Center for Biomedical Engineering

Permanent URI for this college

http://etd.aau.edu.et/handle/123456789/45

Browse

Now showing 1 - 20 of 61

A Trilingual Android Application with Automatic Malaria Detection from Microscopic Images of Red Blood Cells
(Addis Ababa University, 2023-12) Berihun Nigussa; Dawit Assefa (PhD)
Malaria, which is a mosquito-borne blood disease caused by Plasmodium parasites, is one of the virulent infectious diseases affecting human beings and other animals since antiquity. Even though there were promising progresses in the reduction of malaria morbidity and mortality in the past two decades before the outbreak of COVID-19, the latest two reports of theWorld Health Organization (WHO) statistics indicate that malaria has been overlooked due to the COVID pandemics. Malaria is still prevalent specifically in low resource setting areas such as the sub-Saharan African countries, including Ethiopia. WHO reported that there were 229 million new cases of malaria and 409,000 deaths globally in 2019, alone. Whereas in the year 2021, the morbidity and mortality was reported to rise up to 247 million and 619,000, respectively. Timely diagnosis and treatment as well as good awareness about the disease play a major role to combat malaria. In the current project work, it was intended to design and develop a multi-lingual Android App that offers useful information about the malaria disease and is capable of automatically detecting malaria infected red blood cells (RBCs) from color microscopic images based on a deep learning approach. The Convolutional Neural Network (CNN) based deep learning model was trained, validated and tested on a publicly available dataset composed of microscopic images of RBCs taken from individuals with confirmed malaria infection as well as normal control groups. Experimental results generated from the deep learning model showed that the detection capability of the model achieved 100% training accuracy, 96% validation accuracy and 96% testing accuracy. The developed App avails useful information about malaria disease in general and tips users with fundamental information regarding its prevention and transmission mechanisms acting as an m-health system.
Application of CNN in Hypertensive Retinopathy Classification
(Addis Ababa University, 2022-01) Degel, Sharon; Dawit, Assefa (PhD)
Hypertension (HTN) as defined by the world health organization (WHO) is when the blood pressure is elevated with a systolic blood pressure equal to or greater than 140mmHg and a diastolic blood pressure equal to or greater than 90mmHg when measured on 2 separate days. HTN, according to Center for Disease Control and Prevention (CDC), raises the risk for stroke and other heart diseases. In a previous study, it was estimated that there were 972 million adults with HTN in 2000; 333 million in developed countries and 639 million people in developing countries. In a review of current trends by WHO, there has been an increase in adults with HTN from 594 million in 1975 to 1.13 billion in year 2015 with a prevalence of 27% in Africa and 18% in the America. Hypertensive Retinopathy (HPR) is damage to the retinal blood vessels caused by HTN. The use of artificial intelligence has recently been applied in Medical Imaging. One of such areas is in classification of eye diseases. In HPR, previous studies only focused on classification according to two classes; normal or HPR. In the current study, a classification method based on four classes according to the Wong and Mitchell classification which gives a classification based on the severity of HTN severity is presented; Normal, Mild, Moderate, and Severe or Malignant HPR. The study explored the idea of applying Convolutional Neural Network (CNN) in classifying HPR. The work focused on images collected solely from patients with HTN and will play an important role in assisting physicians to speed up the process of diagnosis. The study used the HPR features of the retina for classification. Transfer learning using existing pretrained models was employed in classifying the dataset. Five state-of-the-art pretrained models were employed for classification into two classes; normal and abnormal and into four classes: normal and three stages of HTR. The five different neural architectures namely: ResNet-101, GoogleNet, AlexNet, VGG-19 and Xception achieved an accuracy of 91.61%, 93.01%, 87.41%, 90.21% & 85.31% for the 4 Classification task and 100%, 100%,100%, 99.30% & 97.89% for the 2 Classification task respectively. The best accuracy was obtained using GoogleNet. The pretrained models show that with proper tuning of training parameters like the learning rate, number of epochs, batch size, type of
Application of Virtual Reality for Teaching in the Dissection Room
(Addis Ababa University, 2021-09) Alemseged, Getachew; Dawit, Assefa (PhD)
Virtual reality (VR) is a computer graphics technology which can be used to develop applications to support teaching human anatomy. Using VR, virtual 3D models of various parts of the human body can be developed and interacted with. The present thesis work addresses the problem that medical students often encounter in their practice in wet specimens with the absence of cadavers, using kidney as an example. The main objective of the thesis is to reconstruct highly detailed 3D virtual model of the kidney based on small interval cross- sectional images (CT-DICOM dataset), suitable for undergraduate students teaching virtual dissection through animation using kidney as an example, to supplement the wet specimen dissection procedure.. A 3D slicer software is used to develop 3D model of the kidney from the imported CT-DICOM data. Output of the 3D slicer is high poly object which is exported in an STL file format to the Transmutr software for simplifying the number of faces of the model by preserving the overall shapes of the original object which is used by Google sketch up & SimLab Composer software. Google Sketch Up is used for dissecting the kidney while SimLab Composer is used for rendering the 3D image and finally used for interactive VR viewing purpose. The final results of the developed 3D model of the kidney are used for interactive virtual reality viewing using desktop computer and virtually dissecting the kidney using Google sketch Up for visualization of the internal structure of the developed model. The developed 3D model allows students to perform detaching and interactively viewing a dissected kidney sequentially without the need of a supervisor. It can be also used inside the dissecting room to practice virtual kidney dissection by other medical practitioners.
