Visual Qualitative Evaluation Research Articles

Volumetric apparent diffusion coefficient histogram analysis in term neonatal asphyxia treated with hypothermia.

Our aim is to estimate the long-term neurological sequelae and prognosis in term neonatal asphyxia treated with hypothermia via volumetric apparent diffusion coefficient (ADC) map histogram analysis (HA). Brain MRI studies of eighty-three term neonates with asphyxia who received whole-body hypothermia treatment and examined between postnatal (PN) 4 to 6th days were retrospectively re-evaluated by two radiologists. Volumetric HA was performed for the areas frequently affected in deep and superficial asphyxia (thalamus, lentiform nucleus, posterior limb of internal capsule, corpus callosum forceps major and perirolandic cortex-subcortical white matter) on ADC map. The quantitative ADC values were obtained separately for each region. Qualitative-visual (conventional) MRI findings were also re-evaluated. Neonates were examined neurodevelopmentally according to the Revised Brunet-Lezine scale. The distinguishability of long-term neurodevelopmental outcomes was statistically investigated. With HA, the adverse neurodevelopmental outcomes could only be distinguished from mild-moderated impairment and normal development at the thalamus with 10th percentile ADC (p = 0.02 and p = 0.03; respectively) and ADCmin (p = 0.03 and p = 0.04; respectively). Also with the conventional MRI findings, adverse outcome could be distinguished from mild-moderated impairment (p = 0.04) and normal development (p = 0.04) via cytotoxic edema of the thalamus, corpus striatum and diffuse cerebral cortical. The long-term adverse neurodevelopmental outcomes in newborns with asphyxia who received whole-body hypothermia treatment can be estimate similarly with volumetric ADC-HA and the conventional assessment of the ADC map. This study compare early MRI ADC-HA with neurological sequelae in term newborns with asphyxia who received whole-body hypothermia treatment. We couldn't find any significant difference in predicting adverse neurological sequelae between the visual-qualitative evaluation of the ADC map and HA.

Diagnosis of cervical lymphoma using a YOLO-v7-based model with transfer learning

To investigate the ability of an auxiliary diagnostic model based on the YOLO-v7-based model in the classification of cervical lymphadenopathy images and compare its performance against qualitative visual evaluation by experienced radiologists. Three types of lymph nodes were sampled randomly but not uniformly. The dataset was randomly divided into for training, validation, and testing. The model was constructed with PyTorch. It was trained and weighting parameters were tuned on the validation set. Diagnostic performance was compared with that of the radiologists on the testing set. The mAP of the model was 96.4% at the 50% intersection-over-union threshold. The accuracy values of it were 0.962 for benign lymph nodes, 0.982 for lymphomas, and 0.960 for metastatic lymph nodes. The precision values of it were 0.928 for benign lymph nodes, 0.975 for lymphomas, and 0.927 for metastatic lymph nodes. The accuracy values of radiologists were 0.659 for benign lymph nodes, 0.836 for lymphomas, and 0.580 for metastatic lymph nodes. The precision values of radiologists were 0.478 for benign lymph nodes, 0.329 for lymphomas, and 0.596 for metastatic lymph nodes. The model effectively classifies lymphadenopathies from ultrasound images and outperforms qualitative visual evaluation by experienced radiologists in differential diagnosis.

Evaluation of alternate image capturing devices for recording fingermarks at crime scenes

ABSTRACT This project evaluated the fingermark imaging ability of several alternate image capturing devices selected for their potential to assist with the recording of fingermarks in difficult-to-reach locations commonly encountered at crime scenes, e.g. the inside surface of door handles. Four small form-factor non-DSLR camera systems — intraoral camera, borescope, fibre-optic camera and an iPhone — were evaluated under controlled conditions for their ability to image fingermarks with sufficient detail for identification. Evaluation was conducted using black powdered fingermarks and standard photographic test targets. A qualitative visual evaluation approach was also developed to gauge whether image quality was appropriate for fingermark identification. The results obtained indicated that most of the systems were capable of capturing fingermark details suitable for identification in controlled environments; however, the inherent contrast of the fingermark had an impact on the success of various imaging devices. The practicality and capability of alternate camera devices for fingermark recovery in pseudo-operational conditions is part of ongoing research.

