Effect Of Image Resolution Research Articles

Breast density prediction from low and standard dose mammograms using deep learning: effect of image resolution and model training approach on prediction quality

Purpose. To improve breast cancer risk prediction for young women, we have developed deep learning methods to estimate mammographic density from low dose mammograms taken at approximately 1/10th of the usual dose. We investigate the quality and reliability of the density scores produced on low dose mammograms focussing on how image resolution and levels of training affect the low dose predictions. Methods. Deep learning models are developed and tested, with two feature extraction methods and an end-to-end trained method, on five different resolutions of 15,290 standard dose and simulated low dose mammograms with known labels. The models are further tested on a dataset with 296 matching standard and real low dose images allowing performance on the low dose images to be ascertained. Results. Prediction quality on standard and simulated low dose images compared to labels is similar for all equivalent model training and image resolution versions. Increasing resolution results in improved performance of both feature extraction methods for standard and simulated low dose images, while the trained models show high performance across the resolutions. For the trained models the Spearman rank correlation coefficient between predictions of standard and low dose images at low resolution is 0.951 (0.937 to 0.960) and at the highest resolution 0.956 (0.942 to 0.965). If pairs of model predictions are averaged, similarity increases. Conclusions. Deep learning mammographic density predictions on low dose mammograms are highly correlated with standard dose equivalents for feature extraction and end-to-end approaches across multiple image resolutions. Deep learning models can reliably make high quality mammographic density predictions on low dose mammograms.

Using Remote Sensing Multispectral Imagery for Invasive Species Quantification: The Effect of Image Resolution on Area and Biomass Estimation

This study assesses the applicability of different-resolution multispectral remote sensing images for mapping and estimating the aboveground biomass (AGB) of Carpobrotus edulis, a prominent invasive species in European coastal areas. This study was carried out on the Cávado estuary sand spit (Portugal). The performance of three sets of multispectral images with different Ground Sample Distances (GSDs) were compared: 2.5 cm, 5 cm, and 10 cm. The images were classified using the supervised classification algorithm random forest and later improved by applying a sieve filter. Samples of C. edulis were also collected, dried, and weighed to estimate the AGB using the relationship between the dry weight (DW) and vegetation indices (VIs). The resulting regression models were evaluated based on their coefficient of determination (R2), Normalised Root Mean Square Error (NRMSE), p-value, Akaike information criterion (AIC), and the Bayesian information criterion (BIC). The results show that the three tested image resolutions allow for constructing reliable coverage maps of C. edulis, with overall accuracy values of 89%, 85%, and 88% for the classification of the 2.5 cm, 5 cm, and 10 cm GSD images, respectively. The best-performing VI-DW regression models achieved R2 = 0.87 and NRMSE = 0.09 for the 2.5 cm resolution; R2 = 0.77 and NRMSE = 0.12 for the 5 cm resolution; and R2 = 0.64 and NRMSE = 0.15 for the 10 cm resolution. The C. edulis area and total AGB were 3441.10 m2 and 28,327.1 kg (with an AGB relative error (RE) = 0.08) for the 2.5 cm resolution; 3070.04 m2 and 29,170.8 kg (AGB RE = 0.08) for the 5 cm resolution; and 2305.06 m2 and 22,135.7 kg (AGB RE = 0.11) for the 10 cm resolution. Spatial and model differences were analysed in detail to determine their causes. Final analyses suggest that multispectral imagery of up to 5 cm GSD is adequate for estimating C. edulis distribution and biomass.

Investigating the impact of deformation on foam permeability through CT scans and the Lattice‐Boltzmann method

AbstractFoam has a wide range of applications, where the study of its properties and mechanical response has gained a lot of attention in recent years. To deepen the understanding of foam's behavior, the underlying research proposes a new study to determine the deformation‐dependent permeability of foam using a combination of CT scans and the Lattice‐Boltzmann method (LBM). Specifically, the three‐dimensional mesoscopic structure of the foam is reconstructed using nano‐CT images at different compression levels including uncompressed, 15, 25, 35, and 50 percentages of compression. After processing the data from the CT scans, the LBM is applied to simulate single‐phase fluid flow in the deformed porous domain. The permeability at each corresponding deformation stage is characterized and determined by the LBM using the open‐access Palabos software. In this work, the deformation‐dependent intrinsic permeability tensor is considered by applying a meso‐macro hierarchic upscaling scheme. According to the CT scan outcomes, the compression level is inversely proportional to the porosity. The physical phenomenon is discussed in detail and the effect of image resolution on the permeability computation accuracy is investigated.

