Automatic Thresholding Algorithms Research Articles

Evolution of morphology and microstructure of coarsened nanoporous gold studied by automatic thresholding and image recognition algorithms

The morphology and microstructure parameters of nanoporous gold (NPG) during thermal coarsening were investigated by automatic thresholding and image recognition algorithms. The ligaments width distributions, pore width distributions and curvature distributions are determined to quantitatively characterize microstructure of NPG. The results show a clear temperature dependence of these parameters. The measured activation energy (∼100 kJ/mol) for coarsening process supports the surface diffusion mechanism during ligaments coarsening of NPG. It was also found that the ligament and pore size are topological equivalent. Because of the dangling ligaments formed by pinch-off process have convex curvature, the ratio of positive curvature is higher than that of negative curvature. This structural geometry and microstructure characteristics analysis of NPG could help researchers better understanding the structure evolution, physical and chemical properties.

Automated lesion segmentation and quantification for prediction of paradoxical worsening in patients with tubercular serpiginous-like choroiditis

To develop and evaluate a fully automated pipeline that analyzes color fundus images in patients with tubercular serpiginous-like choroiditis (TB SLC) for prediction of paradoxical worsening (PW). In this retrospective study, patients with TB SLC with a follow-up of 9 months after initiation of anti-tubercular therapy were included. A fully automated custom-designed pipeline was developed which was initially tested using 12 baseline color fundus photographs for assessment of repeatability. After confirming reliability using Bland–Altman plots and intraclass correlation coefficient (ICC), the pipeline was deployed for all patients. The images were preprocessed to exclude the optic nerve from the fundus photo using a single-shot trainable WEKA segmentation algorithm. Two automatic thresholding algorithms were applied, and quantitative metrics were generated. These metrics were compared between PW + and PW- groups using non-parametric tests. A logistic regression model was used to predict probability of PW for assessing binary classification performance and receiver operator curves were generated to choose a sensitivity-optimized threshold. The study included 139 patients (139 eyes; 92 males and 47 females; mean age: 44.8 ± 11.3 years) with TB SLC. Pilot analysis of 12 images showed an excellent ICC for measuring the mean area, intensity, and integrated pixel intensity (all ICC > 0.89). The PW + group had significantly higher mean lesion area (p = 0.0152), mean pixel intensity (p = 0.0181), and integrated pixel intensity (p < 0.0001) compared to the PW- group. Using a sensitivity optimized threshold cut-off for mean pixel intensity, an area under the curve of 0.87 was achieved (sensitivity: 96.80% and specificity: 72.09%). Automated calculation of lesion metrics such as mean pixel intensity and segmented area in TB SLC is a novel approach with good repeatability in predicting PW during the follow-up.

Comparison of thresholding methods for shoreline extraction from Sentinel-2 and Landsat-8 imagery: Extreme Lake Salda, track of Mars on Earth

Lake Salda's extreme environment is geologically similar to Jezero Crater paleolake on Mars due to the formation of stromatolites and its extremely alkaline and cold water. It is critical to accurately determine the shoreline in the littoral zone where stromatolite formation occurs, in alkaline clean lakes like Salda, which contain traces of life on Mars, and in monitoring the change that occurs with climate and anthropogenic effect. The performance of global automatic thresholding algorithms on shoreline determination from NDWI and mNDWI water indices is compared in this study using Sentinel-2 and Landsat-8 images atmospherically corrected by different algorithms. Satellite images data acquired on August 12, 2020 for Sentinel-2 and August 11, 2020 for Landsat-8 on Lake Salda were used to determining the shoreline. The shoreline data measured in situ concurrently with the Sentinel-2 satellite acquisition was used as reference data. In the accuracy analysis, ground control points created inside and outside the lake at a distance of 1 pixel and 0.5 pixel to the reference shoreline for each satellite image were used. The performance of the optimal threshold values determined by each thresholding algorithm in the water index images was assessed using Kappa coefficient, Overall Accuracy (OA), %OA of Inside (%OAinside) and %OA of Outside (%OAoutside) statistics metrics. The optimal threshold values vary depending on the image and the atmospheric correction algorithm applied to the image. The NDWI index produces more accurate results in both Sentinel-2 and Landsat-8 satellite images. While atmospheric correction algorithms do not affect the results in Landsat-8 images, the Sen2Cor algorithm outperforms iCOR in Sentinel-2 images. For thresholding algorithms to be used in different water index and satellite images, Intermode, Isodata, IJ_Isodata, Minimum and Otsu algorithms in Landsat-8_LaSRC_NDWI and Landsat-8_iCOR_NDWI images, and Intermode, Minimum and Huang algorithms in Sentinel-2_Sen2Cor_NDWI images produce the best results. Because the Minimum algorithm causes significant gaps in the lake surface, the Huang and Intermode algorithms should be used for Sentinel-2_Sen2Cor_NDWI images. The 0 (zero) threshold value in the water indices images has a high accuracy only in the NDWI water indices generated from the Landsat-8 image.

