The management of medical equipment raises a range of complex problems including those associated... more The management of medical equipment raises a range of complex problems including those associated with replacement processes. One of the most significant challenges is to identify a proper list of medical equipment that requires replacement and then to optimize this list. In this article, we present a new approach to solve this problem by integrating Quality Function Deployment (QFD) and Genetic Algorithm (GA) in one framework. In a previous application, QFD has proven its validity to solve the priority problem; meanwhile GA is an optimization method. Hence, the proposed model, QFD-GA, was developed to prioritize the medical equipment for replacement process taking into account a set of criteria; in addition, the prioritized list is optimized according to the available budget of the hospital to maximize the number of replaced devices. The validation of the proposed model was carried out on sixty devices that include different types of medical equipment in one public hospital. Results show that the proposed model can efficiently classify the priority into four subcategories, and simultaneously maximize the number of medical equipment to be replaced considering the budget constraint.
Photodiagnosis and Photodynamic Therapy, Mar 1, 2021
SIGNIFICANCE The estimation of tissue oxygenation is vital in the diagnosis and therapeutic evalu... more SIGNIFICANCE The estimation of tissue oxygenation is vital in the diagnosis and therapeutic evaluation of a huge assortment of diseases. However, the hyperspectral (HS) imaging system is a rising innovation that can be utilized to build a highly sensitive, non-invasive, and tissue hemoglobin immersion map. OBJECTIVE As a result of the urgent need to design and implement early detection devices and applications for the COVID-19 pandemic, we propose building a non-invasive custom optical imaging system to assist with phlebotomy and vascular approach to survey the reliability of blood oxygen saturation (SpO2) levels recovered from spectral images. MATERIALS AND METHODS HS images were gathered from 15 healthy subjects without previous medical history complications and with an average age range of 20 to 38 years, who were undergoing phlebotomy. The forearm was vigorously illuminated utilizing an HS camera with polychromatic source light of spectrum range (400∼980 nm). Spectroscopic reflectance images were caught by a focal plane exhibit for the region of interest (ROI). Then the custom algorithm comprising normalization and moving average filtering for noise removal was applied, followed by K-mean clustering for image segmentation to visualize and highlight the arteries and the veins in the investigated forearm. RESULTS The investigations show that after normalization of the recorded signal from the HS camera of the participating subjects it was noticed that at wavelength of 460 nm the oxygenated arteries had a stronger signal than the de-oxygenated veins, and at a wavelength of 750 nm the de-oxygenated veins had a stronger signal than the oxygenated arteries. Thus, the ideal wavelength to reveal the oxygenated arteries was 460 nm, and the ideal wavelength to reveal the de-oxygenated veins was 750 nm. CONCLUSIONS HSI is a prospective technique to assist with phlebotomy and non-contact oxygen saturation approach. Additionally, it may permit future surgical or pharmacological intercessions that titrate or limit ischemic injury continuously.
Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death... more Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death of one third of the world’s population. The main cause of cardiovascular diseases is atherosclerosis which makes arteries less elastic (called ‘‘hardening of the arteries” or ‘‘arterial stiffness’’). The optical Micro Electro Mechanical System (MEMS) pressure sensor has shown its potential in the diagnosis of arterial stiffness that can be conducted by detecting the pulse pressure in the radial artery. In this paper, we attempt to improve the sensitivity of micro-diaphragm deflection in optical Micro-electromechanical System (MEMS) sensors as applied to pulse pressure detection, thus aiming to determine the safety of a person’s measured pulse of cardiovascular disease. The deflection sensitivity improvement was evidenced using Finite Element Analysis ANSYS software. Corrugation for periphery-clamped silicon nitride (Si3N4) micro-diaphragm based on the variation of the diaphragm thicknes...
The technology of next generation sequencing brought about evolution in research that based on se... more The technology of next generation sequencing brought about evolution in research that based on sequence which replaces the microarray due to its advantages. The RNA-Seq is a high-throughput gene expression data that uses NGS technologies. The problem of high-throughput RNA-Seq datasets is the high of dimensionally (variables > observations) that need to reduce their dimensions to predict and classify. In this work, ensemble-based on the approach of selection the feature is proposed to choose small optimal genes from various RNA-Seq cancer datasets. The approach combines two filters to select the features (SNR and t-test). In addition, SVM-RFE is used as an embedded feature selection. The SVM classifier used for validating and testing the selected genes using the accuracy measure. The results show the ability to select small optimal genes with high accuracy. Consequently, the genes which are selected will be used as biomarkers to diagnose cancer.