Effect of image resolution on automatic detection of whitefly (Hemiptera: Aleyrodidae) species on tomato leaflets using deep learning

Deep learning approaches facilitate the rapid and accurate identification of insect pest species in agriculture. However, challenges arise when detecting tiny pests, such as whiteflies (Hemiptera: Aleyrodidae). Although high-resolution images are preferred for accurate detection, they reduce the image acquisition throughput. Hence, this study investigates the impact of image resolution on the accuracy of detecting the adults of two whitefly species, Bemisia tabaci Gennadius and Trialeurodes vaporariorum Westwood, on tomato leaflets. The following state-of-the-art object detectors were used: Faster R-CNN, SSD, YOLOv3, detectoRS, YOLOX, and deformable DETR. The results showed that the models with high-resolution datasets for training and testing performed well in insect detection, although performance declined with decreased image resolution. While, for many models, most errors occurred in species classification, those trained on low-resolution images had a higher proportion of localization errors. Poor performance was also observed when image resolution differed between training and inference. Moreover, the models in which the training dataset was assembled from high- and low-resolution images exhibited similar performance to those trained over each image resolution. Ultimately, detectoRS demonstrated the best performance for lower-resolution images or a mixture of multiple-resolution images, while YOLOX exhibited the highest accuracy for high-resolution images. This study highlights the importance of image resolution for optimal performance in insect pest-detection tasks.

Effect of image resolution on automated classification of chest X-rays.

Deep learning (DL) models have received much attention lately for their ability to achieve expert-level performance on the accurate automated analysis of chest X-rays (CXRs). Recently available public CXR datasets include high resolution images, but state-of-the-art models are trained on reduced size images due to limitations on graphics processing unit memory and training time. As computing hardware continues to advance, it has become feasible to train deep convolutional neural networks on high-resolution images without sacrificing detail by downscaling. This study examines the effect of increased resolution on CXR classification performance. We used the publicly available MIMIC-CXR-JPG dataset, comprising 377,110 high resolution CXR images for this study. We applied image downscaling from native resolution to , , , and and then we used the DenseNet121 and EfficientNet-B4 DL models to evaluate clinical task performance using these four downscaled image resolutions. We find that while some clinical findings are more reliably labeled using high resolutions, many other findings are actually labeled better using downscaled inputs. We qualitatively verify that tasks requiring a large receptive field are better suited to downscaled low resolution input images, by inspecting effective receptive fields and class activation maps of trained models. Finally, we show that stacking an ensemble across resolutions outperforms each individual learner at all input resolutions while providing interpretable scale weights, indicating that diverse information is extracted across resolutions. This study suggests that instead of focusing solely on the finest image resolutions, multi-scale features should be emphasized for information extraction from high-resolution CXRs.

The effect of image resolution on convolutional neural networks in breast ultrasound

PurposeThe objective of this research was to investigate the efficacy of various parameter combinations of Convolutional Neural Networks (CNNs) models, namely MobileNet and DenseNet121, and different input image resolutions (REZs) ranging from 64×64 to 512×512 pixels, for diagnosing breast cancer. Materials and methodsDuring the period of June 2015 to November 2020, two hospitals were involved in the collection of two-dimensional ultrasound breast images for this retrospective multicenter study. The diagnostic performance of the computer models MobileNet and DenseNet 121 was compared at different resolutions. ResultsThe results showed that MobileNet had the best breast cancer diagnosis performance at 320×320pixel REZ and DenseNet121 had the best breast cancer diagnosis performance at 448×448pixel REZ. ConclusionOur study reveals a significant correlation between image resolution and breast cancer diagnosis accuracy. Through the comparison of MobileNet and DenseNet121, it is highlighted that lightweight neural networks (LW-CNNs) can achieve model performance similar to or even slightly better than large neural networks models (HW-CNNs) in ultrasound images, and LW-CNNs' prediction time per image is lower.