Digital three-dimensional visualization of intrabony periodontal defects for regenerative surgical treatment planning

BackgroundIn the regenerative treatment of intrabony periodontal defects, surgical strategies are primarily determined by defect morphologies. In certain cases, however, direct clinical measurements and intraoral radiographs do not provide sufficient information on defect morphologies. Therefore, the application of cone-beam computed tomography (CBCT) has been proposed in specific cases. 3D virtual models reconstructed with automatic thresholding algorithms have already been used for diagnostic purposes. The aim of this study was to utilize 3D virtual models, generated with a semi-automatic segmentation method, for the treatment planning of minimally invasive periodontal surgeries and to evaluate the accuracy of the virtual models, by comparing digital measurements to direct intrasurgical measurements.MethodsFour patients with a total of six intrabony periodontal defects were enrolled in the present study. Two months following initial periodontal treatment, a CBCT scan was taken. The novel semi-automatic segmentation method was performed in an open-source medical image processing software (3D Slicer) to acquire virtual 3D models of alveolar and dental structures. Intrasurgical and digital measurements were taken, and results were compared to validate the accuracy of the digital models. Defect characteristics were determined prior to surgery with conventional diagnostic methods and 3D virtual models. Diagnostic assessments were compared to the actual defect morphology during surgery.ResultsDifferences between intrasurgical and digital measurements in depth and width of intrabony components of periodontal defects averaged 0.31 ± 0.21 mm and 0.41 ± 0.44 mm, respectively. In five out of six cases, defect characteristics could not be assessed precisely with direct clinical measurements and intraoral radiographs. 3D models generated with the presented semi-automatic segmentation method depicted the defect characteristics correctly in all six cases.ConclusionIt can be concluded that 3D virtual models acquired with the described semi-automatic segmentation method provide accurate information on intrabony periodontal defect morphologies, thus influencing the treatment strategy. Within the limitations of this study, models were found to be accurate; however, further investigation with a standardized validation process on a large number of participants has to be conducted.

A versatile technique for high-resolution three-dimensional imaging of human arterial segments using microcomputed tomography

BackgroundQuantitative methods for evaluating microstructure of arterial specimens typically rely on histologic techniques that involve random sampling, which cannot account for the unique spatial distribution of features in three dimensions. MethodsTo overcome this limitation, we demonstrate a nondestructive method for three-dimensional imaging of intact human blood vessels using microcomputed tomography (microCT). Human artery segments were dehydrated and stained in an iodine solution then imaged with a standard laboratory microCT scanner. Image visualization and segmentation was performed using commercially available and open source software. ResultsStaining of cadaveric vessels with iodine enabled clear visualization of the arterial wall with microCT, preserved tissue morphology, and generated high-resolution images with a voxel size of 5.4 μm. Various components of the arterial wall were segmented using a combination of manual and automatic thresholding algorithms. ConclusionsOur approach allows for spatial mapping of human artery tissue samples that can guide targeted histologic analysis of smaller tissue segments, provide geometric data to inform finite element models, quantify degree of atherosclerosis, and help to evaluate the foreign body response to intravascular medical implants. (JVS–Vascular Science 2021;2:13-19.) Clinical RelevanceIn this article, we describe a powerful technique for whole artery analysis of pathologic human tissue specimens that provides high-resolution spatial detail regarding composition of the blood vessel wall. The protocol described here is a valuable adjunct that can be used as a research tool to inform finite element modeling of arteries, quantify pathologic response (ie, neointimal hyperplasia and vascular calcification), and evaluate the tissue/device interface of implanted medical devices.

Progressive Thresholding: Shaping and Specificity in Automated Neurofeedback Training.

