جستجوی مقالات مرتبط با کلیدواژه "یادگیری عمیق" در نشریات گروه "پزشکی"

Colorectal cancer (CRC) is one of the most common cancers globally. Prevention, precise early detection and diagnosis, selecting the best treatment strategy, and correct prediction of prognosis are essential to reduce the burden of CRC and to improve the patients’ survival and quality of life. Along with the explosion of available clinical, omics, pathological, and radiological data, artificial intelligence (AI) has shown a great potential in CRC clinical management, and has provided clinicians with new auxiliary approaches to identify high-risk patients, to select personalized treatment strategies, and to predict prognosis. This review is a comprehensive study on the research progress and clinical applications of AI in management of CRC, including screening, diagnosis, treatment, and prognosis. In short, this review demonstrates the current status of the AI in CRC management. Limitations, challenges, and future prospects of AI in the clinical management of CRC are also discussed.

Larynx cancer can be benign or malignant based on various factors. This research aimed to provide a machine learning-based model to improve the diagnosis of individuals with larynx cancer.

In the first step, the voices of the people who visited the medical centers (including the sounds (A), (E), and (O)) were recorded and considered as a data set. In the second step, the data were classified into three classes (benign cancer, malignant cancer, and healthy) by a specialist. In the third step, the data cleaning was done. In the fourth step, the features related to sound were extracted from the data. In the fifth step, five machine learning models including SVM, Decision Tree, Naïve Bayes, MLP, and Random Forest were implemented on the data set. Finally, the performance of the models was evaluated using evaluation criteria such as accuracy, F-score, and other evaluation criteria.

The results of the implementation showed that the SVM model had a higher accuracy than other models for the sound (A) and sound (O) with an accuracy of 0.818, and the sound (E) with an accuracy of 0.818 in the model MLP had the highest accuracy.

The present study evaluated machine learning models for the diagnosis of laryngeal cancer based on audio data. The results showed that the use of the SVM model for the diagnosis of laryngeal cancer can help diagnose this disease more accurately and provide reliable results.

In this study, a method for automatic prediction of the survival rate of patients with glioblastoma tumor based on machine learning methods and MRI images is presented.

The data set used in this study is the BraTS 2017 database with 163 samples. Each sample of database images has four different imaging modalities as well as information such as the patient's overall life expectancy according to the patient's day and age. Database images are labeled into three categories, short-term, medium-term, and long-term, based on patient longevity after treatment. To improve the prediction results, different types of features were extracted and taught by different machine learning methods. The considered features include texture, volumetric, statistical, and deep features. The machine learning methods used include support vector machine, nearest neighbors, linear discriminant analysis, and decision tree.

The best prediction accuracy based on the classification was obtained using deep features extracted by a pre-trained convolutional neural network (CNN) and by linear discriminant analysis.

Deep learning approaches showed a good performance in the prediction of medical parameters such as survival rate time.

Pulmonary emphysema is one of the lung diseases that usually remains unknown until old age and does not have a definitive treatment. A quick diagnosis of this disease helps a lot to the people involved in this disease and prevents the growth of emphysema masses. This research tries to in early diagnosis of this disease with the help of deep learning methods.

This research tries to diagnose this disease faster with the help of Unet neural network optimized with GPC meta-heuristic algorithm. The data of this research were collected from Imam Ali and Bu Ali Sina hospitals, Zahedan city, Sistan and Baluchistan province. The data include 300 pieces with emphysema, including 65 cases of CLE, 97 cases of PSE, 138 cases of PLE, and 45 cases of normal data. These data were analyzed by Unet deep neural network and GPC optimization algorithm, and finally, with the help of accuracy criteria, recall, specificity, and F-measure were compared and investigated with other methods.

In this research, the criteria used have much better results compared to other emphysema diagnosis methods with the help of the optimized Unet network, with accuracy of 18.97, prediction of 40.98, sensitivity of 48.23, and f score of 97.50, respectively, which shows a faster, more accurate, and more effective diagnosis of this disease with the help of the proposed method.

Using the right deep learning methods in combination with strong optimization algorithms can enable faster and more accurate treatment of diseases.