Minkowski functional evaluation of representative elementary volume of rock microtomography images at multiple resolutions

Pore-scale properties can be obtained by building a reliable digital twin of porous media through the digital rock physics (DRP) workflow. The two prerequisites of DRP are reliable imaging and computing power. Determining a proper image resolution that can reveal the actual pore-scale properties is challenging as there is a trade-off between the resolution and the representative elementary volume (REV). The REV is the smallest volume that reproduces the properties of the whole porous medium. The REV is a function of heterogeneities on the pore scale, the parameter of interest, and the scale range. Although the REV analysis for hydraulic properties is straightforward, it is computationally expensive. This study aims to estimate hydraulic pore-scale properties during REV evaluations by the geometric characterization of porous media using the Minkowski morphological functionals. Two sandstone and one carbonate rock samples were scanned at multiple imaging resolutions by both laboratory and synchrotron tomography. The REVs of various parameters of interest (porosity, permeability, surface area, tortuosity, Minkowski functionals) were computed, and the effect of image resolution and artificial rebinning on the final REV values was examined. After reaching the REV for porosity, the REV for the integrals of mean and total curvature agreed well with the permeability REV for large-volume image sizes. At constant porosity, the Minkowski integrals were found to be indicators for pore throat sizes. We also showed that the properties obtained from the rebinned (or coarsened) images differ entirely from that of actual scans at the same resolution.

Effect of Camera Resolution on the Determination of Mechanical and Multiple Crack Properties of Engineered Cementitious Composites via Digital Image Correlation

In this study, the effect of image resolution on the determination of mechanical parameters and multiple crack properties of Engineered Cementitious Composites (ECC) was investigated by digital image correlation (DIC). For this purpose, low and high resolution images were captured during the tensile loading of ECC by using two cameras simultaneously. DIC analyzes were performed on the images and results were compared both within themselves and with the traditional method. Mechanical parameters have been successfully calculated with DIC analysis and consistent values obtained with traditional methodology. However, high resolution images were found to be more effective for detecting pattern properties than low resolution images. It resulted in low strain error when using high resolution images compared to low resolution images in DIC analysis. Finally, when the high resolution images were used in DIC analysis as opposed to the low resolution images, both the detected number of cracks and calculated crack widths were obtained accurately. This situation was proven by local strain maps provided by DIC.

Accuracy of in vivo microCT imaging in assessing the microstructural properties of the mouse tibia subchondral bone.

Osteoarthritis (OA) is one of the most common musculoskeletal diseases. OA is characterized by degeneration of the articular cartilage as well as the underlying subchondral bone. Post-traumatic osteoarthritis (PTOA) is a subset of OA caused by mechanical trauma. Mouse models, such as destabilization of the medial meniscus (DMM), are useful to study PTOA. Ex vivo micro-Computed Tomography (microCT) imaging is the predominant technique used to scan the mouse knee in OA studies. Nevertheless, in vivo microCT enables the longitudinal assessment of bone microstructure, reducing measurement variability and number of animals required. The effect of image resolution in measuring subchondral bone parameters was previously evaluated only for a limited number of parameters. The aim of this study was to evaluate the ability of in vivo microCT imaging in measuring the microstructural properties of the mouse tibia trabecular and cortical subchondral bone, with respect to ex vivo high resolution imaging, in a DMM model of PTOA. Sixteen male C57BL/6J mice received DMM surgery or sham operation at 14 weeks of age (N=8 per group). The right knee of each mouse was microCT scanned in vivo (10.4μm voxel size) and ex vivo (4.35μm voxel size) at the age of 26 weeks. Each image was aligned to a reference image using rigid registration. The subchondral cortical bone plate thickness was measured at the lateral and medial condyles. Standard morphometric parameters were measured in the subchondral trabecular bone. In vivo microCT imaging led to significant underestimation of bone volume fraction (-14%), bone surface density (-3%) and trabecular number (-16%), whereas trabecular thickness (+3%) and separation (+5%) were significantly overestimated. Nevertheless, most trabecular parameters measured in vivo were well correlated with ex vivo measurements (R2 = 0.69-0.81). Degree of anisotropy, structure model index and connectivity density were measured in vivo with lower accuracy. Excellent accuracy was found for cortical thickness measurements. In conclusion, this study identified what bone morphological parameters can be reliably measured by in vivo microCT imaging of the subchondral bone in the mouse tibia. It highlights that this approach can be used to study longitudinal effects of diseases and treatments on the subchondral cortical bone and on most subchondral trabecular bone parameters, but systematic over- or under-estimations should be considered when interpreting the results.