Neurofeedback has long been proposed as a promising form of adjunctive non-pharmaceutical treatment for a variety of neuropsychological disorders. However, there is much debate over its efficacy and specificity. Many suggest that specificity can only be achieved when a specially trained clinician manually updates reward thresholds that indicate to the trainee when they are modulating their brain activity correctly, during training. We present a novel fully automated reward thresholding algorithm called progressive thresholding and test it with a frontal alpha asymmetry neurofeedback protocol. Progressive thresholding uses dynamic difficulty tuning and individual-specific progress models to simulate the shaping a clinician might perform when setting reward thresholds manually. We demonstrate in a double-blind comparison that progressive thresholding leads to significantly better learning outcomes compared with current automatic reward thresholding algorithms.

An Energy-Based Model Encoding Nonlocal Pairwise Pixel Interactions for Multisensor Change Detection

Image change detection (CD) is a challenging problem, particularly when images come from different sensors. In this paper, we present a novel and reliable CD model, which is first based on the estimation of a robust similarity-feature map generated from a pair of bitemporal heterogeneous remote sensing images. This similarity-feature map, which is supposed to represent the difference between the multitemporal multisensor images, is herein defined, by specifying a set of linear equality constraints, expressed for each pair of pixels existing in the before-and-after satellite images acquired through different modalities. An estimation of this overconstrained problem, also formulated as a nonlocal pairwise energy-based model, is then carried out, in the least square sense, by a fast linear-complexity algorithm based on a multidimensional scaling mapping technique. Finally, the fusion of different binary segmentation results, obtained from this similarity-feature map by different automatic thresholding algorithms, allows us to precisely and automatically classify the changed and unchanged regions. The proposed method is tested on satellite data sets acquired by real heterogeneous sensor, and the results obtained demonstrate the robustness of the proposed model compared with the best existing state-of-the-art multimodal CD methods recently proposed in the literature.

Threshold-based segmentation of fluorescent and chromogenic images of microglia, astrocytes and oligodendrocytes in FIJI

BackgroundImage segmentation is often imperfect, particularly in complex image sets such z-stack micrographs of slice cultures and there is a need for sufficient details of parameters used in quantitative image analysis to allow independent repeatability and appraisal. New methodFor the first time, we have critically evaluated, quantified and validated the performance of different segmentation methodologies using z-stack images of ex vivo glial cells. The BioVoxxel toolbox plugin, available in FIJI, was used to measure the relative quality, accuracy, specificity and sensitivity of 16 global and 9 local threshold automatic thresholding algorithms. ResultsAutomatic thresholding yields improved binary representation of glial cells compared with the conventional user-chosen single threshold approach for confocal z-stacks acquired from ex vivo slice cultures. The performance of threshold algorithms varies considerably in quality, specificity, accuracy and sensitivity with entropy-based thresholds scoring highest for fluorescent staining. Comparison with existing methodsWe have used the BioVoxxel toolbox to correctly and consistently select the best automated threshold algorithm to segment z-projected images of ex vivo glial cells for downstream digital image analysis and to define segmentation quality. The automated OLIG2 cell count was validated using stereology. ConclusionsAs image segmentation and feature extraction can quite critically affect the performance of successive steps in the image analysis workflow, it is becoming increasingly necessary to consider the quality of digital segmenting methodologies. Here, we have applied, validated and extended an existing performance-check methodology in the BioVoxxel toolbox to z-projected images of ex vivo glia cells.

Fully convolutional neural network for removing background in noisy images of uranium bearing particles.

A fully convolutional neural network (FCN) was developed to supersede automatic or manual thresholding algorithms used for tabulating SIMS particle search data. The FCN was designed to perform a binary classification of pixels in each image belonging to a particle or not, thereby effectively removing background signal without manually or automatically determining an intensity threshold. Using 8000 images from 28 different particle screening analyses, the FCN was trained to accurately predict pixels belonging to a particle with near 99% accuracy. Background eliminated images were then segmented using a watershed technique in order to determine isotopic ratios of identified particles. A comparison of the isotopic distributions of an independent data set segmented using the neural network with a commercially available automated particle measurement (APM) program developed by CAMECA was performed. This comparison highlighted the necessity for effective background removal to ensure that resulting particle identification is not only accurate, but preserves valuable signal that could be lost due to improper segmentation. The FCN approach improves the robustness of current state-of-the-art particle searching algorithms by reducing user input biases, resulting in an improved absolute signal per particle and decreased uncertainty of the determined isotope ratios.

Standardizing the protocol for hemispherical photographs: accuracy assessment of binarization algorithms.