Alzheimer's disease is an irreversible neurological condition characterized by cognitive, behavioral, and memory impairments. Early prediction before the transition from mild cognitive impairment to Alzheimer's disease is still a challenging issue. This study aimed to identify factors associated with Alzheimer's disease.

This study proposes a framework for predicting Alzheimer's disease using data collected from the OASIS project, made available by the Washington University Research Center. In this study, a deep neural network was used for prediction. A particle swarm optimization (PSO) algorithm was employed for selecting appropriate features. The combination of these two methods increases the accuracy of the proposed prediction method.

The results indicate that the proposed method achieves higher accuracy with fewer features. Among the 11 features in this dataset, six features (age, socioeconomic status, Mini-mental state examination score, clinical dementia rating scale, estimated total intracranial volume, and normalized whole-brain volume) have a significant impact on predicting the disease. Among these six features, the clinical dementia rating scale is of great importance.

This study investigated the influential factors and prediction of Alzheimer's disease. Early diagnosis of Alzheimer's disease allows for the provision of appropriate diagnostic and therapeutic services, as well as an improvement in patients' quality of life. The proposed method in this study is compared with various machine learning algorithms that have shown good accuracy in predicting Alzheimer's disease. The results indicate that the accuracy of the proposed method is higher with fewer features.

With the increase in the number of infections and deaths of covid-19, the need is felt to use artificial intelligence and machine learning tools to diagnose the corona virus quickly and on time. In this study, using chest x-ray images and VGG-16 deep neural network, an automatic system was designed and implemented to detect cases of covid-19.

In this research, chest x-ray images obtained from the Kaggle database were used. The design of this study included: data sampling, training division, creating directories, transferring images to their own folders, classifying images, creating the proposed VGG-16 model in Python programming language and Keras and Tensorflow libraries, evaluation of the proposed model and finally the creation of the confusion matrix and their analysis and interpretation.

The accuracy and Precision of the proposed model for the positive Covid-19 class were 91 and 93%, respectively. Also, the recall rate and F1-Score for cases with covid-19 were 94% and 88% respectively.

Due to the high accuracy and precision of the proposed model, it can be used to diagnose covid-19 and separate COVID-19 cases from healthy cases and also be used as an auxiliary tool to help doctors diagnose this disease.

In recent years, artificial intelligence (AI) techniques are rapidly becoming clinical practices such as diagnosis and prognosis processes, treatment effectiveness evaluation, and disease monitoring. Previous studies have shown interesting results regarding the diagnostic efficiency of artificial intelligence methods in differentiating patients with multiple sclerosis (MS) from healthy individuals or other demyelinating diseases. There is a lack of a comprehensive systematic review on the role of AI in MS diagnosis. Our aim was to conduct a systematic review to document the performance of artificial intelligence in MS diagnosis. In this study, we conducted a comprehensive and systematic search using the PubMed database. All original studies that focused on deep learning or artificial intelligence to analyze methods aimed at diagnosing MS using MRI images were included in our study.

For this review, we searched PubMed for studies on the application of artificial intelligence in MS using MRI images published in English during the period 2010-2023. The search strategy was based on the words Mesh and their combinations. All studies were reviewed, but only the most relevant ones were used in this review.

Artificial intelligence, using deep learning methods, can predict the incidence of MS and its complications based on the risk factors of the disease and reduces the cost and time spent for various medical tests. Artificial intelligence makes this possible by extracting information and performing the necessary processing using methods such as CNN.

MS diagnosis based on new markers and artificial intelligence is a growing field of research with MRI images. All these results show that with advances in artificial intelligence, the way MS patients are monitored and diagnosed can change. However, several challenges remain, including better understanding of information selected by AI algorithms, appropriate multicenter and longitudinal validation of results, and practical aspects related to hardware and software integration. In general, the critical importance of human supervision to optimize and fully utilize the potential of artificial intelligence approaches cannot be ignored.