Aesthetic caries infiltration – Long-term masking efficacy after 6 years

ObjectivesThis study aimed to evaluate the masking efficacy of caries infiltration technique of initial caries lesions (ICL) six years after debonding and single treatment. MethodsIn 10 adolescents, 74 ICL (ICDAS 2) in 74 teeth were treated by resin infiltration (Icon, DMG) at a mean (SD) of 1.2 (1.2) months after bracket removal. The etching procedure was performed up to 3 times. Standardized digital images were taken before treatment (T0), seven days (T7) and 6 years (T2190) after treatment. Outcomes included the evaluation of the color differences between carious and healthy enamel at T0, T7 and T2190 by quantitative colorimetric analysis (ΔE), ICDAS scores, quantitative light-induced fluorescence (QLF; ΔF,ΔQ,WS Area) and qualitative visual evaluation (5-point Likert-scale [deteriorated (1), unchanged (2), improved, but not satisfying (3), improved and no further treatment required (4), completely masked (5)). ResultsThe median color difference ΔΕ0 (25th/75th percentiles) at T0 was 10.3 (8.56/13.0). At T7 a significant decrease was observed (ΔΕ7=3.7 (2.0/5.8); p<0.001; Friedmann-test; ICDAS p<0.001; Chi-square test). No significant changes based on ΔΕ (p=0.972; Friedmann-test) and ICDAS grading (p=0.511, chi-square test) were observed between T7 and T2190 (ΔΕ2190=2.9 (1.8/4.2)). Furthermore, at T2190 four experienced dentists classified 50% and 37% of the lesions as “improved and no further treatment required” and “completely masked”, respectively (Fleiss kappa: T2190: 0.782 (substantial agreement)). ConclusionAesthetic caries infiltration can effectively mask post-orthodontic initial caries lesions for at least 6 years. These results for most of the teeth could not only be observed by quantitative but also by qualitative analysis. Clinical significanceResin infiltration efficaciously masks post-orthodontic initial carious lesions. The optical improvement can be observed directly after treatment and remains stable for at least six years.

VLSI Architecture of Decision Based Adaptive Denoising Filter for Removing Salt &amp; Pepper Noise

A new Decision Based Adaptive Denoising Filter (DBADF) algorithm &amp; hardware architecture are proposed for restoring the digital image that is highly corrupted with impulse noise. The proposed DBADF detects only the corrupted pixels and that pixel is restored by the noise-free median value or previous value based upon the noise density in the image. The proposed DBADF uses a window initially and adaptively goes up to window based on the noise corruption more than 50% by impulse noise in the current processing window. The proposed architecture was found to exhibit better visual qualitative and quantitative evaluation based on PSNR, IEF, EKI, SSIM, FOM, and error rate. The DBAMF architecture also preserves the original information of digital image with a high density of salt &amp; pepper noise, when compared to many standard conventional algorithms. The proposed architecture has been simulated using the VIRTEX7 FPGA device and the reported maximum post place and route frequency are 149.995MHz and the dynamic power consumption is 179mW.

Maskless 2-Dimensional Digital Subtraction Angiography Generation Model for Abdominal Vasculature using Deep Learning

PurposeTo develop a deep learning (DL) model to generate synthetic, 2-dimensional subtraction angiograms free of artifacts from native abdominal angiograms. Materials and MethodsIn this retrospective study, 2-dimensional digital subtraction angiography (2D-DSA) images and native angiograms were consecutively collected from July 2019 to March 2020. Images were divided into motion-free (training, validation, and motion-free test datasets) and motion-artifact (motion-artifact test dataset) sets. A total of 3,185, 393, 383, and 345 images from 87 patients (mean age, 71 years ± 10; 64 men and 23 women) were included in the training, validation, motion-free, and motion-artifact test datasets, respectively. Native angiograms and 2D-DSA image pairs were used to train and validate an image-to-image translation model to generate synthetic DL-based subtraction angiography (DLSA) images. DLSA images were quantitatively evaluated by the peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) using the motion-free dataset and were qualitatively evaluated via visual assessments by radiologists with a numerical rating scale using the motion-artifact dataset. ResultsThe DLSA images showed a mean PSNR (± standard deviation) of 43.05 dB ± 3.65 and mean SSIM of 0.98 ± 0.01, indicating high agreement with the original 2D-DSA images in the motion-free dataset. Qualitative visual evaluation by radiologists of the motion-artifact dataset showed that DLSA images contained fewer motion artifacts than 2D-DSA images. Additionally, DLSA images scored similar to or higher than 2D-DSA images for vascular visualization and clinical usefulness. ConclusionsThe developed DL model generated synthetic, motion-free subtraction images from abdominal angiograms with similar imaging characteristics to 2D-DSA images.