2010 5th Cairo International Biomedical Engineering Conference, 2010
Abstract Observation of maintenance programs in healthcare facilities is an essential requirement... more Abstract Observation of maintenance programs in healthcare facilities is an essential requirement for almost all equipment in order to guarantee its performance, prevent sudden failures and to extend the equipment's life expectancy. In this work availability, reliability and performance efficiency were used for equipment performance judgment, in order to support the clinical engineer's decision making process. Time series models were the suggested technical tools used for prediction. Two models were used in this study. Raw data were ...
Technology in its own right is now universally recognized as the engine of economic growth. Techn... more Technology in its own right is now universally recognized as the engine of economic growth. Technological innovation generates increases in productivity, keeps companies competitive, and ultimately enables increases in the nation’s standard of living. It is precisely through doing a better job of managing technology that an improved standard of living can be created for all citizens. Unlike manufacturing companies that produce particular goods, hospitals must offer the highest quality of care possible to their customers, the patients and their families. Although manufacturers must produce high quality products and/or services (to remain in business), the health care industry’s failure to offer quality care, can lead to life threatening consequences. Repair and PM are important to the success of any clinical engineering program. Since the goal of the clinical engineering department is to provide a safe and effective for delivery of the best health care
Modern DNA sequencing instruments are able to generate huge amounts of genomic data. Those huge v... more Modern DNA sequencing instruments are able to generate huge amounts of genomic data. Those huge volumes of data require effective storage, fast transmission, provision of quick access to any record, and superior functionality. Data storage costs have an appreciable proportion of total cost in the creation and analysis of DNA sequences. In particular, the increase in the DNA sequences is highly manageable due to a tremendous increase in the disk storage capacity. Standard compression techniques failed to compress these sequences. Recently, new algorithms have been introduced specifically for this purpose. In this paper, we comparatively survey the main ideas and results of lossless compression algorithms that have been developed for DNA sequences.
Clinical Engineering Department in hospitals need to take decision for Health Care Technology ass... more Clinical Engineering Department in hospitals need to take decision for Health Care Technology assessment which includes purchase, maintenance, use, and replacement. In this paper we proposed a analytical hierarchy processes -group decision making (AHP-GDM) model to prioritize the medical equipment for replacement. The model structure includes 11 quantitative and qualitative factors descending as main and sub criteria which can affect of the replacement decision. Each device should be assessed with respect to every covering criterion .and the devices ranking with respect to the criteria independent of other devices . the final score for each devices was calculated as replacement priority index(RPI) . the model was applied for 30 devices in ICU where devices with higher RPI take a higher priority for replacement according for available budget in hospital and Devices with lower RPI take a lower priority for replacement. This model can be integrated as an automated system for medical de...
Nurse Call systems are very important for any medical center / ICU center / hospital, thus affect... more Nurse Call systems are very important for any medical center / ICU center / hospital, thus affecting indirectly the patient’s life. The basic idea of this paper is to design and implement a complete wireless nurse call system in the hospital, displaying the patient room number on a desktop computer LCD. Our proposed system implements a smart controller and several wireless switches using RF technology to continuously monitor and display the state of any room / patient / medical help or assistance needed, in order to provide fast and respectable medical service without any human errors or medical assistance delay, which could occur at any instant of time during system preparation or installation. Our system has succeeded in monitoring up to 5 rooms at the same time collecting real-time data as a prototype. It can be extended to up to 254 monitoring points.
The technology of next generation sequencing brought about evolution in research that based on se... more The technology of next generation sequencing brought about evolution in research that based on sequence which replaces the microarray due to its advantages. The RNA-Seq is a high-throughput gene expression data that uses NGS technologies. The problem of high-throughput RNA-Seq datasets is the high of dimensionally (variables > observations) that need to reduce their dimensions to predict and classify. In this work, ensemble-based on the approach of selection the feature is proposed to choose small optimal genes from various RNA-Seq cancer datasets. The approach combines two filters to select the features (SNR and t-test). In addition, SVM-RFE is used as an embedded feature selection. The SVM classifier used for validating and testing the selected genes using the accuracy measure. The results show the ability to select small optimal genes with high accuracy. Consequently, the genes which are selected will be used as biomarkers to diagnose cancer.
