling of waste material. Thus, this study aims to use recycled glass bottles as sustainable materials.
Our challenge is to use nano glass powder by the addition or replacement of the weight of the cement for producing concrete with enhanced strength.
A nano recycled glass p
The speaker identification is one of the fundamental problems in speech processing and voice modeling. The speaker identification applications include authentication in critical security systems and the accuracy of the selection. Large-scale voice recognition applications are a major challenge. Quick search in the speaker database requires fast, modern techniques and relies on artificial intelligence to achieve the desired results from the system. Many efforts are made to achieve this through the establishment of variable-based systems and the development of new methodologies for speaker identification. Speaker identification is the process of recognizing who is speaking using the characteristics extracted from the speech's waves like pi
... Show MoreMethods of speech recognition have been the subject of several studies over the past decade. Speech recognition has been one of the most exciting areas of the signal processing. Mixed transform is a useful tool for speech signal processing; it is developed for its abilities of improvement in feature extraction. Speech recognition includes three important stages, preprocessing, feature extraction, and classification. Recognition accuracy is so affected by the features extraction stage; therefore different models of mixed transform for feature extraction were proposed. The properties of the recorded isolated word will be 1-D, which achieve the conversion of each 1-D word into a 2-D form. The second step of the word recognizer requires, the
... Show MoreUltraviolet light radiation is applied to treat Plaque Psoriasis disease by targeted phototherapy. This is available through Narrowband-UVB light radiation devices peaked at wavelength 311 nm. Ten cases were chosen as a study group, 8 males aged 22-40 years old, and 2 females aged 25 and 32 years old who were exposed to ultraviolet light radiation. Their recovery or improvement was followed weekly. Different doses were used according to the severity of the lesion and as a trial for the outcome. The dose was given two times a week, starting with 200mJ/cm2, and subsequently increased by 100 or 200 mJ/cm2 reaching a maximum dose as tolerated by each individual patient. Improvement was observed after 4 – 6 weeks. The
... Show MoreDelays occur commonly in construction projects. Assessing the impact of delay is sometimes a contentious
issue. Several delay analysis methods are available but no one method can be universally used over another in
all situations. The selection of the proper analysis method depends upon a variety of factors including
information available, time of analysis, capabilities of the methodology, and time, funds and effort allocated to the analysis. This paper presents computerized schedule analysis programmed that use daily windows analysis method as it recognized one of the most credible methods, and it is one of the few techniques much more likely to be accepted by courts than any other method. A simple case study has been implement