Imaging Evaluation of Intervertebral Disc Degeneration and Painful Discs-Advances and Challenges in Quantitative MRI.

In recent years, various quantitative and functional magnetic resonance imaging (MRI) sequences have been developed and used in clinical practice for the diagnosis of patients with low back pain (LBP). Until now, T2-weighted imaging (T2WI), a visual qualitative evaluation method, has been used to diagnose intervertebral disc (IVD) degeneration. However, this method has limitations in terms of reproducibility and inter-observer agreement. Moreover, T2WI observations do not directly relate with LBP. Therefore, new sequences such as T2 mapping, T1ρ mapping, and MR spectroscopy have been developed as alternative quantitative evaluation methods. These new quantitative MRIs can evaluate the anatomical and physiological changes of IVD degeneration in more detail than conventional T2WI. However, the values obtained from these quantitative MRIs still do not directly correlate with LBP, and there is a need for more widespread use of techniques that are more specific to clinical symptoms such as pain. In this paper, we review the state-of-the-art methodologies and future challenges of quantitative MRI as an imaging diagnostic tool for IVD degeneration and painful discs.

Electromyographic analysis of maximal voluntary contraction of female pelvic floor muscles: Intrarater and interrater reliability study.

To investigate the intrarater and interrater reliability of the surface electromyography (sEMG) of female pelvic floor muscles (PFM) by visual determination of the onset-offset during maximal voluntary contraction (MVC). Intrarater and interrater reliability study. Three MVC, recorded for 5 s (s) each and separated by intervals of 60 s, were requested. Two independent raters visually determined the onset and offset of each MVC. The agreement between raters on the quality of the signals was analyzed both qualitatively (visual inspection) and quantitatively (signal-to-noise ratio, SNR). The reliability of integral and median frequency (MDF) of sEMG signals during MVC was analyzed using intraclass correlation coefficient (ICC) models (ICC2,1 , ICC2,3 ) and repeated measures analysis of variance. A total of 58 women (mean age of 54 years) and two independent raters participated in the analyses. Of the total, 22 (37.9%) were considered of high quality by both raters and the SNR ranged 34.3-22.2 decibels (dB). The agreement of qualitative visual evaluation of the sEMG signal was fair (Cohen's κ = 0.306 [0.148; 0.463]). Intrarater and interrater reliability were excellent for the average of the three MVC, both for integral (ICC2,3 = 0.884 [0.821; 0.925]; ω2 = -0.006, p = 0.558) and MDF (ICC2,3 = 0.998 [0.998; 0.999]; ω2 = -0.009, p = 0.992). Qualitative visual analysis of the sEMG signal presents fair interrater agreement and reflects the SNR. Visual determination of the onset and offset of the MVC of PFM using the standard interface of the sEMG software has excellent reliability for determination of muscle activation variables.

Comparison of inner ear MRI enhancement in patients with Meniere’s disease after intravenous injection of gadobutrol, gadoterate meglumine, or gadodiamide

PurposeTo compare the enhancement results of three gadolinium contrast agents in the inner ear of patients with Meniere’s disease 4 h after intravenous injection of gadobutrol, gadoterate meglumine, or gadodiamide. MethodsWe enrolled 60 patients with a definitive diagnosis of unilateral Meniere’s disease and divided them into three groups of 20 patients; each group received a double dose of gadobutrol, gadoterate meglumine, or gadodiamide. The postcontrast signal intensity of the basal cochlear turn was scored quantitatively, and qualitative visual evaluation of the cochlea, vestibule and semi-circular canals was also performed. The results of both evaluations were compared between the three patient groups. ResultsThe cochlear basal turn signal intensity of the gadobutrol group was significantly higher than that of the gadoterate meglumine and gadodiamide groups; however, no significant difference was observed between the gadoterate meglumine and gadodiamide groups. The intensity of visualization of the semi-circular canals was significantly better in the bilateral gadobutrol group than in the gadoterate meglumine and gadodiamide groups; however, there was no significant difference in terms of the intensity of visualization of the semi-circular canals between the gadoterate meglumine and gadodiamide groups. There were no significant differences in the intensity of visualization of the cochlea and vestibule among the three groups. ConclusionsCompared with gadoterate meglumine and gadodiamide, gadobutrol can provide a higher degree of perilymphatic enhancement and better anatomical details of the semi-circular canals in the ears of patients with Meniere’s disease.