Artificial Intelligence-Based Clinical Assistance Tool for COVID-19 Cases in Ethiopia
(Addis Ababa University, 2022-12) Mastewal Mathiwos; Dawit Assefa (PhD); Robel Kebede (PhD) Co-Advisor
The COVID-19 crisis has had a devastating impact in terms of loss of human life and economic disruption. According to a world health organization report on March 2022, more than 6.1 million people have died worldwide as a result of this pandemic. Early and precise recognition of COVID-19 is key for treating patients and slowing the spread of the pandemic. Several artificial intelligence (AI) based solutions have been developed to facilitate the application of chest X-ray (CXR) imaging and anthropomorphic data for use as a COVID-19 screening tool in resource-limited settings. The current study aims to develop a COVID-19 diagnosis scheme based on a deep learning (DL) approach applied on X-ray image samples collected locally in Ethiopia as well as a severity prediction tool based on a machine learning (ML) approach applied on anthropomorphic data collected from the same patients. The study data for the DL approach consisted of 746 CXR labeled images collected from St. Peter’s Specialized Hospital (SPSH) and Millennium COVID-19 care center (MCCC) in Ethiopia. The samples were composed of images acquired from patients treated for COVID-19 and normal controls. The DL model involved data preprocessing steps applied on the raw CXR images and was designed based on transfer learning approaches. The diagnostic prediction ability of the approach was measured in terms of probability score value and occlusion sensitivity map. The study data for the ML model development consisted of 308 anthropomorphic data (including demographic, comorbidity, and COVID-19 symptoms) collected from the MCCC to check the severity level of the COVID-19 cases and classify them into one of the three classes: moderate, sever or critical. A 5-fold cross-validation approach was used to train two popular ML approaches namely Support Vector Machine (SVM) and K-nearest Neighborhood (KNN) models. The DL method using ResNet-50 architecture achieved best classification performance with a validation accuracy of 94.20 % in accurately classifying COVID-19 and normal cases. The SVM model achieved a better prediction ability than K-NN with an overall accuracy of 89.9 % in predicting the severity status of the COVID-19 cases.
Automated Breast Cancer Detection using Computer Aided Diagnosis
(Addis Ababa University, 2018-05) Bruhtesfa, Mouhabaw; Dawit, Assefa (PhD)
Breast cancer is the most prevalent invasive cancer in women and stands second for chief cause of cancer deaths in women, next to lung cancer. The occurrence rate is exceeding in the developing countries though the rate of mortality has decreased which can be credited to the advances in diagnosis and treatment. Initial diagnosis involves histological observation (microscopic observation of cells/tissues) of affected breast tissues for structural changes, irregularities in cell shapes, distribution of cells in the tissue and determining the grade of the cancer. As manual interpretation of the tissues is often labor intensive, expensive and prone to errors and inconsistency, computer-based analysis of microscopic histopathology images is used as an alternative to provide a more accurate, automatic, fast and reproducible procedure to assess breast cancers. One important aspect in this regard is the automatic segmentation of breast cancers and several approaches are available in the literature for use in executing such tasks. However, the segmentation of major pathological structures and their subsequent follow– ups are not easy because of various artifacts such as presence of anatomical structures with highly correlated pixels with that of lesion, illumination variability, noise and magnification of the microscope. This thesis attempts to present a new mathematical scheme for analysis of color breast histopathology images acquired through digital microscopy or whole slide imaging. The proposed scheme uses a holistic representation of the color images in the three (trinion) space and applies trinion based Fourier transforms to extract useful imaging features for the purpose of classification and segmentation of histopathology images. A suitable color space transformation and a way of extracting robust higher order features are included in the method. The scheme has been applied in analyzing images acquired from standard histopathology image databases and results have shown that the algorithm achieved commendable results with 91% sensitivity, 92.7% specificity, and 92% overall accuracy.
