A New Approach for Detecting Fundus Lesions Using Image Processing and Deep Neural Network Architecture Based on YOLO Model.

Diabetic Retinopathy is one of the main causes of vision loss, and in its initial stages, it presents with fundus lesions, such as microaneurysms, hard exudates, hemorrhages, and soft exudates. Computational models capable of detecting these lesions can help in the early diagnosis of the disease and prevent the manifestation of more severe forms of lesions, helping in screening and defining the best form of treatment. However, the detection of these lesions through computerized systems is a challenge due to numerous factors, such as the characteristics of size and shape of the lesions, noise and the contrast of images available in the public datasets of Diabetic Retinopathy, the number of labeled examples of these lesions available in the datasets and the difficulty of deep learning algorithms in detecting very small objects in digital images. Thus, to overcome these problems, this work proposes a new approach based on image processing techniques, data augmentation, transfer learning, and deep neural networks to assist in the medical diagnosis of fundus lesions. The proposed approach was trained, adjusted, and tested using the public DDR and IDRiD Diabetic Retinopathy datasets and implemented in the PyTorch framework based on the YOLOv5 model. The proposed approach reached in the DDR dataset an mAP of 0.2630 for the IoU limit of 0.5 and F1-score of 0.3485 in the validation stage, and an mAP of 0.1540 for the IoU limit of 0.5 and F1-score of 0.2521, in the test stage. The results obtained in the experiments demonstrate that the proposed approach presented superior results to works with the same purpose found in the literature.

Detection of Fundus Lesions through a Convolutional Neural Network in Patients with Diabetic Retinopathy.

Diabetic Retinopathy is a major cause of vision loss caused by retina lesions, including hard and soft exudates, microaneurysms, and hemorrhages. The development of a computational tool capable of detecting these lesions can assist in the early diagnosis of the most severe forms of the lesions and assist in the screening process and definition of the best treatment form. This paper proposes a computational model based on pre-trained convolutional neural networks capable of detecting fundus lesions to promote medical diagnosis support. The model was trained, adjusted, and evaluated using the DDR Diabetic Retinopathy dataset and implemented based on a YOLOv4 architecture and Darknet framework, reaching an mAP of 11.13% and a mIoU of 13.98%. The experimental results show that the proposed model presented results superior to those obtained in related works found in the literature.

A Mathematical Morphology-Based Filter for Noise Reduction and Detail Preservation in Low-Dose Dental CT Images.

During the acquisition on a low-dose radiation computed tomography (CT) scan, images are usually marked by heavy noise and undesired artifacts, which dramatically reduce its applicability in the image processing workflow. A noise reduction and detail preservation filter based on mathematical morphology is presented in this paper. The filter is geared to allow control of an opening operator followed by a systematic contrast limited adaptive histogram equalization (CLAHE) in conjunction with a reconstruction by dilation in last stage. A quantitative metric built on peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and mean-squared error (MSE) were applied to check noise reduction, detail preservation, and performance. The results obtained by the proposed filter were compared with those obtained in the literature, showing very good results: compared with the best-tested filter, the filter had a gain of 7.91% on PSNR, 7.57% on SSIM and 37.8% on MSE.

Use of Flowchart for Automation of Clinical Protocols in mHealth.

For healthcare professionals to use mobile applications we need someone who knows software development, provide them. In healthcare institutions, health professionals use clinical protocols to govern care, and sometimes these documents are computerized through mobile applications to assist them. This work aims to present a proposal of an application of flow as a way of describing clinical protocols for automatic generation of mobile applications to assist health professionals. The purpose of this research is to enable health professionals to develop applications from the description of their own clinical protocols. As a result, we developed a web system that automates clinical protocols for an Android platform, and we validated with two clinical protocols used in a Brazilian hospital. Preliminary results of the developed architecture demonstrate the feasibility of this study.

Locally Adaptive Operators for Red Lesions Detection in Eye Fundus Images.

One of the major features required by automated software tools of screening for diabetic retinopathy is the detection of red lesions. This paper presents a new automatic method in order to locate red lesions in color eye fundus images. The method relies on mathematical morphology operators and has a coarse and a fine detection stages, respectively. The former detection stage detects structures of low-intensity values in the retina, such as microaneurysms, hemorrhages, blood vessels and the fovea center. Additionally, the latter stage proposes to improve the detection of red lesions identified in the previous stage. For experiments, we use the well-known publicly available DIARETDB1 database. The results indicate that our method detected red lesions with 75.81% and 93.48% of mean sensitivity and mean specificity, respectively.

Automatic Identification of Glaucoma Using Deep Learning Methods.

This paper proposes an automatic classification method to detect glaucoma in fundus images. The method is based on training a neural network using public image databases. The network used in this paper is the GoogLeNet, adapted for this proposal. The methodology was divided into two stages, namely: (1) detection of the region of interest (ROI); (2) image classification. We first used a sliding-window approach combined with the GoogLeNet network. This network was trained using manually extracted ROIs and other fundus image structures. Afterwards, another GoogLeNet model was trained using the previous resulting images. Then those images were used to train another GoogLeNet model to automatically detect glaucoma. To prevent overfitting, data augmentation techniques were used on smaller databases. The results demonstrated that the network had a good accuracy, even with poor quality images found in some databases or generated by the data augmentation algorithm.

CTdBem - A New Protocol for Ultra Low Radiation Dose MDCT.

