Manual Thresholding Research Articles

Applications of LIF to Document Natural Variability of Chlorophyll Content and Cu Uptake in Moss.

Chlorophyll has long been used as a natural indicator of plant health and photosynthetic efficiency. Laser-induced fluorescence (LIF) is an emerging technique for understanding broad spectrum organic processes and has more recently been used to monitor chlorophyll response in plants. Previous work has focused on developing a LIF technique for imaging moss mats to identify metal contamination with the current focus shifting toward application to moss fronds and aiding sample collection for chemical analysis. Two laser systems (CoCoBi a Nd:YGa pulsed laser system and Chl-SL with two blue continuous semiconductor diodes) were used to collect images of moss fronds exposed to increasing levels of Cu (1, 10, and 100 nmol/cm2) using a CMOS camera. The best methods for the preprocessing of images were conducted before the analysis of fluorescence signatures were compared to a control. The Chl-SL system performed better than the CoCoBi, with dynamic time warping (DTW) proving the most effective for image analysis. Manual thresholding to remove lower decimal code values improved the data distributions and proved whether using one or two fronds in an image was more advantageous. A higher DTW difference from the control correlated to lower chlorophyll a/b ratios and a higher metal content, indicating that LIF, with the aid of image processing, can be an effective technique for identifying Cu contamination shortly after an event.

Semi-automated software improves interrater reliability and reduces processing time of magnetic resonance imaging-based exocrine pancreatic assessments in pediatric patients

ObjectivesMagnetic resonance (MR) imaging with secretin stimulation (MR-PFTs) is a non-invasive test for pancreatic exocrine function based on assessing the volume of secreted bowel fluid in vivo. Adoption of this methodology in clinical care and research is largely limited to qualitative assessment of secretion as current methods for secretory response quantification require manual thresholding and segmentation of MR images, which can be time-consuming and prone to interrater variability. We describe novel software (PFTquant) that preprocesses and thresholds MR images, performs heuristic detection of non-bowel fluid objects, and provides the user with intuitive semi-automated tools to segment and quantify bowel fluid in a fast and robust manner. We evaluate the performance of this software on a retrospective set of clinical MRIs.MethodsTwenty MRIs performed in children (< 18 years) were processed independently by two observers using a manual technique and using PFTquant. Interrater agreement in measured secreted fluid volume was compared using intraclass correlation coefficients, Bland-Altman difference analysis, and Dice similarity coefficients.ResultsInterrater reliability of measured bowel fluid secretion using PFTquant was 0.90 (0.76–0.96 95% C.I.) with − 4.5 mL mean difference (-39.4–30.4 mL 95% limits of agreement) compared to 0.69 (0.36–0.86 95% C.I.) with − 0.9 mL mean difference (-77.3–75.5 mL 95% limits of agreement) for manual processing. Dice similarity coefficients were better using PFTquant (0.88 +/- 0.06) compared to manual processing (0.85 +/- 0.10) but not significantly (p = 0.11). Time to process was significantly (p < 0.001) faster using PFTquant (412 +/- 177 s) compared to manual processing (645 +/- 305 s).ConclusionNovel software provides fast, reliable quantification of secreted fluid volume in children undergoing MR-PFTs. Use of the novel software could facilitate wider adoption of quantitative MR-PFTs in clinical care and research.

GammaGateR: semi-automated marker gating for single-cell multiplexed imaging.

Multiplexed immunofluorescence (mIF) is an emerging assay for multichannel protein imaging that can decipher cell-level spatial features in tissues. However, existing automated cell phenotyping methods, such as clustering, face challenges in achieving consistency across experiments and often require subjective evaluation. As a result, mIF analyses often revert to marker gating based on manual thresholding of raw imaging data. To address the need for an evaluable semi-automated algorithm, we developed GammaGateR, an R package for interactive marker gating designed specifically for segmented cell-level data from mIF images. Based on a novel closed-form gamma mixture model, GammaGateR provides estimates of marker-positive cell proportions and soft clustering of marker-positive cells. The model incorporates user-specified constraints that provide a consistent but slide-specific model fit. We compared GammaGateR against the newest unsupervised approach for annotating mIF data, employing two colon datasets and one ovarian cancer dataset for the evaluation. We showed that GammaGateR produces highly similar results to a silver standard established through manual annotation. Furthermore, we demonstrated its effectiveness in identifying biological signals, achieved by mapping known spatial interactions between CD68 and MUC5AC cells in the colon and by accurately predicting survival in ovarian cancer patients using the phenotype probabilities as input for machine learning methods. GammaGateR is a highly efficient tool that can improve the replicability of marker gating results, while reducing the time of manual segmentation. The R package is available at https://github.com/JiangmeiRubyXiong/GammaGateR.