An Automated Segmentation at Retinal Images for Use in Diabetic Retinopathy Studies
(Addis Ababa University, 2014-10) Moges, Daniel; Assefa, Dawit (PhD.)
Automated computer aided detection of retinal lesions associated with Diabetic Retinopathy (DR) offers many potential benefits. In a screening setting, it allows the examination of large number of images in less time and more objectively than traditional observer driven techniques. In a clinical setting, it can be an important diagnostic aid by reducing the workload of trained graders and other costs. However, the segmentation of major pathological structures and their subsequent follow–ups are not easy because of various artifacts such as presence of anatomical structures with highly correlated pixels with that of lesion, illumination variability, noise and movement of the eye during multiple visits by the patient. This study presents a novel mathematical scheme for analysis of color retinal images acquired through digital fundus cameras from patients treated for DR. The proposed scheme uses a holistic representation of the color images in the three (trinion) space and applies trinion based Fourier transforms to extract useful imaging features for the purpose of classification and segmentation of retinal images. A suitable color space transformation and a way of extracting robust higher order features are included in the method. The scheme has been applied in analyzing images acquired from standard retinal image databases. Results have showed that the algorithm achieved 86.06% sensitivity, 96.06% specificity, and 92.65% accuracy for pixel base segmentation of Hard Exudates (HEs) the most prevalent lesions that appears in the earliest stages of DR, while it achieved 96.67% sensitivity, 100% specificity and 97.3% accuracy for image base classification of abnormalities due to DR.
An Automated System Design for Medical Image Storage and Distribution
(Addis Ababa University, 2018-11) Abdela, Kemal; Dawit, Assefa (PhD)
The advent of information and communication technologies (ICT) and their incorporation into the medical domain especially medical imaging have created opportunities to enhance medical services and provide improvement to patient care. To implement such services, the current medical system needs to be integrated to different imaging modalities and at the same time be available to health professionals and patients. Picture Archiving and Communication System (PACS) is one means of storing and uses a server for image transmission through a network. The images could be acquired using any given modalities like Computed Tomography (CT), Ultra Sound (US) or Magnetic Resonance Imaging (MRI) and stored digitally. Digital Imaging and Communications in Medicine (DICOM) and Health Level Seven (HL 7) are often standards used to exchange medical images and patient data within the health institutions and outside of the institution using public network (internet). The images taken from the different imaging modalities are often difficult to transfer over the internet because of their size. We could think of for example storing and transferring an isotropic MRI data set between systems even for few patients. In order to transmit such slices of pictures via the internet, these images need to be compressed first. One of the primary obstacles in the development of a compression scheme is the loss of image resolution and contrast. However, we can improve the transmission time and also upload and download time. In the existing setup of the health facilities in Ethiopia, working medical image sharing systems are available only is some hospitals and function only within the hospitals. In low resource settings with chronic shortage of medical experts to examine the ever increasing amount of imaging data, there is always a need to locate a physician to consult and hence having a system that allows two ways communication remotely could be an enormous benefit. This thesis focuses on the development of a working Medical Image File Storage and Distribution System (MIFSDS) for use in storage and exchange/sharing of medical images within and across hospitals. The system includes development of an image classification algorithm and design of a web page system for image distribution and communication. The image classification algorithm which has been developed based on higher order statistical image feature extraction is shown to be accurate and robust while the system that has been developed for easy storage and transmission of imaging data allowing two way communication is proved to be simple and effective.