Asymmetric CycleGAN for image-to-image translations with uneven complexities

CycleGAN is one of the famous and basic methods for unpaired image-to-image translation tasks. Inspired by the experiments of the NIR-RGB translation, which is a kind of translation where images are translated from simple to complex or vice versa, we concluded the definition of asymmetric translation task. Because of the complexity difference between two domains, the complexity inequality in bidirectional translations is significant. We analyzed and witnessed the limitation of the original CycleGAN in asymmetric translation tasks and proposed an Asymmetric CycleGAN model with generators of unequal sizes to adapt to the asymmetric need in asymmetric translations. An empirical metric was also given to determine the asymmetric task from the aspect of image entropy and could be treated as the auxiliary guidance to design the asymmetric generators. Besides, the edge-retain loss between the input and the generated images was introduced to enhance the structural visual quality. Residual-block-net based and U-net based generators were both applied here to verify the Asymmetric CycleGAN. The performance of different depth of generators for Asymmetric CycleGAN was also discussed on the basis of experiments. The qualitative visual evaluation demonstrated that our model had achieved great improvements compared to original CycleGAN.

Quantification of Osteosarcoma Mineralization on Plain Radiographs - Novel Software Applications to Assess Response to Chemotherapy.

Osteosarcoma is the most common primary malignant bone tumor in children. Patient survival with osteosarcoma is heavily influenced by the response to chemotherapy, measured by tumor necrosis upon histological analysis. Unfortunately, response is not measurable until the time of surgery and therefore modifications to chemotherapy protocol are only made after several weeks of treatment and surgery. Osteosarcoma tumors often demonstrate increased mineralization following the onset of chemotherapy. Furthermore, it has been hypothesized that this mineralization-apparent on radiographs-may correlate with chemotherapy response, however, this has not been demonstrated with qualitative visual evaluation. The ability to non-invasively measure a patient's response to chemotherapy using plain radiographs, which is currently included in the normal clinical workflow, would guide the medical oncologists to tailor treatment for patients with osteosarcoma. After obtaining appropriate multi-center institutional review board approvals, we identified 31patients that possess a pair of pre-and post-chemotherapy radiograph along with the necrosis measure. The images were digitized scans of physical radiographs between 1999 and 2013. Software was designed to measure the signal intensities in the tumor, a region of the soft tissue, air, and healthy bone. The tumor signals were normalized based on the random combination of air, soft tissue or bone, by subtraction or division. The differences in tumor signal between pre-and post-image were plotted against the percent necrosis determined by histological analysis. Different combinations of the normalization methods were compared 2based on the slope, coefficient of determination (R2) and Pearson correlation coefficient (ρ).

Functional aromatic polyamides for the preparation of coated fibres as smart labels for the visual detection of biogenic amine vapours and fish spoilage

We have prepared high-performance functional aromatic polyamides with sensory pendant groups toward amines. These polymers are colourless. The pendant groups have bromonaphthalimide motifs that react with amines rendering coloured arylamines by nucleophilic aromatic substitution of the halide. Thus, these materials are colorimetric sensors toward amines. In the interest of saving costs, while having a sensory material with a high specific surface, cotton fabrics were coated, in order to render coated fibres as smart colorimetric labels toward amines. Moreover, as fish spoilage by microorganism increases the biogenic amines in food, we have applied the smart labels to visually follow the food spoilage. Also, a picture taken to the films allowed to obtain, in seconds, the digital colour definition parameters (RGB) that were correlated with the results of expensive and time-consuming conventional techniques used to obtain relevant food quality data, such as total amine concentration (treatment of the sample and HPLC), colony-forming unit (microbiological assays), total volatile basic nitrogen (TVB-N), and organoleptic test (sensory test). The smart labels are inexpensive, granting in seconds the visual qualitative evaluation of the food quality, or even quantitative, and they comply with the European food contact materials regulation.