Examining the performances of true color RGB bands from Landsat-8, Sentinel-2 and UAV as stand-alone data for mapping artisanal and Small-Scale Mining (ASM)

Conventional satellite sensors provide periodic data for monitoring Artisanal and Small-Scale Mining (ASM). Incorporating Unmanned Aerial Vehicle (UAV) sensors in monitoring ASM can increase data availability and also mitigate the challenge of cloud cover that causes data gaps in satellite data availability. However, most UAV sensors measure only in the Red-Green-Blue (RGB) bands and the utility of RGB for Land Use Land Cover (LULC) mapping is still in the preliminary stages. This study contributes to research by examining the accuracies of true color RGB images for mapping ASM, using the Birim Basin in Ghana as a case study. The study first compared Landsat 8 (L8) and Sentinel-2 (S2) full reflective bands for LULC mapping of the basin. The study then used true color RGB from UAV, L8 and S2 to map a subsection of the basin where there was active ASM. A test experiment was also performed in which other band combinations, single band, and indices were used to map ASM. From the results, the L8 and S2 full reflective bands attained Overall accuracy (OA) and User Accuracy (UA) over 97%, with less than 1% difference between their accuracies. All stand-alone RGB precisely mapped ASM with accuracies of 90% and above, indicating their fit for mapping ASM. However, the effect of image resolution caused differences in the surface area of the ASM estimated from the different sensors. Also, the different band combinations and indices could not accurately map ASM. Nonetheless, the results of the study showed increased scientific grounds for using UAV for mapping ASM. The overall results showed the capabilities of remote sensors in providing frequent and sustainable data for mapping ASM. Additionally, the results from the three images (L8, S2 and UAV) reavelaed much destruction of the forest cover due ASM. These results would be relevant in monitoring and managing the natural resources, which will help achieve environmental sustainability in meeting the SDG goals in 2030.

Remote sensing of forest diversities: the effect of image resolution and spectral plot extent

ABSTRACT Detecting field diversities via remote sensing is becoming important to monitor vegetation dynamics at large scale. The characteristics of the remotely sensed image, depending on the study organism and habitat, affect the efficiency of measuring α-and β-diversities. Therefore, we examined the impact of image resolutions and spectral plot extents on the accuracy of estimating forest α-diversities and compositional variances on the active volcano Mount Usu, northern Japan. Low- (3.2 m) and high-resolution (0.8 m) IKONOS multispectral images were used to create spectral indicators from pixels covering the field plots (narrow extent) and from pixels including neighbouring area (wide extent). Six forest diversity indices were obtained for canopy and for canopy-herb layer (total diversity): species richness (S), Shannon (H’), evenness (J’), Gini-Simpson (D), and true diversity of order 1 (N 1 = expH’) and order 2 (N 2 = 1/D). Changes in species composition were assessed by dissimilarity matrices. The spectral diversity indicators were calculated from the combination of image resolutions and spectral plot extents, and then compared with field diversities. The low-resolution–narrow extent based spectral indicators showed the highest correlations with canopy and total diversities. The best spectral indicators were derived from the scores of the first axis of principal component analysis and from the near infrared band, reaching high correlations with both canopy and total field diversity indices. Of the six field diversities, J’ showed the highest correlations with single spectral indicators, and N 1, 2 showed the highest correlations with pairs of spectral indicators. The correlations between spectral and field dissimilarities were lower than the correlations between α-diversities and spectral indicators, and were unaffected by the resolution and extent. In conclusion, the best spectral indicators were obtained from the low-resolution–narrow extent combination, and the indicators estimated canopy and total field diversity indices of temperate forests equally.