Nation-scale multidecadal shoreline extraction and coastal spatio-temporal change monitoring using cross-mission remote sensing data

Mapping shorelines is critical for coastal management and conservation, as shorelines are essential indicators for understanding coastal geomorphology and dynamics. Nonetheless, as shorelines are often lengthy and rugged, in-situ surveys are labor-intensive and costly. Oppositely, spaceborne remote sensing data provide cost-efficient alternatives, yet, most earlier studies rely on manual digitization or thresholding band ratio-based water indices to delineate instantaneous shorelines, which locations are affected by varying water color and tides. Consequently, cross-temporal assessments of environmental and anthropogenic impacts on coastal stability remain challenging. The present study develops an edge detection-based, dual-temporal, nationwide, tide-coordinated, and sediment cell-based shoreline dataset in Japan, i.e., the TSSD-JP. Both 1980s and 2020s shorelines along the Japanese coasts are extracted using cross-mission multispectral data and an automated shoreline extraction method. The results are validated with national coast surveys and global shoreline databases, confirming a positional accuracy better than 80% at 50 m. Moreover, founded on the extracted dual-temporal shorelines, 30+ year coastal changes are measured. An increased land area of 293.21 km2 at a rate of 8.50 km2/yr is estimated. Evident accretions in bay regions of southern Honshu due to extensive land reclamation and engineering constructions are observed. Since the analysis is conducted on natural management units with a physical basis, potential biases caused by jurisdictional boundaries are overcome. Practically, because TSSD-JP is derived from freely accessible data and tools, its universal applicability is guaranteed. Thence, the proposed approach can be applied to different nations' intertidal zones to assess the coastal vulnerability and resilience as well as support local governments’ shoreline management plans (SMPs).

Quantification of myocardial scar of different etiology using dark- and bright-blood late gadolinium enhancement cardiovascular magnetic resonance

Dark-blood late gadolinium enhancement (LGE) has been shown to improve the visualization and quantification of areas of ischemic scar compared to standard bright-blood LGE. Recently, the performance of various semi-automated quantification methods has been evaluated for the assessment of infarct size using both dark-blood LGE and conventional bright-blood LGE with histopathology as a reference standard. However, the impact of this sequence on different quantification strategies in vivo remains uncertain. In this study, various semi-automated scar quantification methods were evaluated for a range of different ischemic and non-ischemic pathologies encountered in clinical practice. A total of 62 patients referred for clinical cardiovascular magnetic resonance (CMR) were retrospectively included. All patients had a confirmed diagnosis of either ischemic heart disease (IHD; n = 21), dilated/non-ischemic cardiomyopathy (NICM; n = 21), or hypertrophic cardiomyopathy (HCM; n = 20) and underwent CMR on a 1.5 T scanner including both bright- and dark-blood LGE using a standard PSIR sequence. Both methods used identical sequence settings as per clinical protocol, apart from the inversion time parameter, which was set differently. All short-axis LGE images with scar were manually segmented for epicardial and endocardial borders. The extent of LGE was then measured visually by manual signal thresholding, and semi-automatically by signal thresholding using the standard deviation (SD) and the full width at half maximum (FWHM) methods. For all quantification methods in the IHD group, except the 6 SD method, dark-blood LGE detected significantly more enhancement compared to bright-blood LGE (p < 0.05 for all methods). For both bright-blood and dark-blood LGE, the 6 SD method correlated best with manual thresholding (16.9% vs. 17.1% and 20.1% vs. 20.4%, respectively). For the NICM group, no significant differences between LGE methods were found. For bright-blood LGE, the 5 SD method agreed best with manual thresholding (9.3% vs. 11.0%), while for dark-blood LGE the 4 SD method agreed best (12.6% vs. 11.5%). Similarly, for the HCM group no significant differences between LGE methods were found. For bright-blood LGE, the 6 SD method agreed best with manual thresholding (10.9% vs. 12.2%), while for dark-blood LGE the 5 SD method agreed best (13.2% vs. 11.5%). Semi-automated LGE quantification using dark-blood LGE images is feasible in both patients with ischemic and non-ischemic scar patterns. Given the advantage in detecting scar in patients with ischemic heart disease and no disadvantage in patients with non-ischemic scar, dark-blood LGE can be readily and widely adopted into clinical practice without compromising on quantification.