Application of a multi-material artifact reduction algorithm in a wide-detector CT in the evaluation of the portal venous angiography of postoperative TIPS and embolization

Objective: To assess the effect of monochromatic images and metal artifact reduction (MAR) on the image quality of spectral CT portal venous angiography in patients with operation of after the performing transjugular intrahepatic portosystemic stent shunt(TIPS) and embolization. Methods: From December 2017 to April 2018, the examination data of 28 patients with portal hypertension due to cirrhosis who underwent portal vein angiography 1 month after TIPS and embolization were prospectively collected. After spectral CT scanning in revolution CT, the monochromatic energy levels(60 keV, 120 keV), 60 keV + 120 keV, 120kV-like + 120 keV fused images combined with MAR algorithm were reconstructed. Quantitative parameters such as image artifact index (AI) and qualitative visual evaluation scores were recorded and compared. Results: The 120 keV monochromatic images showed the lowest AI value(30.8±8.5, 18.2±4.3) and highest metal artifacts reduction effect. The 60 keV monochromatic images showed the highest AI value (57.3±15.7, 32.1±7.9) and the lowest metal artifacts reduction effect. The AI value of 60 keV + 120 keV fused images was lower than that of 60 keV images(26.2%, 24.7%). The difference of AI value between each group was statistically significant(all P<0.05). The interobserver agreement in the subjective image scores was moderate with kappa value of 0.824. The overall image quality score of 60 keV + 120 keV fused image and the noise score of 120 kV-like+120 keV were higher than the remaining groups. The differences of the subjective scores among each group were statistically significant(all P<0.05). Conclusion: The spectral CT with MAR algorithm can effectively improve the image quality of portal vein angiography after the TIPS and embolization therapy and the 60 keV + 120 keV fused images can eliminate artifacts and ensure a clear display of blood vessels.

Visual qualitative evaluation of diffusion-weighted imaging, apparent diffusion coefficient quantification and Ki-67 proliferation index for predicting atypia in surgical meningiomas

Object. The goal in this study was to investigate the efficiency of qualitative evaluation using DW imaging, ADC quantification, and Ki-67 proliferation labeling index in grading intracranial meningiomas compared to our reference standard histopathology.Methods. From our database, we enrolled 35 consecutive patients with pathologic report of a meningioma, conventional MR imaging, contrast-enhanced MR imaging, DW imaging and ADC maps.Results. Eight patients (23%) had atypical meningioma (5 women and 3 men; mean age: 57.1±15.8 years); 27 (77%) had low-grade-typical (19 women, 8 men; mean age: 54.3±14.0 years). The most frequent histological subtype was the meningothelial type in both low-grade (37%) and atypical meningiomas (59%). The mean Ki-67 proliferation index for low-grade and atypical meningiomas were 2.31±1.44% (range: 1-5) and 7.37 ± 2.72% (range: 3-10), respectively. Meningiomas with Ki-67 proliferation index higher than 4% (except one) were atypical (p&amp;lt;0.001). In visual qualitative evaluation; diffusion restriction was significantly more common in atypical meningiomas (p&amp;lt;0.05). In quantitative evaluation, benign and atypical tumor groups had no significant difference in mean ADC values (p=0.471). Conclusion. We suggest that visual qualitative evaluation of DW imaging may be helpful to predict atypia and the risk of recurrence in patients with surgically treated meningiomas.

Evaluation of Penalized-Likelihood Estimation Reconstruction on a Digital Time-of-Flight PET/CT Scanner for 18F-FDG Whole-Body Examinations.