Identifying and diagnosing Alzheimer's disease in brain tissue is one of the serious challenges in diagnosis in the field of medical image processing. Currently, MRI is the most common way to diagnose Alzheimer's disease, and failure to correctly identify the tissue involved in it can lead to incorrect diagnosis as healthy brain tissue. Deep learning algorithm as a process of detecting features related to damaged tissue and extracting useful information. In this research, we decided to use the convolutional neural network in the processing of medical images so that we can perform the diagnosis with better accuracy than the previous works.

Using a convolutional neural network, the features of MRI images have been extracted. Alzheimer's images have been analyzed using Matlab2023a software and the intended outputs have been obtained.

Brain Alzheimer's images have been analyzed after pre-processing and entering the deep neural network, and in the output of the proposed algorithm, the identification accuracy and identification speed of the algorithm with the improvement of cloud parameters was higher compared to other common methods, which was 96% accuracy. presented in identification.

The purpose of using deep learning is to make image data with large dimensions and a large number into a conceptual form for machines. It is expected that in the future feature extraction will be done more accurately and more details will be available to machine vision systems to recognize objects in the image.

Breast cancer is the most common cancer in women that causes more deaths than other cancers. Thermography is one of the methods of breast cancer diagnosis. The most important challenge in early detection of these images can be human error or lack of access to a skilled person. The use of artificial intelligence methods in image processing can be effective in early detection and reduction of human error. The main aim of this research was to introduce hybrid networks for intelligent diagnosis of breast cancer from thermographic images.

The thermographic images used in this study were collected from the DMR-IR database. First, the main features of the images were extracted by deep convolutional network (CNN). Then, FCNNs and SVM algorithms were used to classify breast cancer from thermographic images.

The accuracy rate for CNN_FC and CNN-SVM algorithms was 94.2% and 0.95%, respectively. In addition, the reliability parameters for these classifiers were calculated as 92.1%, and 97.5%, and the sensitivity for each of these classifiers as 95.5%, and 94.1%, respectively.

The proposed model based on the deep hybrid network has good accuracy compared to similar algorithms; therefore, it can help doctors in the early diagnosis of breast cancer through thermographic images and minimize human error.

In recent years, medicine supply chain management has become more significant, especially after the Covid-19 pandemic. The most important issue is supply chain cost control. If the drug inventory is not properly managed, it will lead to issues such as the lack of inventory of certain drugs, provision of excess inventory, increased costs, and, finally, patient dissatisfaction.

In this study, an attempt has been made to predict and manage the pharmaceutical needs of hospitals using an efficient deep-learning algorithm. The drug consumption data for ten years of Besat General Hospital in Hamedan are extracted from the HIS database. As a case study, the accuracy of the predictive model is evaluated, especially for cefazolin. We use a deep model to analyze the medical time-series data efficiently. This model consists of a Long Short-Term Memory network, which can sufficiently recognize the change history in time-series prediction applications. The proposed model with many adjustable parameters in the deep architecture will bring good performance to overcome the complexities of the learning problem.

Using the deep learning method can increase robustness by reducing the effects of complexity and uncertainty in medical data. The average forecasting error for the proposed method is 0.043, and the measured values for RMSE, MAE, and R2  are 0.335, 0.260, and 0.851, respectively.

A comprehensive comparison between some other predictive methods and the implemented model shows the outperformance of the proposed approach. Additionally, the evaluation results indicate the efficiency of the proposed approach.

Brain-computer interface systems provide the possibility of communicating with the outside world without using physiological mediators for people with physical disabilities through brain signals. A popular type of BCIs is the motor imagery-based systems and one of the most important parts in the design of these systems is the classification of brain signals into different motor imagery classes in order to transform them into control commands. In this paper, a new method of brain signal classifying based on deep learning methods is presented.

This cross-sectional study was conducted at Isfahan University of Medical Sciences, School of Advanced Technologies in Medicine, from February 2020 to June 2022. In the pre-processing block, segmentation of brain signals, selection of suitable channels and filtering by Butterworth filter have been done; then data has transformed to the time-frequency domain by three different kinds of mother wavelets including Cmor, Mexicanhat, and Cgaus. In the classification step, two types of convolutional neural networks (one-dimensional and two-dimensional) were applied whereas each one of them was utilized in two different architectures. Finally, the performance of the networks has been investigated by each one of these three types of input data.