Automatic Detection of Malaria Parasite based on Microscopic Image Analysis
(Addis Ababa University, 2017-02) Bekele, Abebe; Demelash, Masreshaw(PhD)
Automatic Detection of Malaria Parasite based on Microscopic Image Analysis Abebe Bekele Addis Ababa University, 2017 Malaria is a serious global health problem and its diagnosis is usually done manually by compound light microscopy which is time consuming, tiresome and subjective. To support this manual method, in this master thesis, we designed and developed a system which is able to automatically detect plasmodium parasites from images of blood smears acquired by ourselves using a digital light microscope. In this method, blood smears taken from patients who were infected with plasmodium parasites were prepared. Digital images were then acquired by the light microscope and saved in the computer. Red blood cells (RBCs) are first segmented by marker control watershed algorithm, where the foreground markers are obtained from circular Hough transform and background markers from distance transform. The plasmodium infected RBCs are then detected in the Hue-Saturation-Intensity (HSI) color space. Thresholding on hue component of HSI color space is used to detect the chromatin dots of the parasite. Plasmodium falciparum and plasmodium vivax, the two dominant plasmodium species which cause the vast deaths in Ethiopia, are differentiated based on the size of infected RBCs. The performance of the proposed system for RBC segmentation, parasite detection and species differentiations was analyzed by comparing with the gold standard manual method for the total of 91 images of thin blood smears. The result shows that 97% of the RBC counts are similar to the gold standard with 97.5% sensitivity and 84.4% positive predictive value for plasmodium parasite detection at the cellular level. The species differentiations were done for each image with the accuracy level of 91.46%.The result showed the potential of the method for supporting the mass screening of malaria parasite. Keywords: Digital Microscope, Plasmodium, Thin Blood Smears, Watershed Algorithm, Circular Hough Transform, Distance Transform, Hue-Saturation-Intensity (HSI), Thresholding
Automatic Image Analysis for Bone Age Determination
(Addis Ababa University, 2022-12) Andualem Wube; Dawit Assefa (PhD); Asfaw Atinafu (PhD)
Bone age evaluation is commonly performed through radiological assessment of the skeletal development of the left hand, and then is compared with the chronological age. However, despite the time taking process, such image based automated processing remains incredibly challenging to implement. An accurate method is important to study the development of different wrist and pelvis structures which can predict age effectively. The proposed age determination scheme is composed of two major steps a watershed based segmentation scheme used to generate segmented images which are used as inputs to a deep learning algorithm that is able to determine the age of a given subject. The deep learning scheme utilized the InceptionV3 architecture for its implementation. The model was trained, validated, and tested on radiological hand images acquired from the RSNA database. The model resulted in a mean square error of 4.4 months when compared against the available ground truth information. Overall, the results showed that the proposed bone age determination scheme comes with great promises.
Brain Tumor Detection Based on Magnetic Resonance Image Analysis
(Addis Ababa University, 2018-01) Amare, Ambaw; Dawit, Assefa (PhD)
Automatic detection of brain tumors based on magnetic resonance (MR) image processing has been developed in this thesis. Improving the ability to accurately identify early-stage tumors is important goal for physicians, because early detection of brain tumors is a key factor in producing successful treatments. In this regard, an automatic brain tumor detection and segmentation framework has been proposed in this thesis work based on contrast enhanced T1 weighted (T1-W) images acquired from a cohort of patients with confirmed high grade brain tumors. Gray scale T1-W images have been represented in the three component Trinion space and Trinion Fourier transform has been applied aiming to extract useful features that could be used to automatically detect and segment brain tumors from their surrounding background. The performance of the proposed scheme has been evaluated by comparing its segmentation outputs with the ground truth information (based on manual contours by radiologists) that came with the MR data set. Results have showed that the algorithm achieved 99.6% sensitivity, 100% specificity, and 99.8% accuracy for pixel based segmentation while it achieved 91.5% sensitivity, 90% specificity and 90.5% accuracy for image based classification of tumors.
Cepstral Analysis of Wideband Ultrasound
(Addis Ababa University, 2021-03) Eyob, Adugnaw; Dawit, Assefa (PhD)
Ultrasound is a pressure wave with frequency beyond20 �� . Wideband ultrasound is ultrasound pulse with mega range of frequency components. Most of the currently available ultrasound transducers are narrow banded and fail to provide better axially resolved images for tissues with multiple layers. The methodology used in this thesis models multi-layered biological medium based on linear acoustics of pulse-echo detection principle, normal incidence, longitudinal ultrasonic propagation and investigates the feasibility to detect periodicity. Ultrasound propagation data was generated synthetically and subjected to cepstral analysis to detect periodicity in a multilayered skin tissue model. Rectangular pulses with a center frequency of 16-30 MHz and pulse duration of 1.2 × 10−7s and 0.8 × 10−7s were applied to a 4-layered medium. The output is measured and cepstral analysis was applied to determine the feasibility of periodicity detection. For layers separated by equal thicknesses, the cepstral peaks existed at equal intervals where as in the case of different layer thicknesses, peaks existed at integer multiple of the thinnest layer thickness. In cepstral analysis, when the cepstral peaks exist at equal intervals or integer multiples of the shortest time of flight (time of flight to the thinnest layer), periodicity detection is guaranteed. It is concluded that periodicity is detected with wideband ultrasound pulses and the minimum and maximum bandwidth are determined based on the duration of the pulse. The possible limitations with the thesis are the assumption of normal incidence planar waveforms, linear ultrasound propagation, and parallel surfaces.