The resolution and quantitative accuracy of PET are highly influenced by the reconstruction method. Penalized-likelihood estimation algorithms allow for fully convergent iterative reconstruction, generating a higher image contrast than ordered-subsets expectation maximization (OSEM) while limiting noise. In this study, a type of penalized reconstruction known as block-sequential regularized expectation maximization (BSREM) was compared with time-of-flight OSEM (TOF OSEM). Various strengths of noise penalization factor β were tested along with various acquisition durations and transaxial fields of view (FOVs) with the aim of evaluating the performance and clinical use of BSREM for 18F-FDG PET/CT, both quantitatively and in a qualitative visual evaluation. Methods: Eleven clinical whole-body 18F-FDG PET/CT examinations acquired on a digital TOF PET/CT scanner were included. The data were reconstructed using BSREM with point-spread function recovery and β-factors of 133, 267, 400, and 533-and using TOF OSEM with point-spread function-for various acquisition times per bed position and various FOVs. Noise level, signal-to-noise ratio (SNR), signal-to-background ratio (SBR), and SUV were analyzed. A masked evaluation of visual image quality, rating several aspects, was performed by 2 nuclear medicine physicians to complement the analysis. Results: The lowest levels of noise were reached with the highest β-factor, resulting in the highest SNR, which in turn resulted in the lowest SBR. A β-factor of 400 gave noise equivalent to TOF OSEM but produced a significant increase in SUVmax (11%), SNR (22%), and SBR (12%). BSREM with a β-factor of 533 at a decreased acquisition duration (2 min/bed position) was comparable to TOF OSEM at a full acquisition duration (3 min/bed position). Reconstructed FOV had an impact on BSREM outcome measures; SNR increased and SBR decreased when FOV was shifted from 70 to 50 cm. The evaluation of visual image quality resulted in similar scores for reconstructions, although a β-factor of 400 obtained the highest mean whereas a β-factor of 267 was ranked best in overall image quality, contrast, sharpness, and tumor detectability. Conclusion: In comparison with TOF OSEM, penalized BSREM reconstruction resulted in an increased tumor SUVmax and an improved SNR and SBR at a matched level of noise. BSREM allowed for a shorter acquisition than TOF OSEM, with equal image quality.

Road Extraction From a High Spatial Resolution Remote Sensing Image Based on Richer Convolutional Features

The extraction and vectorization of roads from high spatial resolution remote sensing (HSRRS) images are of great significance to city planning and development. However, significant as they are, it is usually an arduous task to put them into practice because the HSRRS images are often filled with complex ground information. Furthermore, extracted roads may suffer from netsplit or brokenness. This paper thus proposes Richer convolutional features (RCFs)-based road extraction (Road-RCF) as a method which targets these issues. A modified roads sample set and RCF network are applied to generate road probabilities in order to extract initial road information. After the road centerlines extraction by the refinement algorithm, vectorized roads are ultimately extracted. The compared experiment results show that the Road-RCF method produce better road extraction results than the other four state-of-the-art methods, in both quantitative road extraction accuracy metrics and the qualitative visual evaluation. The benefits of this model are threefold. First, the image-to-image network structure of side-output realizes multi-scale and multi-level road feature fusion in order to make a full use of the information from a low level to a high level. Second, according to the deep supervision of the side-output, it guides the learning of the correct road information. Third, after the detection of the road, the road centerlines are vectorized to facilitate the attribute information management and electronic map production. In a word, the proposed Road-RCF method is both practical and meaningful toward updating the geo-information system database.

DE BRUIJN STRUCTURED ILLUMINATION STUDYING WITHIN THE TASK OF RESTORING HANDS RELIEF

In the course of studies on the problem of restoring hands relief, using the de Bruijn structured illumination, methods of solving this problem are proposed. This is a method of simple quantitative detection of Hough segments on the skin of the hand, a method of qualitative visual evaluation of the effectiveness of the color palette using the dominant color, and a method of the weight coefficients of the components of the color palette. The proposed methods make it possible to quantitatively determine the optimal choice of the color scheme for generating the de Bruijn bands when illumination of the hand, to restore its relief. The work describes the stages of this study, led from visual observation to a full quantitative calculation of the quality of calibration illuminations, with the possibility of their optimal choice. In the course of experiments and observations, the requirements for the technical support of research were developed to achieve the best quality of the images of the hands. Also, the paper presents a high-speed de Bruijn sequence generating algorithm using Lyndon's words, which excludes the search for Euler chains or Hamiltonian cycles, for various kinds of de Bruijn graphs. With its help, the generation of structured light patterns with various color schemes was carried out, with the purpose of further analysis of their use in 3D reconstruction systems of hands.

