31 research outputs found
Improved thresholding method for cell image segmentation based on global homogeneity information
No full textCell segmentation provides opportunity to
highlight abnormalities in the human body with a view to assist
medical experts to diagnose objectively. In order to achieve
this, a robust segmentation tool that gives high segmentation
accuracy is desirable. Cell images can be classified as
homogeneous and heterogeneous. Their existence in any of the
two categories is a function of how they are captured. This
however hinders the deployment of existing segmentation
models such as graph cut, Otsu thresholding, k-means and
watershed to cater for these categories of cell images. Our
contribution in this paper is to develop in the first instance a
segmentation model that automatically categorizes cell images
as homogeneous and heterogeneous. Secondly, based on a
category, a suitable and improved Otsu thresholding method is
proposed for cell segmentation. Experimental results on
heterogeneous cell images show improved segmentation
accuracy of 91.36% over that derived from traditional Otsu
thresholding (74%)
Investigating a Denoising Approach to an Improved Otsu Segmentation on Cell Images
No full textImage denoising provides an opportunity to minimize unwanted signal from any image in order to improve its interpretation by either human or machine. In the medical context, image denoising serves as a critical element of image processing as it helps to improve the quality of data presented for manual or automatic diagnosis. While there exist a number of image denoising methods such as the median, diffusion and Gaussian filtering, selecting a suitable one for cell segmentation may be challenging as one is tasked with ensuring adopted denoising method preserves critical object structures, like boundaries, while at the same time minimizing noise. In this paper, we discuss two popular denoising methods (diffusion filtering and Gaussian filtering) and investigates their significance, in improving the accuracy of segmented cell images, both individually and by their combinations. Experiment carried out on public and private datasets of cell images indicates an improved segmentation accuracy when cell images are first denoised with the combination of diffusion and Gaussian filtering as against individual denoising methods.</jats:p
A fast and non-trainable facial recognition system for schools
Get PDFDeep learning models have been at the forefront of facial recognition because they deliver improved classification accuracy over traditional ones. Regardless, deep learning models require an extensive dataset for training. To significantly cut down on its training time and dataset volume, pretrained models, have been used although, they are still required to undergo the usual training process for custom facial recognition tasks. This research focuses on an improved facial recognition system that lacks the training and retraining requirements. The system uses an existing deep learning feature extraction model. First, a user stands before a camera-enabled system. After that, the user supplies a unique identification number to fetch a corresponding face image from the database. This process generates two face feature vectors. One from the camera and that retrieved from the database. The cosine distance function determines the similarity value of these vectors. When the cosine distance value falls below a set threshold, the face is recognized and access granted. If the cosine distance of the two vectors gives a value above this threshold, access is denied. The proposed model performs satisfactorily on publicly available datasets
Using the ResNet-50 pre-trained model to improve the classification output of a non-image kidney stone dataset
Get PDFKidney stone detection based on urine samples seems to be a cost-effective way of detecting the formation of stones. Urine features are usually collected from patients to determine if there is a likelihood of kidney stone formation. There are existing machine learning models that can be used to classify if a stone exists in the kidney, such as the support vector machine (SVM) and deep learning (DL) models. We propose a DL network that works with a pre-trained (ResNet-50) model, making non-image urine features work with an image-based pre-trained model (ResNet-50). Six urine features collected from patients are projected onto 172,800 neurons. This output is then reshaped into a 240 by 240 by 3 tensors. The reshaped output serves as the input to the ResNet-50. The output of this is then sent into a binary classifier to determine if a kidney stone exists or not. The proposed model is benchmarked against the SVM, XGBoost, and two variants of DL networks, and it shows improved performance using the AUC-ROC, Accuracy and F1-score metrics. We demonstrate that combining non-image urine features with an image-based pre-trained model improves classification outcomes, highlighting the potential of integrating heterogeneous data sources for enhanced predictive accuracy
Youthhood
Get PDFTESTING-GROUND issue 03, Youthhood, examines worlds through youthful eyes, makes evident young ambitions, and questions how we can better empower young people to design cities, landscapes, and a planet that works for them. The issue includes contributions from: Carmel Keren, Jude Daniel Smith, Claire Edwards, Kazeem Kuteyi, Emmanuel Adarkwah, Reza Nik, Dan Cui, Kristofer Cullum-Fernandez, Fida Sassi, Simeon Shtebunaev, Daze Aghaji, Averill Dimabuyu, Sarri Elfaitouri, Rebecca McDonald-Balfour, and Ed Wall.