Three channels were selected as the best ones for nine subjects. To separate 8-30 Hz, a 5th degree Butterworth filter was used. After finding the optimal parameters in the proposed networks, wavelet transform with Cgauss mother wavelet has the highest percentage in the both proposed architectures. Two-dimensional convolutional neural network has higher convergence speed, higher accuracy and more complexity of calculations. In terms of accuracy, precision, sensitivity and F1-score, two-dimensional convolutional neural network has performed better than one-dimensional convolutional neural network. The accuracy of 92.53%, which is obtained from the second architecture, as the best result, is reported.

The results obtained from the proposed network indicate that suitable, and well-designed deep learning networks can be utilized as an accurate tool for data classification in application of motion perception.

Hand radiographs are commonly used to evaluate bone maturity. So that the significant difference between the estimated bone age and the chronological age can indicate a developmental disorder. However, the manual evaluation of images is usually a time-consuming and observer-dependent process. Therefore, in this paper, an automatic method for the assessment of bone age using radiographs of children's hands is proposed.

In this fundamental-applied research, the collection of radiographic images of the Radiological Society of North America (RSNA) was used, and transfer learning methods were proposed. The input images were first pre-processed due to low quality. Then a pre-trained model based on DenseNet-121 was used to extract the discriminating spatial features.

Evaluations using five pre-trained models on the RSNA dataset showed that the DenseNet-121 model, after adjustment, could perform better than other models, with a mean absolute error of 9.8 months.

Skeletal maturity can be estimated with satisfactory accuracy using the DenseNet-121 model, and this method can help radiologists in quick and accurate measurement of bone age.

Accurate delineation of myocardial fibrosis in Late Gadolinium Enhancement Cardiac Magnetic Resonance (LGE-CMR) has a crucial role in the assessment and risk stratification of HCM patients. As this is time-consuming and requires expertise, automation can be essential in accelerating this process. This study aims to use Unet-based deep learning methods to automate the mentioned process.

This study used three consecutive Unet-based networks for Region of Interest (ROI) detection, myocardial segmentation, and fibrosis delineation. The study was conducted on LGE images of 41 images diagnosed with HCM, which were contoured by two experts.

This model reported a Dice similarity coefficient and accuracy of 89.74 and 98.22 in myocardial segmentation and 88.42 and 94.66 in fibrosis delineation, respectively, and could outperform the previous methods

The results confirm that using deep learning methods for delineating myocardial fibrosis not only can automate the process, but also helps improve the results and decrease the required time.

Glioma is the most common primary brain tumor, and early detection of tumors is important in the treatment planning for the patient. The precise segmentation of the tumor and intratumoral areas on the MRI by a radiologist is the first step in the diagnosis, which, in addition to the consuming time, can also receive different diagnoses from different physicians. The aim of this study was to provide an automated method for segmenting the tumor and intratumoral areas.

This is a fundamental-applied study that was conducted from May 2020 to September 2021 using multimodal MRI images of 285 patients with glioma tumors from the BraTS 2018 Database. This database was collected from 19 different MRI imaging centers, including multimodal MRI images of 210 HGG patients, and 75 LGG patients. In this study, a 2D U-Net architecture was designed with a patch-based method for training, which comprises an encoding path for feature extraction and a symmetrical decoding path. The training of this network was performed in three separate stages, using data from high-grade gliomas (HGG), and low-grade gliomas (LGG), and combining two groups of 210, 75, and 220 patients, respectively.

The proposed model estimated the Dice Similarity Coefficient (DSC) results in HGG datasets 0.85, 0.85, 0.77, LGG datasets 0.80, 0.66, 0.51, and the combination of the two groups 0.88, 0.79, 0.77 for regions the whole tumor, tumor core, and enhancing region in the training dataset, respectively. The results related to Hussdorf Distance (HD) for HGG datasets were 8.24, 9.92, 4.43, LGG datasets 11.5, 11.31, 2.23, and the combination of the two groups 7.20, 8.82, 4.43 for regions the whole tumor, tumor core, and enhancing region in the training dataset, respectively.