Hemispherical photography is a well-established method to optically assess ecological parameters related to plant canopies; e.g. ground-level light regimes and the distribution of foliage within the crown space. Interpreting hemispherical photographs involves classifying pixels as either sky or vegetation. A wide range of automatic thresholding or binarization algorithms exists to classify the photographs. The variety in methodology hampers ability to compare results across studies. To identify an optimal threshold selection method, this study assessed the accuracy of seven binarization methods implemented in software currently available for the processing of hemispherical photographs. Therefore, binarizations obtained by the algorithms were compared to reference data generated through a manual binarization of a stratified random selection of pixels. This approach was adopted from the accuracy assessment of map classifications known from remote sensing studies. Percentage correct () and kappa-statistics () were calculated. The accuracy of the algorithms was assessed for photographs taken with automatic exposure settings (auto-exposure) and photographs taken with settings which avoid overexposure (histogram-exposure). In addition, gap fraction values derived from hemispherical photographs were compared with estimates derived from the manually classified reference pixels. All tested algorithms were shown to be sensitive to overexposure. Three of the algorithms showed an accuracy which was high enough to be recommended for the processing of histogram-exposed hemispherical photographs: “Minimum” ( 98.8%; 0.952), “Edge Detection” ( 98.1%; 0.950), and “Minimum Histogram” ( 98.1%; 0.947). The Minimum algorithm overestimated gap fraction least of all (11%). The overestimation by the algorithms Edge Detection (63%) and Minimum Histogram (67%) were considerably larger. For the remaining four evaluated algorithms (IsoData, Maximum Entropy, MinError, and Otsu) an incompatibility with photographs containing overexposed pixels was detected. When applied to histogram-exposed photographs, these algorithms overestimated the gap fraction by at least 180%.

A comparative study of automatic thresholding approaches for 3D x‐ray microtomography of trabecular bone

This paper presents a comparative study of automatic thresholding algorithms for segmenting trabecular bone volume in x-ray microtomography (μCT). First, a preprocessing stage was established, which considered noise reduction by applying anisotropic diffusion filtering and contrast enhancement by using morphological top-hats. Next, four automatic thresholding algorithms were implemented: clustering, maximum entropy, moment preservation, and concavity-based. These approaches analyze the preprocessed 3D μCT image histogram to optimize some parameters to find the best gray-level threshold. Thirty-eight vertebra bone samples were acquired from 19 normal Wistar rats, specifically the L3 and L4 vertebrae. The μCT images were acquired with a microfocus x-ray device at 100 slices/sample. Next, three human operators segmented the entire 3D μCT images manually to establish ground-truth segmentations so as to associate the segmentation problem with perceptual grouping. The normalized probabilistic Rand index (NPRI) was used to quantify the agreement between each computerized segmentation and the corresponding set of three ground-truth segmentations. Hence, the NPRI value should tend toward unity for an acceptable performance. Finally, a statistical analysis was done to determine which thresholding approach achieved the best performance. Besides, 3D morphometric indices were also measured. The Games-Howell test (α = 0.05) was used to compare the equality of means from the NPRI results considering the four thresholding algorithms (multiple comparisons). This statistical analysis indicated that the clustering and moment preservation techniques performed similarly, with NPRI values of 0.594 ± 0.126 and 0.607 ± 0.127, respectively. The main advantage of computerized segmentation is that it is fully automatic; that is, no interaction with the user is required. Thus, the method could be considered objective. Besides, the proposed preprocessing stage plays an important role in enhancing the μCT image quality to achieve better separation between the background volume and the trabecular bone volume.

Twelve automated thresholding methods for segmentation of PET images: a phantom study

Tumor volume delineation over positron emission tomography (PET) images is of great interest for proper diagnosis and therapy planning. However, standard segmentation techniques (manual or semi-automated) are operator dependent and time consuming while fully automated procedures are cumbersome or require complex mathematical development. The aim of this study was to segment PET images in a fully automated way by implementing a set of 12 automated thresholding algorithms, classical in the fields of optical character recognition, tissue engineering or non-destructive testing images in high-tech structures. Automated thresholding algorithms select a specific threshold for each image without any a priori spatial information of the segmented object or any special calibration of the tomograph, as opposed to usual thresholding methods for PET. Spherical 18F-filled objects of different volumes were acquired on clinical PET/CT and on a small animal PET scanner, with three different signal-to-background ratios. Images were segmented with 12 automatic thresholding algorithms and results were compared with the standard segmentation reference, a threshold at 42% of the maximum uptake. Ridler and Ramesh thresholding algorithms based on clustering and histogram-shape information, respectively, provided better results that the classical 42%-based threshold (p < 0.05). We have herein demonstrated that fully automated thresholding algorithms can provide better results than classical PET segmentation tools.