A Spectral-Texture Kernel-Based Classification Method for Hyperspectral Images

Classification of hyperspectral images always suffers from high dimensionality and very limited labeled samples. Recently, the spectral-spatial classification has attracted considerable attention and can achieve higher classification accuracy and smoother classification maps. In this paper, a novel spectral-spatial classification method for hyperspectral images by using kernel methods is investigated. For a given hyperspectral image, the principle component analysis (PCA) transform is first performed. Then, the first principle component of the input image is segmented into non-overlapping homogeneous regions by using the entropy rate superpixel (ERS) algorithm. Next, the local spectral histogram model is applied to each homogeneous region to obtain the corresponding texture features. Because this step is performed within each homogenous region, instead of within a fixed-size image window, the obtained local texture features in the image are more accurate, which can effectively benefit the improvement of classification accuracy. In the following step, a contextual spectral-texture kernel is constructed by combining spectral information in the image and the extracted texture information using the linearity property of the kernel methods. Finally, the classification map is achieved by the support vector machines (SVM) classifier using the proposed spectral-texture kernel. Experiments on two benchmark airborne hyperspectral datasets demonstrate that our method can effectively improve classification accuracies, even though only a very limited training sample is available. Specifically, our method can achieve from 8.26% to 15.1% higher in terms of overall accuracy than the traditional SVM classifier. The performance of our method was further compared to several state-of-the-art classification methods of hyperspectral images using objective quantitative measures and a visual qualitative evaluation.

Spectral-Spatial Classification of Hyperspectral Images Using Joint Bilateral Filter and Graph Cut Based Model

Hyperspectral image classification can be achieved by modeling an energy minimization problem on a graph of image pixels. In this paper, an effective spectral-spatial classification method for hyperspectral images based on joint bilateral filtering (JBF) and graph cut segmentation is proposed. In this method, a novel technique for labeling regions obtained by the spectral-spatial segmentation process is presented. Our method includes the following steps. First, the probabilistic support vector machines (SVM) classifier is used to estimate probabilities belonging to each information class. Second, an extended JBF is employed to perform image smoothing on the probability maps. By using our JBF process, salt-and-pepper classification noise in homogeneous regions can be effectively smoothed out while object boundaries in the original image are better preserved as well. Third, a sequence of modified bi-labeling graph cut models is constructed for each information class to extract the desirable object belonging to the corresponding class from the smoothed probability maps. Finally, a classification map is achieved by merging the segmentation maps obtained in the last step using a simple and effective rule. Experimental results based on three benchmark airborne hyperspectral datasets with different resolutions and contexts demonstrate that our method can achieve 8.56%–13.68% higher overall accuracies than the pixel-wise SVM classifier. The performance of our method was further compared to several classical hyperspectral image classification methods using objective quantitative measures and a visual qualitative evaluation.

A Combination of Geographically Weighted Regression, Particle Swarm Optimization and Support Vector Machine for Landslide Susceptibility Mapping: A Case Study at Wanzhou in the Three Gorges Area, China.

In this study, a novel coupling model for landslide susceptibility mapping is presented. In practice, environmental factors may have different impacts at a local scale in study areas. To provide better predictions, a geographically weighted regression (GWR) technique is firstly used in our method to segment study areas into a series of prediction regions with appropriate sizes. Meanwhile, a support vector machine (SVM) classifier is exploited in each prediction region for landslide susceptibility mapping. To further improve the prediction performance, the particle swarm optimization (PSO) algorithm is used in the prediction regions to obtain optimal parameters for the SVM classifier. To evaluate the prediction performance of our model, several SVM-based prediction models are utilized for comparison on a study area of the Wanzhou district in the Three Gorges Reservoir. Experimental results, based on three objective quantitative measures and visual qualitative evaluation, indicate that our model can achieve better prediction accuracies and is more effective for landslide susceptibility mapping. For instance, our model can achieve an overall prediction accuracy of 91.10%, which is 7.8%–19.1% higher than the traditional SVM-based models. In addition, the obtained landslide susceptibility map by our model can demonstrate an intensive correlation between the classified very high-susceptibility zone and the previously investigated landslides.