The tremendous growth in genomic data generation present an important issues in handling, storage... more The tremendous growth in genomic data generation present an important issues in handling, storage and transmission of these data. General purpose compression tools such as Bzip2 fail in this respect; hence many algorithms have been specially suggested for the genomic data. This work introduces a new, loss-less and reference-free compression method to increase the compression performance of Bacterial DNA Sequences using Bzip2. The results show a decreasing in compression ratio by 7.74% on average with compression speed about ten times faster than ever before.
Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death... more Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death of one third of the world’s population. The main cause of cardiovascular diseases is atherosclerosis which makes arteries less elastic (called ‘‘hardening of the arteries” or ‘‘arterial stiffness’’). The optical Micro Electro Mechanical System (MEMS) pressure sensor has shown its potential in the diagnosis of arterial stiffness that can be conducted by detecting the pulse pressure in the radial artery. In this paper, we attempt to improve the sensitivity of micro-diaphragm deflection in optical Micro-electromechanical System (MEMS) sensors as applied to pulse pressure detection, thus aiming to determine the safety of a person’s measured pulse of cardiovascular disease. The deflection sensitivity improvement was evidenced using Finite Element Analysis ANSYS software. Corrugation for periphery-clamped silicon nitride (Si3N4) micro-diaphragm based on the variation of the diaphragm thicknes...
Thermal ablation modalities, for example radiofrequency ablation (RFA) and microwave ablation, ar... more Thermal ablation modalities, for example radiofrequency ablation (RFA) and microwave ablation, are intended to prompt controlled tumour removal by raising tissue temperature. However, monitoring the size of the resulting tissue damage during the thermal removal procedures is a challenging task. The objective of this study was to evaluate the observation of RFA on an ex vivo liver sample with both a commercial and a low-cost system to distinguish between the normal and the ablated regions as well as the thermally affected regions. RFA trials were conducted on five different ex vivo normal bovine samples and monitored initially by a custom hyperspectral (HS) camera to measure the diffuse reflectance (Rd) utilising a polychromatic light source (tungsten halogen lamp) within the spectral range 348–950 nm. Next, the light source was replaced with monochromatic LEDs (415, 565 and 660 nm) and a commercial charge-coupled device (CCD) camera was used instead of the HS camera. The system algo...
DNA microarray is an innovative tool for gene studies in biomedical research, and its application... more DNA microarray is an innovative tool for gene studies in biomedical research, and its applications can vary from cancer diagnosis to human identification. Image processing is an important aspect of microarray experiments, the primary purpose of the image analysis step is to extract numerical foreground and background intensities for the red and green channels for each spot on the microarray. The background intensities are used to correct the foreground intensities for local variation on the array surface, resulting in corrected red and green intensities for each spot that can be considered as a primary data for subsequent analysis. Most techniques divide the overall microarray image processing into three steps: gridding, segmentation, and quantification. In this paper, a  simple automated gridding technique is developed with a great effect on noisy microarray images. A segmentation technique based on ‘edge-detection’ is applied to identify the spots and separate the foregroun...
BACKGROUND AND OBJECTIVES White blood cells (WBCs) differential counting yields valued informatio... more BACKGROUND AND OBJECTIVES White blood cells (WBCs) differential counting yields valued information about human health and disease. The current developed automated cell morphology equipments perform differential count which is based on blood smear image analysis. Previous identification systems for WBCs consist of successive dependent stages; pre-processing, segmentation, feature extraction, feature selection, and classification. There is a real need to employ deep learning methodologies so that the performance of previous WBCs identification systems can be increased. Classifying small limited datasets through deep learning systems is a major challenge and should be investigated. METHODS In this paper, we propose a novel identification system for WBCs based on deep convolutional neural networks. Two methodologies based on transfer learning are followed: transfer learning based on deep activation features and fine-tuning of existed deep networks. Deep acrivation featues are extracted from several pre-trained networks and employed in a traditional identification system. Moreover, a novel end-to-end convolutional deep architecture called "WBCsNet" is proposed and built from scratch. Finally, a limited balanced WBCs dataset classification is performed through the WBCsNet as a pre-trained network. RESULTS During our experiments, three different public WBCs datasets (2551 images) have been used which contain 5 healthy WBCs types. The overall system accuracy achieved by the proposed WBCsNet is (96.1%) which is more than different transfer learning approaches or even the previous traditional identification system. We also present features visualization for the WBCsNet activation which reflects higher response than the pre-trained activated one. CONCLUSION a novel WBCs identification system based on deep learning theory is proposed and a high performance WBCsNet can be employed as a pre-trained network.