Using the U-Net network can help physicians in the accurate segmentation of the tumor and its various areas, as well as increase the survival rate of these patients and improve their quality of life through accurate diagnosis and early treatment.

Evaluation of treatment response is one of the most challenging tasks in the treatment planning of cancer cases. Regarding the significant effect of tumor recurrence in the treatment planning of patients with lung cancer, finding an approach to predict the recurrence of these tumors is of great importance. Nowadays, this process is done experimentally, and its accuracy depends on the experience and proficiency of the oncologist. This study aimed to provide an automated method to detect the recurrence of lung cancer based on imaging and clinical features.

Our proposed method was evaluated in 162 patients with non-small cell lung cancer (NSCLC) using the NSCLC radiogenomic database in the Cancer Imaging Archive (TCIA) portal. After pre-processing, segmentation was performed using the Otsu method. In the next step, the radiomic features were extracted using pre-trained AlexNet and GoogleNet models, and along with clinical features, they were used to detect lesion recurrence. Finally, all cases were classified into two classes using machine learning methods.

The proposed method used clinical and deep features. The classification was done using various models, and the accuracy of the support vector machine by AlexNet features resulted in the highest performance. The mean values of accuracy, sensitivity, and specificity for this model are 99.76, 99.77, and 99.76%, respectively.

The main finding of this study was revealing the capability of deep learning methods in extracting features from medical images. For example, the AlexNet was able to extract features from CT images of NSCLC patients, which are very helpful in the recurrence prediction of these lesions.

Background &

Mammography is one of the reliable methods for early detection of breast cancer. However, it is difficult for the radiologist to provide an accurate and uniform assessment of the massive mammograms produced in the extensive screening. Therefore, the presence of an intelligent system that is highly accurate in detecting the location of a cancerous mass will be very necessary and necessary. In this regard, this research, by using mammography images and image processing techniques, has been tried to get an accurate diagnosis of the location of breast cancer. For this purpose, first by using some digital image enhancement techniques, an attempt is made to increase the recognition of cancerous tissues, and then by using classification techniques, the precise separation of cancerous parts from healthy parts of the breast is done. In research, various techniques have been proposed to improve the detection of tumors in mammograms and the accuracy of breast cancer classification. The basic problems in breast mammography in identifying and classifying masses and microcalcifications are caused by various factors. One of these complications is due to the awkward and illogical shape of some clusters of calcifications. The boundaries of each of the microcalcifications in the cluster cannot be well defined, and the radiologist may not be able to make an accurate diagnostic decision about the clinical nature of the microcalcifications in an area, but he can usually identify suspicious areas. In the paper, they presented a CAD system for processing mammographic images. They used the compressed breast thickness parameter for feature selection showed the importance of breast compression and changes in breast composition and then applied it to a variety of mammography image processing tasks. Considering that breast thickness is a key parameter in calculations and is not usually recorded; they showed that breast thickness can be estimated from an image and examined its sensitivity on the estimates. Then they discussed how to simulate X-rays in each examination and also simulate the appearance of anatomical structures inside the breast. In the research, tissue characteristics were used to automatically evaluate breast tissue density in digital mammograms. In this approach, the target area is limited to breast tissue only; so that artifacts, background, and head muscles are removed.Breast cancer is the most common cancer among women and the second cause of cancer-related death in women. Mammography is a simple type of imaging and a tool for early detection of non-palpable breast cancers; however, examining and interpreting a large number of mammogram images is a challenging and time-consuming task, and the possibility of human errors is high. One of the most important deep learning methods is convolutional neural networks. In the article, the digital database for screening mammography from the CBIS version was used to improve data validation.