This study describes a novel method of obtaining low-dose computed tomography (CT) scans followed by imaging postprocessing that provides diagnostic quality to such low-dose exams. In addition, we compared the Total Radiation Doses (DLP) of the 64-channel MDCT x 16-channel MDCT for a new Dental CT - CTdBem protocol for hospital use. DLP data obtained from 20 patients using 16-channel MDCT was compared with 20 other patients using 64-channel MDCT. In both tomographic (Aquilion 64 and Brightspeed 16) FOV was approximately 160(V) x 130(H) mm. An imaging postprocessing algorithm was used to provide diagnostic quality to the obtained low-dose CT scans. Imaging postprocessing included imaging smoothing, multiplanar reconstruction (MPR), and volume rendering (VR), as well as surface rendering (SR) to allow three-dimensional printing of the desired scans. The average DLPs were of 28,5 mGy.cm and 54,65 mGy.cm, using the 64-channel MDCT and 16-channel MDCT, respectively. The effective radiation dose (DLP) from the 64-channel MDCT statistically differs from the DLP data from 16-channel MDCT, resulting in a value of p < 0.05 for all comparisons. A novel low-dose CT protocol for dentomaxillofacial assessment using imaging postprocessing techniques is described. The authors concluded that although the DLP values differ statistically (p < 0.05), both equipment (64 and 16-channel MDCT) produce tomographic images of patients with low radiation doses. The greater the number of detectors, the lower the mAs product and, consequently, the amount of X-radiationproduced.

Geração automática de aplicativos móveis a partir de protocolos clínicos / Automatic generation of mobile application from clinical protocols

Atualmente, com a popularização de smartphones, profissionais da saúde podem contar com aplicativos de auxílio ao atendimento clínico. Este trabalho objetiva apresentar uma ferramenta web para geração automática de aplicativos móveis. Seu propósito é permitir que os profissionais da saúde desenvolvam esses aplicativos a partir da descrição dos seus próprios protocolos clínicos. Para isso, primeiramente verificou-se na literatura trabalhos relacionados com geração de aplicativos automáticos e a área de saúde. Após, foi estudado os protocolos clínicos e então foi desenvolvido a arquitetura e lógica de uma ferramenta de geração automática de aplicativos móveis a partir da descrição gráfica de protocolos clínicos. Como resultado, foi desenvolvido um sistema web que automatiza protocolos clínicos para a plataforma Android. A arquitetura desenvolvida proporcionou resultados preliminares que aprovam a viabilidade destes estudo.

Currently, with the popularization of smartphones, healthcare professionals can take advantage of mobile applications to support clinical care. This paper introduces a software tool for automatic generation of mobile applications. The central idea is to allow health professionals to develop these applications from the description of their ownclinical protocols. Thus, the automatic generation of applications in healthcare have been properly investigated in the existing literature. Afterwards, the clinical protocols were studied and then the architecture and logic of the proposed tool was developed from the graphic description of clinical protocols. Consequently, a web system that automates clinical protocols for the Android platform was developed. Preliminary results of the developed architecture demonstrate the feasibility of this study.

Fovea center detection based on the retina anatomy and mathematical morphology.

In this work, we present a new fovea center detection method for color eye fundus images. This method is based on known anatomical constraints on the relative locations of retina structures, and mathematical morphology. The detection of this anatomical feature is a prerequisite for the computer aided diagnosis of several retinal diseases, such as Diabetic Macular Edema. The proposed method is adaptive to local illumination changes, and it is robust to local disturbances introduced by pathologies in digital color eye fundus images (e.g. exudates). Our experimental results using the DRIVE image database indicate that our method is able to detect the fovea center in 37 out of 37 images (i.e. with a success rate of 100%). Using the DIARETDB1 database, our method was able to detect the fovea center in 92.13% of all tested cases (i.e. in 82 out of 89 images). These results indicate that our approach potentially can achieve a better performance than comparable methods proposed in the literature.

Segmentation of the optic disk in color eye fundus images using an adaptive morphological approach.

The identification of some important retinal anatomical regions is a prerequisite for the computer aided diagnosis of several retinal diseases. In this paper, we propose a new adaptive method for the automatic segmentation of the optic disk in digital color fundus images, using mathematical morphology. The proposed method has been designed to be robust under varying illumination and image acquisition conditions, common in eye fundus imaging. Our experimental results based on two publicly available eye fundus image databases are encouraging, and indicate that our approach potentially can achieve a better performance than other known methods proposed in the literature. Using the DRIVE database (which consists of 40 retinal images), our method achieves a success rate of 100% in the correct location of the optic disk, with 41.47% of mean overlap. In the DIARETDB1 database (which consists of 89 retinal images), the optic disk is correctly located in 97.75% of the images, with a mean overlap of 43.65%.

A coarse-to-fine strategy for automatically detecting exudates in color eye fundus images.

The detection of exudates is a prerequisite for detecting and grading severe retinal lesions, like the diabetic macular edema. In this work, we present a new method based on mathematical morphology for detecting exudates in color eye fundus images. A preliminary evaluation of the proposed method performance on a known public database, namely DIARETDB1, indicates that it can achieve an average sensitivity of 70.48%, and an average specificity of 98.84%. Comparing to other recent automatic methods available in the literature, our proposed approach potentially can obtain better exudate detection results in terms of sensitivity and specificity.

A morphological approach for the fovea location in color fundus images.

This paper presents a novel method for the detection of the fovea center in color fundus images. The method was evaluated using a set of 89 images from the DIARETDB1 project, which contains images presenting normal and pathological situations. Using the Mean Absolute Distance (MAD) as a metric, we report 7.37+/-8.89 (mean +/- standard deviation) detection performance for the fovea center which represents an improvement in comparison to other state-of-the-art methods in the literature.