The management of medical equipment raises a range of complex problems including those associated... more The management of medical equipment raises a range of complex problems including those associated with replacement processes. One of the most significant challenges is to identify a proper list of medical equipment that requires replacement and then to optimize this list. In this article, we present a new approach to solve this problem by integrating Quality Function Deployment (QFD) and Genetic Algorithm (GA) in one framework. In a previous application, QFD has proven its validity to solve the priority problem; meanwhile GA is an optimization method. Hence, the proposed model, QFD-GA, was developed to prioritize the medical equipment for replacement process taking into account a set of criteria; in addition, the prioritized list is optimized according to the available budget of the hospital to maximize the number of replaced devices. The validation of the proposed model was carried out on sixty devices that include different types of medical equipment in one public hospital. Results show that the proposed model can efficiently classify the priority into four subcategories, and simultaneously maximize the number of medical equipment to be replaced considering the budget constraint.
Photodiagnosis and Photodynamic Therapy, Mar 1, 2021
SIGNIFICANCE The estimation of tissue oxygenation is vital in the diagnosis and therapeutic evalu... more SIGNIFICANCE The estimation of tissue oxygenation is vital in the diagnosis and therapeutic evaluation of a huge assortment of diseases. However, the hyperspectral (HS) imaging system is a rising innovation that can be utilized to build a highly sensitive, non-invasive, and tissue hemoglobin immersion map. OBJECTIVE As a result of the urgent need to design and implement early detection devices and applications for the COVID-19 pandemic, we propose building a non-invasive custom optical imaging system to assist with phlebotomy and vascular approach to survey the reliability of blood oxygen saturation (SpO2) levels recovered from spectral images. MATERIALS AND METHODS HS images were gathered from 15 healthy subjects without previous medical history complications and with an average age range of 20 to 38 years, who were undergoing phlebotomy. The forearm was vigorously illuminated utilizing an HS camera with polychromatic source light of spectrum range (400∼980 nm). Spectroscopic reflectance images were caught by a focal plane exhibit for the region of interest (ROI). Then the custom algorithm comprising normalization and moving average filtering for noise removal was applied, followed by K-mean clustering for image segmentation to visualize and highlight the arteries and the veins in the investigated forearm. RESULTS The investigations show that after normalization of the recorded signal from the HS camera of the participating subjects it was noticed that at wavelength of 460 nm the oxygenated arteries had a stronger signal than the de-oxygenated veins, and at a wavelength of 750 nm the de-oxygenated veins had a stronger signal than the oxygenated arteries. Thus, the ideal wavelength to reveal the oxygenated arteries was 460 nm, and the ideal wavelength to reveal the de-oxygenated veins was 750 nm. CONCLUSIONS HSI is a prospective technique to assist with phlebotomy and non-contact oxygen saturation approach. Additionally, it may permit future surgical or pharmacological intercessions that titrate or limit ischemic injury continuously.
Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death... more Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death of one third of the world’s population. The main cause of cardiovascular diseases is atherosclerosis which makes arteries less elastic (called ‘‘hardening of the arteries” or ‘‘arterial stiffness’’). The optical Micro Electro Mechanical System (MEMS) pressure sensor has shown its potential in the diagnosis of arterial stiffness that can be conducted by detecting the pulse pressure in the radial artery. In this paper, we attempt to improve the sensitivity of micro-diaphragm deflection in optical Micro-electromechanical System (MEMS) sensors as applied to pulse pressure detection, thus aiming to determine the safety of a person’s measured pulse of cardiovascular disease. The deflection sensitivity improvement was evidenced using Finite Element Analysis ANSYS software. Corrugation for periphery-clamped silicon nitride (Si3N4) micro-diaphragm based on the variation of the diaphragm thicknes...
The technology of next generation sequencing brought about evolution in research that based on se... more The technology of next generation sequencing brought about evolution in research that based on sequence which replaces the microarray due to its advantages. The RNA-Seq is a high-throughput gene expression data that uses NGS technologies. The problem of high-throughput RNA-Seq datasets is the high of dimensionally (variables > observations) that need to reduce their dimensions to predict and classify. In this work, ensemble-based on the approach of selection the feature is proposed to choose small optimal genes from various RNA-Seq cancer datasets. The approach combines two filters to select the features (SNR and t-test). In addition, SVM-RFE is used as an embedded feature selection. The SVM classifier used for validating and testing the selected genes using the accuracy measure. The results show the ability to select small optimal genes with high accuracy. Consequently, the genes which are selected will be used as biomarkers to diagnose cancer.