Comparison of automatic and interactive thresholding of hemispherical photography

This study presents the effects of operator bias and variation in interactive thresholding on the estimation of light environment using hemispherical photography. Twenty-one hemispherical photographs taken beneath a wide range of canopy densities were visually converted to binary images twice by 21 operators, and then the gap fraction was computed from the images. The interactive threshold varied greatly among the different operators and within a single operator, which resulted in a considerable operator bias and variation in the gap fraction. This study also compared three widely used automatic thresholding algorithms, which were installed in freely available software LIA for Win32 for analyzing hemispherical photography, with interactive thresholding using the same photographs. The median of the interactive threshold by repetitive interactive thresholdings from 21 operators was assumed to be correct for the comparison. The results indicated that MINIMUM was considered to be a better algorithm than the other ones installed in LIA32 when the gap fraction was over 10%. However, VARIANCE seemed to be superior to MINIMUM under the low gap fraction and the cloudy sky condition with dark and white clouds. This implied that MINIMUM or VARIANCE should be used for analyzing hemispherical photographs with LIA32. In conclusion, we need to pay attention to the selection of the automatic thresholding algorithm and the sky condition when taking hemispherical photographs.

Observer-dependent variability of the thresholding step in the quantitative analysis of soil images and X-ray microtomography data

For the investigation of many geometrical features of soils, computer-assisted image analysis has become a method of choice over the last few decades. This analysis involves numerous steps, regarding which subjective decisions have to be made by the individuals conducting the research. This is particularly the case with the thresholding step, required to transform the original (color or greyscale) images into the type of binary representation (e.g., pores in white, solids in black) needed for fractal analysis or simulation with Lattice–Boltzmann models. Limited information exists at present on whether different observers, analyzing the same soil, would be likely to obtain similar results. In this general context, the first objective of the research reported in this article was to determine, through a so-called “round-robin” test, how much variation exists among the outcomes of various image thresholding strategies (including any image pre-treatment deemed appropriate), routinely adopted by soil scientists. Three test images – of a field soil, a soil thin section, and a virtual section through a 3-dimensional CT data set – were thresholded by 13 experts, worldwide. At the same time, variability of the outcomes of a set of automatic thresholding algorithms, applied to portions of the test images, was also investigated. The experimental results obtained illustrate the fact that experts rely on very different approaches to threshold images of soils, and that there is considerable observer influence associated with this thresholding. This observer dependence is not likely to be alleviated by adoption of one of the many existing automatic thresholding algorithms, many of which produce thresholded images that are equally, or even more, variable than those of the experts. These observations suggest that, at this point, analysis of the same image of a soil, be it a simple photograph or 3-dimensional X-ray CT data, by different individuals can lead to very different results, without any assurance that any of them would be even approximately “correct” or best suited to the objective at hand. Different strategies are proposed to cope with this situation, including the use of physical “standards”, adoption of procedures to assess the accuracy of thresholding, benchmarking with physical measurements, or the development of computational methods that do not require binary images.

Evaluation of biofilm image thresholding methods

To evaluate biomass distribution in heterogeneous biofilms from their microscope images, it is often necessary to perform image thresholding by converting the gray-scale images to binary images consisting of a foreground of biomass material and a background of interstitial space. The selection of the gray-scale intensity used for thresholding is arbitrary but under the control of the operator, which may produce unacceptable levels of variability among operators. The quality of numerical information extracted from the images is diminished by such variability, and it is desirable to find a method that improves the reproducibility of thresholding operations. Automatic methods of thresholding provide this reproducibility, but often at the expense of accuracy, as they consistently set thresholds that differ significantly from what human operators would choose. The performance of five automatic image thresholding algorithms was tested in this study: (1) local entropy; (2) joint entropy; (3) relative entropy; (4) Renyi's entropy; and (5) iterative selection. Only the iterative selection method was satisfactory in that it was consistently setting the threshold level near that set manually. The extraction of feature information from biofilm images benefits from automatic thresholding and can be extended to other fields, such as medical imaging.

Comments on gray-level thresholding of images using a correlation criterion

In our comments we deal with two automatic thresholding algorithms proposed by Otsu and Brink, respectively. These algorithms, in spite of their different approach, lead to one and the same function to be maximized.