Rebecca McDonald-Balfour (Author), Jude Daniel Smith (Author), Daze Aghaji (Author), Carmel Keran (Author), Alexis Liu (Author), Dan Cui (Author), Kristofer Cullum-Fernandez (Author), Fida Sassi (Author), Averill Dimabuyu (Author), Ed
ADAPTIVE PARAMETER SELECTION FOR GRAPH CUT-BASED SEGMENTATION ON CELL IMAGES
Get PDFGraph cut segmentation approach provides a platform for segmenting images in a globally optimised fashion. The graph cut energy function includes a parameter that adjusts its data term and smoothness term relative to each other. However, one of the key challenges in graph cut segmentation is finding a suitable parameter value that suits a given segmentation. A suitable parameter value is desirable in order to avoid image oversegmentation or under-segmentation. To address the problem of trial and error in manual parameter selection, we propose an intuitive and adaptive parameter selection for cell segmentation using graph cut. The greyscale image of the cell is logarithmically transformed to shrink the dynamic range of foreground pixels in order to extract the boundaries of cells. The extracted cell boundary dynamically adjusts and contextualises the parameter value of the graph cut, countering its shrink bias. Experiments suggest that the proposed model outperforms previous cell segmentation approaches.</p
Leveraging graph cut’s energy function for context aware facial recognition in indoor environments.
No full textContext-aware facial recognition regards the recognition of faces in association with their respective environments. This concept is useful for the domestic robot which interacts with humans when performing specific functions in indoor environments. Deep learning models have been relevant in solving facial and place recognition challenges; however, they require the procurement of training images for optimal performance. Pre-trained models have also been offered to reduce training time significantly. Regardless, for classification tasks, custom data must be acquired to ensure that learning models are developed from other pre-trained models. This paper proposes a place recognition model that is inspired by the graph cut energy function, which is specifically designed for image segmentation. Common objects in the considered environment are identified and thereafter they are passed over to a graph cut inspired model for indoor environment classification. Additionally, faces in the considered environment are extracted and recognised. Finally, the developed model can recognise a face together with its environment. The strength of the proposed model lies in its ability to classify indoor environments without the usual training process(es). This approach differs from what is obtained in traditional deep learning models. The classification capability of the developed model was compared to state-of-the-art models and exhibited promising outcome
A sample-free Bayesian-like model for indoor environment recognition.
No full textVisual localization of indoor environments enables an autonomous system to recognize its current location and environment using sensors such as a camera. This paper proposes a method for visual recognition of indoor environments leveraging on existing object detection, ontology, Bayesian-like framework, and speeded-up robust features (SURF) algorithms. Objects detected in such an environment are fed into a Bayesian-like framework for domain recognition. Finally, the SURF localizes the predicted environment. One of the objectives of the proposed model is to eliminate the image-based training phase encountered in traditional place recognition algorithms. The proposed model does not rely on any visual information on the environment for training. Experiments are carried out on two publicly available datasets with promising results
Managing workplace violence in an outpatient mental health clinic
No full textWorkplace violence (WPV) is a complex and hazardous occupational problem healthcare workers face daily. The prevalence of WPV has increased exponentially in the healthcare industry in the past decade, specifically in outpatient mental health clinics. The increase is attributed to limited access to quality mental healthcare, poverty, substance abuse, and hereditary factors predisposing many individuals to mental health disorders. The purpose of this project is to determine the effect of WPV prevention training program and the use of violence risk assessment tools on the occurrence of patient-on-provider violence in outpatient mental health settings. Methodology: This project used a quasi-experimental design as the most suitable approach to assess the implementation of a violence risk assessment tool and WPV prevention training that targets incident of patient aggression. Results: Project results showed an increase in violence risk assessment post-implementation evidenced by a 53% increase in V-RISK-10 utilization. Also, there was a 36.4% decrease in patient-on-provider violent cases. Implications for practice: The result indicates an improved employees' workplace safety and violence prevention competency. It also provides a cost-effective way to manage type II violence.D.N.P.Includes bibliographical reference
African Sartorial Culture and the Question of Identity: Towards an African Philosophy of Dress
Get PDFThis paper is a critical interrogation of the apparel culture as a marker of African identity in traditional and contemporary Africa. The article philosophically discusses the sartorial culture of sub-Saharan Africans in the light of its defining elements, identity, and non-verbal communicative proclivities. Focusing on the Yoruba and the Ashanti people, the author argues that African dress expresses some symbolic, linguistic, and sometimes hidden, complex and immanent meaning(s) requiring extensive interpretations and meaning construction. With illustrative examples, he defends the position that the identity of some cultural regions in Africa can be grouped together based on the original, specific techniques and essence of dress that they commonly share. Against the present absence of an African philosophy of dress in the African sartorial culture and knowledge production, he argues the imperativeness of an African philosophy of dress, its subject matter, and connections to other cognate branches of African philosophy, and the prospects of such an ancillary African philosophy