Deep learning-based automated kidney and cyst segmentation of autosomal dominant polycystic kidney disease using single vs. multi-institutional data

PurposeThis study aimed to investigate if a deep learning model trained with a single institution's data has comparable accuracy to that trained with multi-institutional data for segmenting kidney and cyst regions in magnetic resonance (MR) images of patients affected by autosomal dominant polycystic kidney disease (ADPKD). MethodsWe used TensorFlow with a Keras custom UNet on 2D slices of 756 MRI images of kidneys with ADPKD obtained from four institutions in the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) study. The ground truth was determined via a manual plus global thresholding method. Five models were trained with 80 % of all institutional data (n = 604) and each institutional data (n = 232, 172, 148, or 52), respectively, and validated with 10 % and tested on an unseen 10 % of the data. The model's performance was evaluated using the Dice Similarity Coefficient (DSC). ResultsThe DSCs by the model trained with all institutional data ranged from 0.92 to 0.95 for kidney image segmentation, only 1–2 % higher than those by the models trained with single institutional data (0.90–0.93).In cyst segmentation, however, the DSCs by the model trained with all institutional data ranged from 0.83 to 0.89, which were 2–20 % higher than those by the models trained with single institutional data (0.66–0.86). ConclusionThe UNet performance, when trained with a single institutional dataset, exhibited similar accuracy to the model trained on a multi-institutional dataset. Segmentation accuracy increases with models trained on larger sample sizes, especially in more complex cyst segmentation.

Standardized and semiautomated workflow for 3D characterization of liberated particles

MSPaCMAn is the recently developed workflow that does the mineralogical quantification of individual particles using its histograms while considering the effects of partial volume artefacts in interphases at particle level detail. This paper demonstrates and validates the new developments in the MSPaCMAn workflow, aiming to minimize user bias and enhance the accuracy of MSPaCMAn.Here, in the new developments of MSPaCMAn workflow, firstly, the recently developed deep learning method, namely ParticleSeg3D, is employed to distinguish particles from the background. Secondly, the particle's size and shape information are considered along with its histogram to classify and quantify the mineral phases in liberated particles. After the new developments, the detection limit of MSPaCMAn to characterize small and thin liberated particles is enhanced. Experimental results demonstrate the effectiveness of the MSPaCMAn's updated workflow. It was found that the mineralogical composition calculated by the MSPaCMAn's updated workflow was precise, with the highest coefficient of variance of 6.56%. Moreover, the mineralogical composition determined by MSPaCMAn's updated workflow had low variance across three different reconstruction parameters and two different voxel sizes (5.5 μm and 10 μm). Comparisons with other quantification methods highlight the accuracy of MSPaCMAn's updated workflow to determine the mineralogical compositions. When analyzing a test sample consisting of quartz, calcite, fluorite, and lepidolite, MSPaCMAn's updated workflow achieved the highest mineralogical deviation of only 7.73% from the reference mineralogy. In contrast, the random forest algorithm resulted in a deviation of 51.47%, while the manual thresholding method yielded a deviation of 41.46%.Overall, these findings emphasize the reliability and accuracy of MSPaCMAn's updated workflow in quantifying mineralogical composition.

Cardiac magnetic field map topology quantified by Kullback-Leibler entropy identifies patients with clinically suspected myocarditis.

Myocarditis is a condition that can have severe adverse outcomes and lead to sudden cardiac death if remaining undetected. This study tested the capability of cardiac magnetic field mapping to detect patients with clinically suspected myocarditis. This could open up the way for rapid, non-invasive, and cost-effective screening of suspected cases before a gold standard assessment via endomyocardial biopsy. Historical cardiac magnetic field maps (n = 97) and data from a state-of-the-art magnetocardiography device (n = 30) were analyzed using the Kullback-Leibler entropy (KLE) for dimensionality reduction and topological quantification. Linear discriminant analysis was used to discern between patients with ongoing myocarditis and healthy controls. The STT segment of a magnetocardiogram, i.e., the section between the end of the S wave and the end of the T wave, was best suited to discern both groups. Using a 250-ms excerpt from the onset of the STT segment gave a reliable classification between the myocarditis and control group for both historic data (sensitivity: 0.83, specificity: 0.85, accuracy: 0.84) and recent data (sensitivity: 0.69, specificity: 0.88, accuracy: 0.80) using the KLE to quantify the topology of the cardiac magnetic field map. The implementation based on KLE can reliably distinguish between clinically suspected myocarditis patients and healthy controls. We implemented an automatized feature selection based on LDA to replace the observer-dependent manual thresholding in previous studies.