2010 5th Cairo International Biomedical Engineering Conference, 2010
Abstract Observation of maintenance programs in healthcare facilities is an essential requirement... more Abstract Observation of maintenance programs in healthcare facilities is an essential requirement for almost all equipment in order to guarantee its performance, prevent sudden failures and to extend the equipment's life expectancy. In this work availability, reliability and performance efficiency were used for equipment performance judgment, in order to support the clinical engineer's decision making process. Time series models were the suggested technical tools used for prediction. Two models were used in this study. Raw data were ...
Technology in its own right is now universally recognized as the engine of economic growth. Techn... more Technology in its own right is now universally recognized as the engine of economic growth. Technological innovation generates increases in productivity, keeps companies competitive, and ultimately enables increases in the nation’s standard of living. It is precisely through doing a better job of managing technology that an improved standard of living can be created for all citizens. Unlike manufacturing companies that produce particular goods, hospitals must offer the highest quality of care possible to their customers, the patients and their families. Although manufacturers must produce high quality products and/or services (to remain in business), the health care industry’s failure to offer quality care, can lead to life threatening consequences. Repair and PM are important to the success of any clinical engineering program. Since the goal of the clinical engineering department is to provide a safe and effective for delivery of the best health care
Modern DNA sequencing instruments are able to generate huge amounts of genomic data. Those huge v... more Modern DNA sequencing instruments are able to generate huge amounts of genomic data. Those huge volumes of data require effective storage, fast transmission, provision of quick access to any record, and superior functionality. Data storage costs have an appreciable proportion of total cost in the creation and analysis of DNA sequences. In particular, the increase in the DNA sequences is highly manageable due to a tremendous increase in the disk storage capacity. Standard compression techniques failed to compress these sequences. Recently, new algorithms have been introduced specifically for this purpose. In this paper, we comparatively survey the main ideas and results of lossless compression algorithms that have been developed for DNA sequences.
Clinical Engineering Department in hospitals need to take decision for Health Care Technology ass... more Clinical Engineering Department in hospitals need to take decision for Health Care Technology assessment which includes purchase, maintenance, use, and replacement. In this paper we proposed a analytical hierarchy processes -group decision making (AHP-GDM) model to prioritize the medical equipment for replacement. The model structure includes 11 quantitative and qualitative factors descending as main and sub criteria which can affect of the replacement decision. Each device should be assessed with respect to every covering criterion .and the devices ranking with respect to the criteria independent of other devices . the final score for each devices was calculated as replacement priority index(RPI) . the model was applied for 30 devices in ICU where devices with higher RPI take a higher priority for replacement according for available budget in hospital and Devices with lower RPI take a lower priority for replacement. This model can be integrated as an automated system for medical de...
Nurse Call systems are very important for any medical center / ICU center / hospital, thus affect... more Nurse Call systems are very important for any medical center / ICU center / hospital, thus affecting indirectly the patient’s life. The basic idea of this paper is to design and implement a complete wireless nurse call system in the hospital, displaying the patient room number on a desktop computer LCD. Our proposed system implements a smart controller and several wireless switches using RF technology to continuously monitor and display the state of any room / patient / medical help or assistance needed, in order to provide fast and respectable medical service without any human errors or medical assistance delay, which could occur at any instant of time during system preparation or installation. Our system has succeeded in monitoring up to 5 rooms at the same time collecting real-time data as a prototype. It can be extended to up to 254 monitoring points.
The technology of next generation sequencing brought about evolution in research that based on se... more The technology of next generation sequencing brought about evolution in research that based on sequence which replaces the microarray due to its advantages. The RNA-Seq is a high-throughput gene expression data that uses NGS technologies. The problem of high-throughput RNA-Seq datasets is the high of dimensionally (variables > observations) that need to reduce their dimensions to predict and classify. In this work, ensemble-based on the approach of selection the feature is proposed to choose small optimal genes from various RNA-Seq cancer datasets. The approach combines two filters to select the features (SNR and t-test). In addition, SVM-RFE is used as an embedded feature selection. The SVM classifier used for validating and testing the selected genes using the accuracy measure. The results show the ability to select small optimal genes with high accuracy. Consequently, the genes which are selected will be used as biomarkers to diagnose cancer.