On the limits of finite element models created from (micro)CT datasets and used in studies of bone-implant-related biomechanical problems

Patient-specific approach is gaining a wide popularity in computational simulations of biomechanical systems. Simulations (most often based on the finite element method) are to date routinely created using data from imaging devices such as computed tomography which makes the models seemingly very complex and sophisticated. However, using a computed tomography in finite element calculations does not necessarily enhance the quality or even credibility of the models as these depend on the quality of the input images. Low-resolution (medical-)CT datasets do not always offer detailed representation of trabecular bone in FE models and thus might lead to incorrect calculation of mechanical response to external loading. The effect of image resolution on mechanical simulations of bone-implant interaction has not been thoroughly studied yet. In this study, the effect of image resolution on the modeling procedure and resulting mechanical strains in bone was analyzed on the example of cranial implant. For this purpose, several finite element models of bone interacting with fixation-screws were generated using seven computed tomography datasets of a bone specimen but with different image resolutions (ranging from micro-CT resolution of 25 μm to medical-CT resolution of 1250 μm). The comparative analysis revealed that FE models created from images of low resolution (obtained from medical computed tomography) can produce biased results. There are two main reasons: 1. Medical computed tomography images do not allow generating models with complex trabecular architecture which leads to substituting of the intertrabecular pores with a fictitious mass; 2. Image gray value distribution can be distorted resulting in incorrect mechanical properties of the bone and thus in unrealistic or even completely fictitious mechanical strains. The biased results of calculated mechanical strains can lead to incorrect conclusion, especially when bone-implant interaction is investigated. The image resolution was observed not to significantly affect stresses in the fixation screw itself; however, selection of bone material representation might result in significantly different stresses in the screw.

Measurement and Analysis of Depth Resolution Using Active Stereo Cameras

We propose a novel method to measure and analyze depth resolution in near range. Depth resolution is the smallest depth difference that can be detected by a depth camera, which is an important parameter but is hard to be measured accurately. Intuitively, a stair-shaped target is preferred to measure depth resolution but has many limitations. To overcome the difficulties, we use a flat target and move a depth camera at specified steps to form a stair-shaped depth map. Then, the depth map is compressed to one-dimensional profile by an average cross-section method. The 68-95-99.7 rule is applied for identifying the depth resolution. In order to obtain accurate data, a computer-controlled positioning platform is used to move or rotate a depth camera and a target in high precision. Three types of targets are made and tested, including a stair-shaped metal plate, a flat acrylic board, and a tensioned electric screen. A procedure is provided for aligning the optical plane of depth camera to the target before measuring. The Intel RealSense D400 series depth cameras are adopted for verification. Their theoretical depth resolution is derived and compared to the measured depth resolution. The effect of image resolution (x, y) on the depth resolution (z) is also analyzed. The experimental results offer useful information for researchers to design a depth camera and verify its depth resolution.

Resolving flame thickness using high-speed chemiluminescence imaging of OH* and CH* in spherically expanding methane–air flames

The coupling of CFD simulations with detailed chemical kinetics presents great progress in predicting the complex behavior of reacting flows, but also requires validated input parameters in the form of experimental data. The spatial profile of a combustion wave represents one such parameter, which can be directly measured using chemiluminescence imaging of a spherically expanding flame. In this work, emission signals from electronically excited methylidyne (CH*) and hydroxyl (OH*) radicals near 434 nm and 315 nm, respectively, from spherically expanding methane–air flames at 1 atm and 298 K were recorded for equivalence ratios of 0.8, 1.0, and 1.2. Spatial profiles of normalized intensity were compared to predicted profiles from AramcoMech2.0. The effect of image resolution was investigated by repeating experiments for three levels of image pixel density. An Abel inversion was employed to extract intensity profiles of CH* and OH* at flame radii up to 6.5 cm. Measured flame thickness increased as flames grew in size, but this behavior diminished as image resolution increased. A linear stretch correlation was used to extrapolate measured thicknesses to an unstretched thickness for each experimental condition. Radical-based flame thicknesses and corresponding spatial profiles were found to be highly dependent on image resolution, and at high resolution, measured flame thickness appeared to approach the kinetically predicted radical-based thicknesses. This paper lays the foundation for future, comprehensive measurements of spherical, laminar flames that can resolve the flame zone details to a level of precision not typically seen in the literature, providing benchmark data for both kinetics model validation and CFD model inputs. As a result, the measurements thus far indicate that the measured flame zone thickness based on electronically excited species is much closer to the length scale typically predicted by kinetics models than what has been seen in most experiments to date.