Comparing ovarian expression of sperm acrosome associated 3 protein in young and adult queens

The purpose of this research was to quantify sperm acrosome associated 3 protein expression in the ovaries of young (3.0 ± 0.9 months, n = 11) and adult (10.4 ± 2.8 months, n = 11) queens. Immunohistochemistry was performed on formalin-fixed, paraffin-embedded feline ovarian sections. Ovaries were obtained following routine ovariohysterectomy of queens. Cellular expression of sperm acrosome associated 3 protein was measured in primordial, primary, secondary, and tertiary follicles using an image-analysis software's red, green, and blue stack and manual thresholding functions. The oocyte nucleus, ooplasm, granulosa cells, and theca cells were outlined using the freehand selection tool and mean grey value was recorded. Results from each cellular location were compared between age groups using a Student's t-test and between follicle stages using an analysis of variance. Compared to adult queens, younger queens had significantly greater sperm acrosome associated 3 protein expression in granulosa cells of primary, secondary, and tertiary follicles. Also, theca cells of secondary and tertiary follicles had significantly greater sperm acrosome associated 3 protein expression in younger queens compared to adult queens. The oocyte nucleus of primordial, primary, and secondary follicles had significantly greater sperm acrosome associated 3 protein expression in younger queens compared to adult queens. However, sperm acrosome associated 3 protein expression within the ooplasm did not differ significantly between age groups of any follicle type. More research is needed to determine what role sperm acrosome associated 3 protein may play in female fertility in animals as well as what mechanisms regulate ovarian sperm acrosome associated 3 protein expression over time.

UCTNet with Dual-Flow Architecture: Snow Coverage Mapping with Sentinel-2 Satellite Imagery

Satellite remote sensing (RS) has been drawing considerable research interest in land-cover classification due to its low price, short revisit time, and large coverage. However, clouds pose a significant challenge, occluding the objects on satellite RS images. In addition, snow coverage mapping plays a vital role in studying hydrology and climatology and investigating crop disease overwintering for smart agriculture. Distinguishing snow from clouds is challenging since they share similar color and reflection characteristics. Conventional approaches with manual thresholding and machine learning algorithms (e.g., SVM and Random Forest) could not fully extract useful information, while current deep-learning methods, e.g., CNNs or Transformer models, still have limitations in fully exploiting abundant spatial/spectral information of RS images. Therefore, this work aims to develop an efficient snow and cloud classification algorithm using satellite multispectral RS images. In particular, we propose an innovative algorithm entitled UCTNet by adopting a dual-flow structure to integrate information extracted via Transformer and CNN branches. Particularly, CNN and Transformer integration Module (CTIM) is designed to maximally integrate the information extracted via two branches. Meanwhile, Final Information Fusion Module and Auxiliary Information Fusion Head are designed for better performance. The four-band satellite multispectral RS dataset for snow coverage mapping is adopted for performance evaluation. Compared with previous methods (e.g., U-Net, Swin, and CSDNet), the experimental results show that the proposed UCTNet achieves the best performance in terms of accuracy (95.72%) and mean IoU score (91.21%) while with the smallest model size (3.93 M). The confirmed efficiency of UCTNet shows great potential for dual-flow architecture on snow and cloud classification.