Characterization and Efficiency Test of Affordable and Ecofriendly Sanitary Pad Made of Natural Fibers from Enset
(Addis Ababa University, 2020-12) Semira, Abdela; Kim, Gyeong-Man (Prof.); Dawit, Assefa (PhD) Co – Advisor
Sanitary pads are one of menstrual hygiene management (MHM) materials that are used by girls and women to absorb their menstruation. However, these products are mainly available in urban area and also expensive for the majority underprivileged girls of our country. Thus, most girls specially the rural girls are forced to use unhygienic and uncomfortable materials to deal with their period. Hence, these girls are forced to live stressful, uncomfortable and unsafe life. These result in a long chain of negative impacts on the health of our girls and women. On the other hand, the materials that are used to fabricate sanitary pads create environmental contamination during disposal. This research aims at providing a better alternative to conventional sanitary pads using sustainable and convenient raw materials like cotton, Enset pulp and a bioplastic to make sanitary pads with a needed performance. By converting Enset fiber to pulp the major part of the sanitary pad, an absorbent core, is produced. Then, cotton fabric that is commonly known by the name of ‘Nethla’ is used as a top sheet during sample pad production and finally the bioplastic is used as the bottom layer to make sample sanitary pad. Results showed that the sample sanitary pad meets all the required criteria such as absorbency, ability to withstand pressure after absorption, pH, wicking property, liquid striking property, fluid retention, disposability and physical parameters (i.e. pad length, width and thickness). Also the research revealed that the Enset pulp did not have antimicrobial activity. The cost estimation of the sample sanitary pad was done and the net cost for single pad production was found 0.85 birr. As compared to conventional sanitary pads that are available in the local market, the pad designed in the current thesis is affordable, sustainable and ecofriendly with the potential to replace not only the pads that are imported from abroad but also the raw materials needed to produce them locally.
Characterization of Antibubbles Response under Sonication In Terms of Size Distribution and Destruction Threshold for Use in Drug Delivery
(Addis Ababa University, 2023-06) Mahider Yifru; Dawit Assefa (PhD); Michiel Postema (PhD) Co-Advisor
Antibubbles are gas bubbles containing liquid droplets. Due to their acoustic property, they are preferable for different medical applications such as drug delivery and ultrasonic harmonic imaging. In drug delivery, they are used as drug carrier for effective treatment of diseases such as cancer. In order to ensure safe and targeted drug delivery, the behaviors of the drug carriers (antibubbles) need to be studied. The aim of the current research was to study response of antibubbles under sonication and determine destruction threshold which can lead to fragmentation under ultrasound pulse having safe mechanical index (MI) with reduced undesired mechanical and thermal bio-effects. Gray-scale video frames of antibubbles were used for this research, which were generated through laboratory experiment. Matlab and ImageJ platforms were used for effective detection of antibubbles and determination of their respective size before, during and after sonication together with destruction threshold. The proposed algorithm involves extraction of texture features form the video frames based on the local gray level co-occurrence matrix (GLCM) and a watershed scheme for segmentation of the antibubbles. Once antibubbles are detected, their size distribution under ultrasound pulse with varying acoustic pressures was analyzed. The investigation carried out under sonication of low pressure ultrasound waves (0.2 MPa and 0.4 MPa) showed that the antibubbles oscillate repeatedly with series of stable contraction and expansion but will not undergo fragmentation. This property of antibubbles interaction with low pressure ultrasound wave makes them very useful and preferable to be used as a drug carrier in a way of manipulating the antibubble movement through the blood vessel to the target region. In case of antibubble sonication under higher acoustic pressure wave (0.6 MPa and 1 MPa), their oscillations become more asymmetric which finally lead to fragmentation. The MI of 0.6 MPa and 1Mpa ultrasound pulses are 0.6 and 1 respectively, where significant risk of cavitation is considered for MI > 0.7. The destruction threshold analysis showed that antibubbles under acoustic pressure of 0.6 MPa experience fragmentation when their size during maximum expansion is approximately twice that of their initial size intermes of area, i.e. Amax ≈ 2.1 Ai. In the case of 1 MPa acoustic pressure Amax ≈ 2.8 Ai, showing that ultrasound pulse with acoustic pressure of 0.6 MPa is considered safe compared to 1 MPa for use in drug delivery with the capacity to induce fragmentation and effective targeted drug release.