The effect of image resolution in the human presence detection: A case study on real-world image data

The automated operation of robots and flying drones is coupled to high security requirements with respect to humans and environment. Sometimes, persons have to be detected from a long distance or high altitude to allow the autonomous system an adequate and timely response. State-of-the-art Convolutional Neural Networks (CNNs) enable high object detection rates for different image data but only within their respective training, validation and test datasets. Recent studies show the limited generalization ability of CNNs for unknown data, even with merely small image changes. A typical source of such problems is the varying resolution of input images and the inevitable scaling of them to match the input-layer size of the network model. While modern cameras are able to capture high-resolution images of humans also from a longer distance, the practical input-layer sizes of networks are comparably small. Hence, we investigate the reliability of a network architecture for human detection with respect to such input-scaling effects. The popular VisDrone dataset with its varying image resolution and many relatively small depictions of humans is surveyed as well as the high-resolution AgriDrone image data from an agricultural context. Our results show that the object detection rate depends on the image scaling factor as well as on the relative size of persons. An enlarged input-layer size of the network can only partially contribute to counteract the observed effects. In addition, the detection algorithm becomes computationally more expensive by the increased effort.

The Effect of Image Resolution on Deep Learning in Radiography.

To examine variations of convolutional neural network (CNN) performance for multiple chest radiograph diagnoses and image resolutions. This retrospective study examined CNN performance using the publicly available National Institutes of Health chest radiograph dataset comprising 112 120 chest radiographic images from 30 805 patients. The network architectures examined included ResNet34 and DenseNet121. Image resolutions ranging from 32 × 32 to 600 × 600 pixels were investigated. Network training paradigms used 80% of samples for training and 20% for validation. CNN performance was evaluated based on area under the receiver operating characteristic curve (AUC) and label accuracy. Binary output networks were trained separately for each label or diagnosis under consideration. Maximum AUCs were achieved at image resolutions between 256 × 256 and 448 × 448 pixels for binary decision networks targeting emphysema, cardiomegaly, hernias, edema, effusions, atelectasis, masses, and nodules. When comparing performance between networks that utilize lower resolution (64 × 64 pixels) versus higher (320 × 320 pixels) resolution inputs, emphysema, cardiomegaly, hernia, and pulmonary nodule detection had the highest fractional improvements in AUC at higher image resolutions. Specifically, pulmonary nodule detection had an AUC performance ratio of 80.7% ± 1.5 (standard deviation) (0.689 of 0.854) whereas thoracic mass detection had an AUC ratio of 86.7% ± 1.2 (0.767 of 0.886) for these image resolutions. Increasing image resolution for CNN training often has a trade-off with the maximum possible batch size, yet optimal selection of image resolution has the potential for further increasing neural network performance for various radiology-based machine learning tasks. Furthermore, identifying diagnosis-specific tasks that require relatively higher image resolution can potentially provide insight into the relative difficulty of identifying different radiology findings. Supplemental material is available for this article. © RSNA, 2020See also the commentary by Lakhani in this issue.

Effects of Image Resolution on Sandstone Porosity and Permeability as Obtained from X-Ray Microscopy

Estimating porosity and permeability for porous rock is a vital component of reservoir engineering, and imaging techniques have to date focused on methodologies to match image-derived flow parameters with experimentally identified values. Less emphasis has been placed on the trade-off between imaging complexity, computational time, and error in identifying porosity and permeability. Here, the effect of image resolution on the permeability derived from micro-X-ray microscopy (micro-XRM) is discussed. A minicore plug of Bentheimer sandstone is imaged at a resolution of $$1024 \times 1024 \times 1024$$ voxels, with a voxel size of 1.53 $${\upmu }\hbox {m}$$ , and progressively rebinned to as low as 32 voxels per side (voxel size 48.96 $${\upmu }\hbox {m}$$ ). Pore-scale flow is modeled using the finite volume method in the open-source program OpenFOAM®. A sharp drop in permeability between images with a voxel size of 24 and 12 $${\upmu }\hbox {m}$$ suggests that an optimal speed/resolution trade-off may be found. The primary source of error is due to reassignment of voxels along the pore–solid interface and the subsequent change in pore connectivity. We observe the error in permeability and porosity due to both image resolution and thresholding values in order to find a method that balances an acceptable error range with reasonable computation time.