#4694 AUTOMATIC SEGMENTATION OF GLOMERULAR SUBSTRUCTURES BY DEEP LEARNING

Abstract Background and Aims Electron microscopy (EM) complements light microscopy (LM) evaluation of the kidney biopsy. Foot process effacement, as assessed by EM, helps in diagnosing podocytopathies. However, human interpretation of EM images is time-intensive and often subjective. In this pilot, we investigate how deep learning techniques can help in adequate segmentation of the glomerular basement membranes (GBM) and podocytes in EM images. The ultimate goal would be to design an automatic tool for reliable and fast assessment of foot process effacement. Method Podocytes and glomerular basement membranes (GBM) of 10 patients, 5 with podocyte disease and 5 without glomerular changes (where biopsy showed tubulointerstitial pathology on LM) were annotated with a combination of manual annotation and thresholding for providing ground truth. After data preprocessing including splitting, flipping, rotating and tiling, a modified U-net architecture was applied (‘baseline model’). This baseline model was compared to a combination of U-net and a self-supervised contrastive learning approach with pretraining on 100 additional images (SimCLR framework, ‘fine-tuned model’). Segmentation performance was measured by IoU score. Results Segmentation of the glomerular basement membrane was best achieved by the baseline model and resulted in an IoU score of 0.711±0.089 compared to a IoU score of 0.675 ±0.093 in the fine-tuned model Segmentation of the podocytes was most successful in the fine-tuned model, with a IoU score of 0.609±0.118 compared to a IoU score of 0.591 ±0.118 in the baseline model Conclusion This study pioneers in segmenting glomerular substructures on EM images by means of a modified U-net architecture. The next step is training and validation in larger datasets. Data annotation remains a challenge. Inclusion of more images is expected to greatly improve the performance of the model. Guideline 1500-3600 characters

An integrated convolutional neural network for classifying small pulmonary solid nodules.

Achieving accurate classification of benign and malignant pulmonary nodules is essential for treating some diseases. However, traditional typing methods have difficulty obtaining satisfactory results on small pulmonary solid nodules, mainly caused by two aspects: (1) noise interference from other tissue information; (2) missing features of small nodules caused by downsampling in traditional convolutional neural networks. To solve these problems, this paper proposes a new typing method to improve the diagnosis rate of small pulmonary solid nodules in CT images. Specifically, first, we introduce the Otsu thresholding algorithm to preprocess the data and filter the interference information. Then, to acquire more small nodule features, we add parallel radiomics to the 3D convolutional neural network. Radiomics can extract a large number of quantitative features from medical images. Finally, the classifier generated more accurate results by the visual and radiomic features. In the experiments, we tested the proposed method on multiple data sets, and the proposed method outperformed other methods in the small pulmonary solid nodule classification task. In addition, various groups of ablation experiments demonstrated that the Otsu thresholding algorithm and radiomics are helpful for the judgment of small nodules and proved that the Otsu thresholding algorithm is more flexible than the manual thresholding algorithm.

Prognostic Value of Different Thresholds for Myocardial Scar Quantification on Cardiac MRI Late Gadolinium Enhancement Images in Patients Receiving Implantable Cardioverter Defibrillators

Purpose To compare the predictive value of different myocardial scar quantification thresholds using cardiac MRI for appropriate implantable cardioverter defibrillator (ICD) shock and mortality. Materials and Methods In this retrospective, two-center observational cohort study, patients with ischemic or nonischemic cardiomyopathy underwent cardiac MRI prior to ICD implantation. Late gadolinium enhancement (LGE) was first determined visually and then quantified by blinded cardiac MRI readers using different SDs above the mean signal of normal myocardium, full-width half-maximum, and manual thresholding. The intermediate signal “gray zone” was calculated as the differences between different SDs. Results Among 374 consecutive eligible patients (mean age, 61 years ± 13 [SD]; mean left ventricular ejection fraction, 32% ± 14; secondary prevention, 62.7%), those with LGE had a higher rate of appropriate ICD shock or death than those without (37.5% vs 26.6%, log-rank P = .04) over a median follow-up of 61 months. In multivariable analysis, none of the thresholds for quantifying scar was a significant predictor of mortality or appropriate ICD shock, while the extent of gray zone was an independent predictor (adjusted hazard ratio per 1 g = 1.025; 95% CI: 1.008, 1.043; P = .005) regardless of the presence or absence of ischemic heart disease (P interaction = .57). Model discrimination was highest for the model incorporating the gray zone (between 2 SD and 4 SD). Conclusion Presence of LGE was associated with a higher rate of appropriate ICD shock or death. Although none of the scar quantification techniques predicted outcomes, the gray zone both in infarct and nonischemic scar was an independent predictor and may refine risk stratification. Keywords: MRI, Scar Quantification, Implantable Cardioverter Defibrillator, Sudden Cardiac Death Supplemental material is available for this article. © RSNA, 2023

An image-based technique for automated root disease severity assessment using PlantCV.