Classification of Fatty Liver Disease Using Deep Learning Technique
(Addis Ababa University, 2023-08) Hume Degebassa; Dawit Assefa (PhD); Million Molla (Mr.) Co-Advisor
Fatty liver disease (FLD) also termed steatosis liver disease can be classified into two broad spectrums: Alcoholic fatty liver disease (AFLD) and Non-alcoholic liver disease (NAFLD). AFLD is a type of liver disease associated with alcohol intake of 40 to 80 grams of ethanol/day for males and of 20 to 40 grams/day for females. NAFLD is a disorder characterized by excess accumulation of fat (in the form of triglycerides) with an amount >5% in the hepatocytes. The progression of the disease is manifested by ongoing inflammation and consequent steatosis grade: G1 (mild steatosis), G2 (moderate steatosis), and G3 (severe steatosis). If no early strategies are adopted, FLD can progress to steatohepatitis (SH) which is a risk factor for liver fibrosis, cirrhosis, and hepatocellular carcinoma (liver cancer). Cirrhosis is currently the 11th most common cause of death globally and liver cancer is the 16th leading cause of death. In Ethiopia, cirrhosis was the 7th leading cause of mortality accounting for 24 deaths per 100,000 populations in 2019. Hence, early diagnosis is important to avoid advanced stages of FLD. In order to diagnose and grade FLDs, Ultrasound imaging is the most well-known technique. The visual assessment of these Ultrasound images is not only time taking and labor-intensive task but it is also prone to inter-observer variability. That calls for the development of an automated system that overcomes subjectivity and inconsistency in the screening process. Different studies in the literature have proposed the classification of FLDs using deep learning and machine learning techniques. However, most of these techniques are limited to binary classification and don’t consider the severity level. The current study aims to develop an automatic fatty liver classification and grading scheme using deep learning techniques. A total of 550 liver ultrasound images were collected from Zenode repository and image pre-processing was employed on the acquired images before the images were used to train the deep learning model. The effect of important deep network hyper parameters including batch size, learning rate, regularizer and epoch was investigated. Different pre-trained deep learning networks were tested for their classification efficacy including VGGNet, MobileNetV2, ResNet and Xception based on useful performance matrices. Accordingly, the Xception model was found to outperform the rest for multiclass (normal, mild, moderate, and severe) fatty liver classification. It offered a precision of 96.25%, recall of 97.75%, F1 score of 96.75%, and overall accuracy of 96.36% showing the great promises of the model.
Color Medical Image Edge Detection based on Higher Dimensional Fourier Transforms Applied in Diabetic Retinopathy Studies
(Addis Ababa University, 2019-04-25) Abebe, Belachew; Dawit, Assefa (PhD)
Various edge detection techniques for color images that have been proposed in the last two decades showed that color images contain 10% additional edge information as compared to their gray scale counterparts. For color image edge detection, the traditional methods used for grayscale images are usually extended and applied to the three-color channels separately. This leads to lose the intrinsic inter-correlation information embedded in color image components in addition to computational complexity incurred. Efficient and accurate edge detection leads to increased performance of subsequent image processing and analysis techniques including image segmentation and quantification. In this thesis, an edge detection algorithm has been proposed that treats color value triplets as vectors based on higher dimensional algebra. A human perception-based color space has been used due to its importance in color image edge detection. The trinion based algorithm has provided an efficient method to represent the color information vectorally. Color edge features are extracted based on a second order statistical technique using the weibull distribution method. A suitable color space transformation and a way of extracting robust higher order features are included in the method. Performance of the proposed scheme is compared against classical and other vectorial approaches which have been proposed in the literature based on objective criteria. Results showed that the proposed approach wins the other techniques available in the literature. Application of the proposed method has been shown in edge detection applied on color images taken from patients treated for diabetic retinopathy acquired from publicly available databases and St. Paul’s Hospital Millennium Medical College. The algorithm performs well in detecting exudates, hemorrhages, optical disc and blood vessels.
Computer Aided Diagnosis System for Melanoma Lesion Detection
(Addis Ababa University, 2018-04) Endalkachew, Wolde; Dawit, Assefa (PhD); Mengistu, Kifle (PhD) Co-Advisor
Detection of skin cancer in the earlier stage is very critical. Nowadays, skin cancer is seen as one of the most hazardous form of cancers found in humans. The most common types of skin cancers are Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC) and Melanoma. Among these melanoma is the deadliest type of skin cancer. The detection of Melanoma cancer in early stage can be helpful to cure it. Computer vision plays an important role in medical image diagnosis and this has been proved by many existing systems. A computer aided skin image diagnosis system has a significant potential for screening and prognosis. It is utmost important in countries where unaided visual diagnosis system is the practice and where there is insufficient number of dermatologists. This thesis presents a computer aided diagnosis system for the detection of melanoma skin cancer using a novel mathematical scheme. The proposed method uses a holistic representation of skin color images to extract useful features for use in effective segmentation of melanoma lesions. The segmentation scheme is preceded by a processing stage composed of noise filtering color space transformation. Vertex component analysis and principal component analysis are integrated to form a hybrid approach for unmixing and feature dimension reduction. An optimized feature selection technique is also used to obtain the best achievable performance in effectively detecting the lesions. Support vector machine (SVM) along with geometrical and color feature threshold values are integrated to make effective detection of melanoma lesions. The effectiveness of the developed scheme to classify lesions into benign, suspicious and malignant melanoma is found to be promising. The proposed scheme has been tested on images taken from standard dermoscopy image databases and achieved 96.4% sensitivity, 99.4% specificity, and 97.9% overall accuracy for pixel based classification of melanomas while it achieved 98% sensitivity, 100% specificity, and 98.6% accuracy for image based classification justifying it‟s great promises.