Contactless Monitoring of Breathing Patterns and Respiratory Rate at the Pit of the Neck: A Single Camera Approach

Vital signs monitoring is pivotal not only in clinical settings but also in home environments. Remote monitoring devices, systems, and services are emerging as tracking vital signs must be performed on a daily basis. Different types of sensors can be used to monitor breathing patterns and respiratory rate. However, the latter remains the least measured vital sign in several scenarios due to the intrusiveness of most adopted sensors. In this paper, we propose an inexpensive, off-the-shelf, and contactless measuring system for respiration signals taking as region of interest the pit of the neck. The system analyses video recorded by a single RGB camera and extracts the respiratory pattern from intensity variations of reflected light at the level of the collar bones and above the sternum. Breath-by-breath respiratory rate is then estimated from the processed breathing pattern. In addition, the effect of image resolution on monitoring breathing patterns and respiratory rate has been investigated. The proposed system was tested on twelve healthy volunteers (males and females) during quiet breathing at different sensor resolution (i.e., HD 720, PAL, WVGA, VGA, SVGA, and NTSC). Signals collected with the proposed system have been compared against a reference signal in both the frequency domain and time domain. By using the HD 720 resolution, frequency domain analysis showed perfect agreement between average breathing frequency values gathered by the proposed measuring system and reference instrument. An average mean absolute error (MAE) of 0.55 breaths/min was assessed in breath-by-breath monitoring in the time domain, while Bland-Altman showed a bias of −0.03 ± 1.78 breaths/min. Even in the case of lower camera resolution setting (i.e., NTSC), the system demonstrated good performances (MAE of 1.53 breaths/min, bias of −0.06 ± 2.08 breaths/min) for contactless monitoring of both breathing pattern and breath-by-breath respiratory rate over time.

Performance dependency of retinal image quality assessment algorithms on image resolution: analyses and solutions

Retinal image quality assessment (RIQA) is the first step performed in retinal image processing systems necessary to assure that the processed images are suitable for analysis and medical diagnosis. RIQA algorithms created for controlled environments can result in degraded performance for cross-dataset experiments in which the train and test images have different resolutions. The effect of image resolution on the performance of four different RIQA algorithms, chosen to include generic, segmentation, and transform-based quality features, is studied using datasets of various resolutions. Analyses showed that for cross-dataset classifications, the performance of some RIQA algorithms was reduced by up to 50% in cases where the train and test dataset image resolutions were significantly different. A statistical analysis was conducted to study how the retinal image quality features are affected by image resolution which resulted in their categorization into resolution-dependent and resolution-independent features. Feature scaling was then introduced to overcome the transform-based RIQA algorithm’s cross-dataset performance degradation resulting in a 100% performance enhancement. Based on this study, the investigation and enhancement of the cross-dataset performance of RIQA algorithms are recommended as a standard part of their design in order to assure their performance reliability in processing images of various resolutions.

Purwarupa Sistem Integrasi Metadata Foto Udara Dengan Data Sistem ADAHRS

This research was done by implementing aerial photographs metadata integrated system with ADAHRS data system. P hotos and ADAHRS data were taken along using the camera and ADAHRS devices were installed on UAV. D ata were taken from ADAHRS consisted of date, time, air speed, GPS latitud e- longitude, altitude, temperature, air pressure, roll, pitch, a nd yaw. The method in the system included custom tag configuration, extracted metadata, the synchronisation and integrated data between ADAHRS and photo metadata. C onfiguration, extraction and integration process were done using ExifTool. Datestamp and timestamp in photo metadata and ADAHRS data were used as a parameter of integrating data o n synchronisatio n . The systematical experiment was done by test on the effect of image resolution and number of photos to the duration of the whole process. This system was able to integrate ADAHRS data from image resolution, were 230 pixels x 1536 pixels, 1920 pixels x 1080 pixels, 1280 pixels x 720 pixels, and 848 pixels and 480 pixels. Amount of time average of process were 3,936 seconds/100 photos, 7,454 seconds/200 photos, 11,156 second/300 photos, 14,740 seconds/400 photos, and 18,460 seconds/500 photos. This system can be able to make a backup from extended version and retain the originality of photo metadata.