Plant disease severity assessments are used to quantify plant-pathogen interactions and identify disease-resistant lines. One common method for disease assessment involves scoring tissue manually using a semi-quantitative scale. Automating assessments would provide fast, unbiased, and quantitative measurements of root disease severity, allowing for improved consistency within and across large data sets. However, using traditional Root System Markup Language (RSML) software in the study of root responses to pathogens presents additional challenges; these include the removal of necrotic tissue during the thresholding process, which results in inaccurate image analysis. Using PlantCV, we developed a Python-based pipeline, herein called RootDS, with two main objectives: (1) improving disease severity phenotyping and (2) generating binary images as inputs for RSML software. We tested the pipeline in common bean inoculated with Fusarium root rot. Quantitative disease scores and root area generated by this pipeline had a strong correlation with manually curated values (R 2 = 0.92 and 0.90, respectively) and provided a broader capture of variation than manual disease scores. Compared to traditional manual thresholding, images generated using our pipeline did not affect RSML output. Overall, the RootDS pipeline provides greater functionality in disease score data sets and provides an alternative method for generating image sets for use in available RSML software.

Feature Based Techniques for a Face Recognition using Supervised Learning Algorithms based on Fixed Monocular Video Camera

Automatic face recognition has ample significance in biometric research. Recent decades have witnessed enormous growth in this research area. Face-based identification is always considered more expedient as compared to other biometric authentications owing to its uniqueness and wide acceptance. The major contribution of this work is twofold; firstly, it comprises an extension of manual thresholding feature-based face recognition approach to an automatic feature-based supervised learning face recognition. Secondly, various new feature sets are proposed and tested on several classifiers for 2, 3, 4, and 5 persons. In addition, the use of slope features of facial components, such as the nose, right eye, left eye, and lips along with other conventional features for face recognition is also a unique contribution of this research. Multiple experiments were performed on the UMT face database. The results demonstrated a comparison of 5 different sets of feature-based approaches on 7 classifiers using the metrics of time efficiency and accuracy. They also depicted that the proposed approaches achieve a percentage accuracy of up to 95.5%.

Automatic lesion detection and segmentation in 18F-flutemetamol positron emission tomography images using deep learning

BackgroundBeta amyloid in the brain, which was originally confirmed by post-mortem examinations, can now be confirmed in living patients using amyloid positron emission tomography (PET) tracers, and the accuracy of diagnosis can be improved by beta amyloid plaque confirmation in patients. Amyloid deposition in the brain is often associated with the expression of dementia. Hence, it is important to identify the anatomically and functionally meaningful areas of the human brain cortex surface using PET to diagnose the possibility of developing dementia. In this study, we demonstrated the validity of automated 18F-flutemetamol PET lesion detection and segmentation based on a complete 2D U-Net convolutional neural network via masking treatment strategies.MethodsPET data were first normalized by volume and divided into five amyloid accumulation zones through axial, coronary, and thalamic slices. A single U-Net was trained using a divided dataset for one of these zones. Ground truth segmentations were obtained by manual delineation and thresholding (1.5 × background).ResultsThe following intersection over union values were obtained for the various slices in the verification dataset: frontal lobe axial/sagittal: 0.733/0.804; posterior cingulate cortex and precuneus coronal/sagittal: 0.661/0.726; lateral temporal lobe axial/coronal: 0.864/0.892; parietal lobe axial/coronal: 0.542/0.759; and striatum axial/sagittal: 0.679/0.752. The U-Net convolutional neural network architecture allowed fully automated 2D division of the 18F-flutemetamol PET brain images of Alzheimer's patients.ConclusionsAs dementia should be tested and evaluated in various ways, there is a need for artificial intelligence programs. This study can serve as a reference for future studies using auxiliary roles and research in Alzheimer's diagnosis.

Detecting Vegetation to Open Water Transitions in a Subtropical Wetland Landscape from Historical Panchromatic Aerial Photography and Multispectral Satellite Imagery