The tremendous growth in genomic data generation present an important issues in handling, storage... more The tremendous growth in genomic data generation present an important issues in handling, storage and transmission of these data. General purpose compression tools such as Bzip2 fail in this respect; hence many algorithms have been specially suggested for the genomic data. This work introduces a new, loss-less and reference-free compression method to increase the compression performance of Bacterial DNA Sequences using Bzip2. The results show a decreasing in compression ratio by 7.74% on average with compression speed about ten times faster than ever before.
Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death... more Cardiovascular diseases are one of the leading causes of death. Globally, they underlie the death of one third of the world’s population. The main cause of cardiovascular diseases is atherosclerosis which makes arteries less elastic (called ‘‘hardening of the arteries” or ‘‘arterial stiffness’’). The optical Micro Electro Mechanical System (MEMS) pressure sensor has shown its potential in the diagnosis of arterial stiffness that can be conducted by detecting the pulse pressure in the radial artery. In this paper, we attempt to improve the sensitivity of micro-diaphragm deflection in optical Micro-electromechanical System (MEMS) sensors as applied to pulse pressure detection, thus aiming to determine the safety of a person’s measured pulse of cardiovascular disease. The deflection sensitivity improvement was evidenced using Finite Element Analysis ANSYS software. Corrugation for periphery-clamped silicon nitride (Si3N4) micro-diaphragm based on the variation of the diaphragm thicknes...
Thermal ablation modalities, for example radiofrequency ablation (RFA) and microwave ablation, ar... more Thermal ablation modalities, for example radiofrequency ablation (RFA) and microwave ablation, are intended to prompt controlled tumour removal by raising tissue temperature. However, monitoring the size of the resulting tissue damage during the thermal removal procedures is a challenging task. The objective of this study was to evaluate the observation of RFA on an ex vivo liver sample with both a commercial and a low-cost system to distinguish between the normal and the ablated regions as well as the thermally affected regions. RFA trials were conducted on five different ex vivo normal bovine samples and monitored initially by a custom hyperspectral (HS) camera to measure the diffuse reflectance (Rd) utilising a polychromatic light source (tungsten halogen lamp) within the spectral range 348–950 nm. Next, the light source was replaced with monochromatic LEDs (415, 565 and 660 nm) and a commercial charge-coupled device (CCD) camera was used instead of the HS camera. The system algo...
DNA microarray is an innovative tool for gene studies in biomedical research, and its application... more DNA microarray is an innovative tool for gene studies in biomedical research, and its applications can vary from cancer diagnosis to human identification. Image processing is an important aspect of microarray experiments, the primary purpose of the image analysis step is to extract numerical foreground and background intensities for the red and green channels for each spot on the microarray. The background intensities are used to correct the foreground intensities for local variation on the array surface, resulting in corrected red and green intensities for each spot that can be considered as a primary data for subsequent analysis. Most techniques divide the overall microarray image processing into three steps: gridding, segmentation, and quantification. In this paper, a  simple automated gridding technique is developed with a great effect on noisy microarray images. A segmentation technique based on ‘edge-detection’ is applied to identify the spots and separate the foregroun...
BACKGROUND AND OBJECTIVES White blood cells (WBCs) differential counting yields valued informatio... more BACKGROUND AND OBJECTIVES White blood cells (WBCs) differential counting yields valued information about human health and disease. The current developed automated cell morphology equipments perform differential count which is based on blood smear image analysis. Previous identification systems for WBCs consist of successive dependent stages; pre-processing, segmentation, feature extraction, feature selection, and classification. There is a real need to employ deep learning methodologies so that the performance of previous WBCs identification systems can be increased. Classifying small limited datasets through deep learning systems is a major challenge and should be investigated. METHODS In this paper, we propose a novel identification system for WBCs based on deep convolutional neural networks. Two methodologies based on transfer learning are followed: transfer learning based on deep activation features and fine-tuning of existed deep networks. Deep acrivation featues are extracted from several pre-trained networks and employed in a traditional identification system. Moreover, a novel end-to-end convolutional deep architecture called "WBCsNet" is proposed and built from scratch. Finally, a limited balanced WBCs dataset classification is performed through the WBCsNet as a pre-trained network. RESULTS During our experiments, three different public WBCs datasets (2551 images) have been used which contain 5 healthy WBCs types. The overall system accuracy achieved by the proposed WBCsNet is (96.1%) which is more than different transfer learning approaches or even the previous traditional identification system. We also present features visualization for the WBCsNet activation which reflects higher response than the pre-trained activated one. CONCLUSION a novel WBCs identification system based on deep learning theory is proposed and a high performance WBCsNet can be employed as a pre-trained network.
Uploads
Papers by Amr Sharawi