CT Based Lung Cancer Detection Using Spatially Localized Integral Transforms
(Addis Ababa University, 2018-06) Abel, Belay; Dawit, Assefa (PhD)
Lung cancer is the leading cause of death among other cancer types. Its survival rate is very limited unless diagnosed/detected early. As a result, early detection is imminent for minimizing the deaths caused by lung cancer. For realizing the early detection and early diagnosis, sophisticated and complex imaging modalities like low dose CT are often utilized. However, the big problem lays ahead on the image interpretation. Due to different factors like image quality, suppressed image features in the spatial domain, radiologists eye sight and radiologists’ expertise level, misdiagnosis and error are the major difficulties to overcome. In that regard, many research works have been reported in the literature to deal with the image interpretation issues. Among other researches, CADe and CADx are considered the most remarkable methods for early detection and diagnosis. However, most of the CADe and CADx systems are not clinically implemented. This is because of their inefficiency, inaccuracy and non-robustness. Aiming for a more robust and accurate detection of lung cancers, a new computer aided scheme has been suggested in this thesis. The scheme implements a rotation invariant, joint space-frequency localized integral transform together with spatial image enhancement for feature extraction of CT lung images. The algorithm has been implemented on a Matlab platform (Matlab 2013a) and validated on CT lung images acquired from TCIA database. Results showed that the algorithm achieved a sensitivity of 97.1%, specificity of 83.33% and overall accuracy of 96.68% in detecting lung cancers showing its great promise.
Customizing the Pressure-based Spirometer for Improving its Accuracy for Chronic Obstructive Pulmonary Disease Tests
(Addis Ababa University, 2021-11) Ashebir, Worku; Masreshaw, Demelash
Breathing is a fundamental and essential process for the proper functioning of human being. It is a process of supplying oxygen to the body and removing of waste carbon dioxide from the body. This process takes place through the respiratory system and organs involving the nose, pharynx, trachea, bronchi, and lungs. There are several diseases and conditions that can affect the respiratory system. Chronic obstructive pulmonary disease (COPD) is the most common one that arises from airway obstructions. A differential pressure-based spirometer is the most dominant device used to perform pulmonary function test in order to diagnose and monitor COPD. This pressure-based spirometer acquires a signal through a pressure difference between two points of the patient breathing tube (PBT) created by the patient against the atmospheric pressure. This means as the COPD patient finds it hard to breath, the differential pressure created by the patient across PBT reduces. The detection sensitivity of the PBT and altitude correction are the main important parameters which need to be considered while designing a differential pressure-based spirometry. So the aim of this thesis is to customize a differential pressure-based spirometer device with respect to altitude in order to improve accuracy of COPD tests. In this thesis, LabVIEW software has been used to generate and analyze the measured signal, and customize the spirometer with respect to altitude. Moreover, Bernoulli’s equation has been used to calculate air flowrate and forced vital capacity of the lung which are used to grade COPD test result. Finally, the designed PBT is simulated using Ansys 19.2 software and a prototype is created using 3D printer. The flowrate vs time, and volume vs time graphs were obtained and displayed using LabVIEW. The result shows that the accuracy of COPD test can be improved by increasing the sensitivity of PBT and customizing a pressure-based spirometer using the altitude correction factor during designing and operating the pressure-based spirometer.