Over the last century, direct human modification has been a major driver of coastal wetland degradation, resulting in widespread losses of wetland vegetation and a transition to open water. High-resolution satellite imagery is widely available for monitoring changes in present-day wetlands; however, understanding the rates of wetland vegetation loss over the last century depends on the use of historical panchromatic aerial photographs. In this study, we compared manual image thresholding and an automated machine learning (ML) method in detecting wetland vegetation and open water from historical panchromatic photographs in the Florida Everglades, a subtropical wetland landscape. We compared the same classes delineated in the historical photographs to 2012 multispectral satellite imagery and assessed the accuracy of detecting vegetation loss over a 72 year timescale (1940 to 2012) for a range of minimum mapping units (MMUs). Overall, classification accuracies were &gt;95% across the historical photographs and satellite imagery, regardless of the classification method and MMUs. We detected a 2.3–2.7 ha increase in open water pixels across all change maps (overall accuracies &gt; 95%). Our analysis demonstrated that ML classification methods can be used to delineate wetland vegetation from open water in low-quality, panchromatic aerial photographs and that a combination of images with different resolutions is compatible with change detection. The study also highlights how evaluating a range of MMUs can identify the effect of scale on detection accuracy and change class estimates as well as in determining the most relevant scale of analysis for the process of interest.

Urban tree generator: spatio-temporal and generative deep learning for urban tree localization and modeling

We present a vision-based algorithm that uses spatio-temporal satellite imagery, pattern recognition, procedural modeling, and deep learning to perform tree localization in urban settings. Our method resolves two primary challenges. First, automated city-scale tree localization at high accuracy typically requires significant acquisition/user intervention. Second, vegetation-index segmentation methods from satellites require manual thresholding, which varies across geographic areas, and are not robust across cities with varying terrain, geometry, altitude, and canopy. In our work, we compensate for the lack of visual detail by using satellite snapshots across twelve months and segment cities into various vegetation clusters. Then, we use multiple GAN-based networks to plant trees by recognizing placement patterns inside segmented regions procedurally. We present comprehensive experiments over four cities (Chicago, Austin, Indianapolis, Lagos), achieving tree count accuracies of 87–97%. Finally, we show that the knowledge accumulated from each model (trained on a particular city) can be transferred to a different city.

Segmentation methods and morphometry of confocal microscopy imaged corneal epithelial cells

PurposeTo develop and explore automated cell identification and segmentation methods for morphometry of confocal microscopy imaged corneal epithelial cells using ImageJ software. MethodsIn vivo confocal microscopy images of the intermediate (wing) and basal cell layers of the central and peripheral corneas of 20 healthy participants were analysed. The intermediate and basal cell areas obtained using the two new techniques (i.e., manual- and auto- thresholding) were compared with the widely used manual tracing technique. A predefined range of epithelial cell morphometric parameters was used as image descriptors to improve cell identification and segmentation. ResultsThe mean intermediate cell area obtained using the manual tracing (central; 120 ± 14 µm2, peripheral; 123 ± 15 µm2) was statistically similar (p > 0.05) to the manual thresholding (central; 119 ± 7 µm2, peripheral; 119 ± 8) but not with the auto thresholding technique (central; 101 ± 8 µm2, peripheral; 101 ± 7 µm2). Bland-Altman limits of agreement for the mean difference (measurement bias) in central and peripheral intermediate cell area determined via manual tracing and manual thresholding techniques were 1 µm2 (+25 to − 23 µm2) and 4 µm2 (+29.8 to − 21.9 µm2). There were statistically significant differences in basal cell area between the three methods. ConclusionThe manual thresholding technique may be used for automated identification and segmentation of corneal epithelial intermediate cells (central and peripheral) for assessing various morphometric parameters. However, measurement of the corneal epithelial basal cells is less reliable using thresholding techniques.

A Machine Learning Perspective on fNIRS Signal Quality Control Approaches.

Despite a rise in the use of functional Near Infra-Red Spectroscopy (fNIRS) to study neural systems, fNIRS signal processing is not standardized and is highly affected by empirical and manual procedures. At the beginning of any signal processing procedure, Signal Quality Control (SQC) is critical to prevent errors and unreliable results. In fNIRS analysis, SQC currently relies on applying empirical thresholds to handcrafted Signal Quality Indicators (SQIs). In this study, we use a dataset of fNIRS signals (N = 1,340) recorded from 67 subjects, and manually label the signal quality of a subset of segments (N = 548) to investigate the pitfalls of current practices while exploring the opportunities provided by Deep Learning approaches. We show that SQIs statistically discriminate signals with bad quality, but the identification by means of empirical thresholds lacks sensitivity. Alternatively to manual thresholding, conventional machine learning models based on the SQIs have been proven more accurate, with end-to-end approaches, based on Convolutional Neural Networks, capable of further improving the performance. The proposed approach, based on machine learning, represents a more objective SQC for fNIRS and moves towards the use of fully automated and standardized procedures.