In this research, three types of architecture were designed in the two-class mode and one type of architecture was designed in the three-class mode. To design the network, the layers were arranged according to Figure 5, which uses an input layer of size 159 x 145 a two-dimensional convolution layer of size 20 x 8, and a maximum integration layer of size 5 x 2, and two fully connected layers. (The maximum integration layer was used because it uses the maximum amount of neuron clusters of the previous layer and also causes faster convergence, and improves generalization and selection of invariant features). The third designed network architecture is shown in such a way that one input layer three 2D convolutional layers three maximum integration layers and two fully connected layers are used, the size of each layer is shown in Table (3). Layering is equal to one. The training time is 6:37 and the accuracy obtained for the validation data is 92.58% and the test data is 86.5%.

The simulation results for 310 data for the second type of two-class architecture, the training time is 6:06, and the accuracy obtained for the validation data is 84.40% and 72.82% for the test data. Also, the simulation results for 1240 The data for the first type of two-class architecture, the training time is 3:44:54, and the accuracy obtained for the validation data is 51.72% and the test data is 51.69%.

After a series of pre-processing, the number of used images was selected as 310. Then two other types of architecture were designed, and by applying the processed data, the accuracy of the architectures for 310 data was 42.39%, 7and 2.82%, respectively. 79.34% was obtained. The accuracy of the architectures for 1240 data was 51.69%, 65.45%, 72.46%. In the three-class mode, 1318 images in the database were used, and due to the lack of the same size, the images were resized. Then the image mask was applied to the images and given to the designed convolutional neural network, and the data was classified into three classes. According to the pre-processing and operations that have been done, the accuracy of the network has increased (72.39%) and the result has improved. The advantage of the method is the increased accuracy of validation and test data.

The modern world today allows images to be received and stored digitally. To get better results, it is sometimes necessary to make changes to these images. These changes pursue three main goals: image processing, analysis, and comprehension. For this reason, computer image processing systems have been developed to perform these operations with better speed and accuracy. Four major processes occur in these systems: preprocessing, image quality enhancement, image conversion, and image classification and analysis. In these methods, using mathematics, rules have been created by the computer to simulate human visual elements, and it is an aspect of image analysis that is used for specific purposes. Skin imaging systems provide the ability to process images in high volume and with minimal time and cost, as well as increase the accuracy of diagnosis and classification of diseases. These systems, fatigue, human error and other weaknesses that the diagnostician can suffer. Do not have it (1). The first step in diagnosing skin diseases and analyzing digital images of patients with skin lesions is to take a color photograph of the lesion area. One of the most valid methods for this is the use of a dermoscopic device (2). Dermoscopy, also known as dermatoscopy, is an effective tool for dermatologists involved in early diagnosis. Using dermoscopically evaluated pigmented lesions, abnormal structural features are detected and the border of the lesions is accurately observed (3). Accordingly, benign lesions can be detected without the need for biopsy. Dermoscopy increases the accuracy of the diagnosis and helps GPs to correctly identify people with suspected lesions who need to be referred to a specialist. Dermoscopy is also effective in diagnosing non-pigmented skin lesions and inflammatory dermatoses. In dermoscopy, the skin is examined using a special microscope (4).

The proposed algorithm of this research can be divided into 7 separate steps (loading data set, data integration: data balancing with data amplification or data augmentation technique, data cleaning: clearing images to remove hair noise, slicing images to separate skin from skin Healthy, data conversion: data preparation, convolution neural network design (CNN) and training of the proposed model for image feature extraction, classification combination and mass learning by majority voting method). Which was implemented in Python language in Google Columbine environment and supervised. For this study, 25,331 dermoscopic images consisting of skin lesions were included (70% educational images, 15% experimental and 15% validation). In data preprocessing, the data were balanced, then the data cleaning operation was performed to remove hair noise, and the data reduction operation was performed to segment the images by separating the lesion from healthy skin. In the next process By designing the convolution neural network, training data were extracted for feature extraction, and by combining the classifiers, an automated system for diagnosing skin diseases was created and evaluated in dermoscopic images.

In the proposed method of hair noise removal, the quality of images is increased and also the separation of the lesion from healthy skin is optimally designed to accelerate image processing to extract high-level features in the convolutional neural network and increase the accuracy of diagnosis and classification to create An automated diagnostic system is a feature of this study compared to other studies. According to the research results, the use of an automated system for the diagnosis and classification of skin diseases and cancers for health-related care is recommended.