Decision Support System for Medical Equipment Standardization
(Addis Ababa University, 2020-06) Tadesse, Minalku; Masreshaw, Demelash (PhD); Mengistu, Kifle (PhD) Co-Advisor
Introduction: Healthcare technology in general and medical equipment in particular is vital for the healthcare provision. However, today’s medical equipment market competition paved a way for existence of lack of Medical Equipment Standardization (MES) in hospitals. Consequently, decision making in MES and managing of medical equipment appropriateness become complex practices. Similarly, although the exact problem in Ethiopia is not yet known, only 72% and 50% of medical equipment in Addis Ababa and regional public hospitals respectively are functional which raise equipment appropriateness issues. Moreover, my field observation helped me see professionals who complained about lack of MES. But no attention is given in Ethiopia, and evidences around investigating the needs, challenges, practices and requirements in MES decision making process are limited. So this research aims to investigate the impact of lack of MES in equipment appropriateness and develop a novel decision support system for MES decision making process. Methodology: Mixed study was applied. For this survey, 457 health professionals from 5 Federal hospitals and 6 hospitals from Addis Ababa Health Bureau participated during survey between March and April 2018. To strength this, 4 biomedical engineers (1 EFDA, 3 EPSA) were also interviewed. To analyze the collected data Statistical Package for Social Sciences (SPSS) software version 23 was used. Input-output (IO) approach and sequential water fall model was used to organize and develop the system respectively. Requirements were assessed and validated before the system was developed. Then C-sharp and SQL were used as a programming language. At the end, this novel system was simulated using analytical model and tested using hypothetical values. Results and Conclusions: Descriptive test result of survey indicated that lack of medical equipment standardization has an impact on medical equipment appropriateness. The cross-tabulation test also supports this and on average more than 257 (72.3%) participants agreed on lack of MES existence and its impact on equipment appropriateness. Similarly, chi-square test result also indicated that, there is a statistically significant relationship between existence of lack of MES and its impact on medical equipment appropriateness. Our findings are in agreement with the WHO findings which stated that 30-50% of world economy is wasted by extra spare part and maintenance requirement resulted from lack of MES. In addition based on survey result, physicians’ preference, manager-supplier relationship, public procurement law, negative attitude, lack of communication and collaboration are major challenges in MES. Moreover, 337 (94.7%) participants reported they have no system for MES standardization has an impact on equipment appropriateness and supports the need of a new system that facilitates MES decision making process. The overall survey result indicated that lack of medical equipment standardization has an impact on equipment appropriateness and supports the need of a new system that facilitates MES decision making process. Following this, a novel system was developed and tested. The system output can support decision makers in MES decision making process.
Deep Learning-Based Murmur Detection and Murmur Characteristics Classification from Phonocardiogram
(Addis Ababa University, 2025-05) Hafiza Moges; Melkamu Hunegnaw (PhD)
CVD is the main cause of death worldwide. The World Heart Federation reported a 20.5 million death from CVD in 2021, by 2030 this number is expected to rise to 23.6 million, and more than 75 percent of CVD death occur in low and middle income nations, as they have less access for equitable health care services. Phonocardiograph (PCG) is an affordable and portable instrument which can record, play back, and provide a visual display for heart sounds. Blood flow that is turbulent may cause the cardiac valves to vibrate enough to produce murmurs, which produces audible heart sounds and abnormal waveforms in the PCG. Most risk factors can be significantly reduced by early diagnosis. Although, PCG signals are useful for detecting heart murmurs it needs a trained medical professional for its interpretation. This study proposes a hierarchical multi-scale convolutional neural network (HMS-Net) based automatic murmur detection and murmur characteristics classification using the publicly available CirCor DigiScope PCG dataset. Previous research studies mainly focus on classifying heart sounds as either murmur vs. normal or detecting a limited type of valvular heart diseases (VHDs) but this study extends it by designing a pipeline that first determines whether a murmur is present or absent; not only these but also it advances the analysis by further classifying five key murmur characteristics: timing, shape, pitch, grade, and quality, which is a significant step beyond traditional binary murmur detection. For the training and evaluation, it adopts the HMS-Net model developed by the winner of the George B. Moody PhysioNet Challenge, with a few important adjustments. In the original work, the dataset included an ’unknown’ murmur class due to expert uncertainty to label them as present or absent, so they used a quality assessment method for this class. In this study, the ’unknown’ class is excluded, as it does not represent a valid clinical category, so that this study focused on clinically meaningful labels. Additionally, a separate prediction pipeline is designed to detect murmur and simultaneously predict the five characteristics. The core model architecture and preprocessing steps remain unchanged. Each task are trained and evaluated independently using metrics such as accuracy, weighted accuracy, F1 score, AUROC, and AUPRC. The proposed model achieved an accuracy of 93.1%, and F1 score of 86.9% for the murmur detection task. For the murmur characteristic tasks, the model achieved accuracy scores between 74.3% and 81.5%. This goes a step beyond simply detecting whether a murmur is present to identifying the key five murmur characteristics. Identifying these characteristics is very important because it helps to determine the type of VHDs, which makes the model’s output more clinically useful. These results are promising, considering the small data and class imbalance. These results suggest that the proposed system can support early screening to improve patient referrals to specialized care with more detailed murmur information of the patient, which could help to identify the underlying VHDs so that it helps patients to get early diagnosis and timly treatment. It will be very helpful to reduce earlier deaths due to VHDs in resource limited communities.

Browse

Browsing Center for Biomedical Engineering by Title

Results Per Page

Sort Options