Today, the applications of artificial intelligence are not hidden from anyone. Among these, machine learning as one of the most important branches of this field has a special place in all sciences. Deep learning has proven its worth by using the basics of artificial neural networks in solving many issues in the field of medical image processing such as classification. Experts also based on various experiences of using training methods to conclude that there is no single specific training algorithm that can be successful for all applications and has the highest accuracy. Hence they suggest combined learning. According to the important results, although each of the algorithms had a successful performance individually, but combining several algorithms with each other has led to higher accuracy and less error-making decisions. This study is a step towards helping physicians and specialists in diagnosing skin diseases and benign and malignant skin cancers and can help GPs or other physicians to better manage high-risk lesions. Secondary triage as well as avoid unnecessary treatments and minimize biopsy, which is an invasive and costly procedure. This research helps to provide health-related care, forecasting and treatment, as well as cost savings for both patients and health care providers. Also, in deprived areas and far from the specialist, dermoscopic devices with the help of this algorithm can cause timely treatment and reduce patients' costs and time in the field of diagnosis and, if necessary, referral of patients to the desired specialist. Available as a commercial software package. This software package has the ability to connect to dermoscopic devices. By connecting this software package to dermoscopy, a device is created to quickly diagnose skin diseases and cancers. The greatest value of this dissertation is that it is used as a benchmark for designing future studies and evaluating skin cancer diagnosis techniques in patients who are usually examined by a general practitioner and specialist. The findings of the present study are also consistent with the results of Andre and Pachko (2019) research on the diagnosis of skin cancer based on deep learning and entropy for Perth samples (6).

Cardiovascular diseases are one of the leading causes of death worldwide. Therefore, early diagnosis of heart disease, evaluation of cardiovascular system using cardiac hearing and Phonocardiogram (PCG) analysis which is a low cost, non-invasive, rapid method, and automatic screening of cardiovascular patients in remote areas is crucial. The aim of this study is to present a new method for screening heart patients based on signal processing (PCG) that is cheap and fast and has sufficient accuracy.

In this study, for screening 2062 labeled PCG signals, by extracting new features and applying them in 1- Random forest network 2- K-nearest neighbors 3- Decision tree 4- Linear discriminant analysis 5- Logistic regression and 6- Deep neural network, six different models were constructed and each of them was evaluated by k fold cross-validation method (K = 10). The test data were applied to the mentioned models and based on the outputs of these models, three indicators of accuracy, sensitivity and specificity were calculated. We showed and developed a new solution in differentiating and screening some heart patients from healthy individuals using PCG analysis.

Evaluation on the mentioned models was calculated by the three indicators, repeated 5 times and their mean and variance values were calculated. The highest sensitivity value is related to deep neural network (DNN) with sensitivity of 96.4 ± 0.14 and accuracy of 93.4 ± 0.11.

The new differential features along with the success of the proposed deep neural network in differentiating and screening between PCGs of healthy individuals and heart patients, shows the efficiency of the proposed algorithm. This method can be further improved with simultaneous multimodal classifier and the application of the voting rule.

Early detection of diabetic retinopathy in the military forces can prevent their performance reduction or avoid the occurrence of operational errors. So an automated and optimal method to diagnose the degree of disease from retinal images is valuable in the prevention of acute phases. The purpose of this article is to present a new method in determining of proliferation based on the deep generative adversarial networks models (GANs).

In this study, which was conducted in 2018-2019, a new method was used to classify 35,126 medical images on the data set available from the Kagel site related to a hospital in the UK. To create a balance between levels, first with the help of a deep GAN, the number of small classes was increased, then using a designed classifier, the degree of disease was determined in different ways.

Using the designed GAN, an accuracy of about 87% was obtained for classification, which was about 7% more than the best similar works. In addition, with the distribution of the model, the efficiency of automation also showed an improvement of 60%.

By solving the problem of imbalance between different levels of retinopathy, by producing new images using the designed GAN and distributing the mentioned operations, while increasing the performance, optimal accuracy has been obtained. Therefore, this new strategy can be used to automate the diagnosis of diabetic retinopathy.