Unsupervised Fuzzy C-means Clustering Research Articles

Multiple sclerosis versus cerebral small vessel disease in MRI: a practical approach using qualitative and quantitative signal intensity differences in white matter lesions.

Multiple sclerosis (MS) and cerebral small vessel disease (CSVD) are relatively common radiological entities that occasionally necessitate differential diagnosis. To investigate the differences in magnetic resonance imaging (MRI) signal intensity (SI) between MS and CSVD related white matter lesions. On 1.5-T and 3-T MRI scanners, 50 patients with MS (380 lesions) and 50 patients with CSVD (395 lesions) were retrospectively evaluated. Visual inspection was used to conduct qualitative analysis on diffusion-weighted imaging (DWI)_b1000 to determine relative signal intensity. The thalamus served as the reference for quantitative analysis based on SI ratio (SIR). The statistical analysis utilized univariable and multivariable methods. There were analyses of patient and lesion datasets. On a dataset restricted by age (30-50 years), additional evaluations, including unsupervised fuzzy c-means clustering, were performed. Using both quantitative and qualitative features, the optimal model achieved a 100% accuracy, sensitivity, and specificity with an area under the curve (AUC) of 1 in patient-wise analysis. With an AUC of 0.984, the best model achieved a 94% accuracy, sensitivity, and specificity when using only quantitative features. The model's accuracy, sensitivity, and specificity were 91.9%, 84.6%, and 95.8%, respectively, when using the age-restricted dataset. Independent predictors were T2_SIR_max (optimal cutoff=2.1) and DWI_b1000_SIR_mean (optimal cutoff=1.1). Clustering also performed well with an accuracy, sensitivity, and specificity of 86.5%, 70.6%, and 100%, respectively, in the age-restricted dataset. SI characteristics derived from DWI_b1000 and T2-weighted-based MRI demonstrate excellent performance in differentiating white matter lesions caused by MS and CSVD.

An innovative integrated modelling of safety data using multiple correspondence analysis and fuzzy discretization techniques

In this study, we have proposed an innovative integrated methodology to handle a mix of categorical and numeric safety data. We have augmented the traditional multiple correspondence analysis (MCA) through the use of fuzzy discretization approach, t-SNE technique and fuzzy c-means clustering. The fuzzy discretization approach transforms the continuous variables to categorical variables to make them analyzable using MCA. R2-profile is adopted to obtain the best number of hidden dimensions representing the maximum categorical information. Then, t-SNE technique is used to represent the high dimensional categorical information in a 2D map to visualize the significant categorical associations. Then, fuzzy c-means clustering (FCM) is used to group the categories in different clusters based on their membership degree. To determine the optimal number of clusters, cluster validity indices are used. Davies-Bouldin (DB) Index, Dunn’s (DU) Index and Silhouette (SW) coefficients are used to determine the quality of clustering solutions. The proposed methodology is tested using electric overhead traveling (EOT) crane related near-miss incidents and found that our approach is effective. From managerial implication point of view, several safety rules are generated and subsequent safety countermeasures are proposed. Further, the results obtained through FCM is compared with K-means (KM) algorithm and unsupervised fuzzy c-means clustering (UPFCM). FCM outperforms KM and UPFCM on the basis of quality of solutions.

A Quantum-Inspired Self-Supervised Network model for automatic segmentation of brain MR images

The classical self-supervised neural network architectures suffer from slow convergence problem and incorporation of quantum computing in classical self-supervised networks is a potential solution towards it. In this article, a fully self-supervised novel quantum-inspired neural network model referred to as Quantum-Inspired Self-Supervised Network (QIS-Net) is proposed and tailored for fully automatic segmentation of brain MR images to obviate the challenges faced by deeply supervised Convolutional Neural Network (CNN) architectures. The proposed QIS-Net architecture is composed of three layers of quantum neuron (input, intermediate and output) expressed as qbits. The intermediate and output layers of the QIS-Net architecture are inter-linked through bi-directional propagation of quantum states, wherein the image pixel intensities (quantum bits) are self-organized in between these two layers without any external supervision or training. Quantum observation allows to obtain the true output once the superimposed quantum states interact with the external environment. The proposed self-supervised quantum-inspired network model has been tailored for and tested on Dynamic Susceptibility Contrast (DSC) brain MR images from Nature data sets for detecting complete tumor and reported promising accuracy and reasonable dice similarity scores in comparison with the unsupervised Fuzzy C-Means clustering, self-trained QIBDS Net, Opti-QIBDS Net, deeply supervised U-Net and Fully Convolutional Neural Networks (FCNNs).

Tissue-specific and interpretable sub-segmentation of whole tumour burden on CT images by unsupervised fuzzy clustering

Background:Cancer typically exhibits genotypic and phenotypic heterogeneity, which can have prognostic significance and influence therapy response. Computed Tomography (CT)-based radiomic approaches calculate quantitative features of tumour heterogeneity at a mesoscopic level, regardless of macroscopic areas of hypo-dense (i.e., cystic/necrotic), hyper-dense (i.e., calcified), or intermediately dense (i.e., soft tissue) portions. Method:With the goal of achieving the automated sub-segmentation of these three tissue types, we present here a two-stage computational framework based on unsupervised Fuzzy C-Means Clustering (FCM) techniques. No existing approach has specifically addressed this task so far. Our tissue-specific image sub-segmentation was tested on ovarian cancer (pelvic/ovarian and omental disease) and renal cell carcinoma CT datasets using both overlap-based and distance-based metrics for evaluation. Results:On all tested sub-segmentation tasks, our two-stage segmentation approach outperformed conventional segmentation techniques: fixed multi-thresholding, the Otsu method, and automatic cluster number selection heuristics for the K-means clustering algorithm. In addition, experiments showed that the integration of the spatial information into the FCM algorithm generally achieves more accurate segmentation results, whilst the kernelised FCM versions are not beneficial. The best spatial FCM configuration achieved average Dice similarity coefficient values starting from 81.94±4.76 and 83.43±3.81 for hyper-dense and hypo-dense components, respectively, for the investigated sub-segmentation tasks. Conclusions:The proposed intelligent framework could be readily integrated into clinical research environments and provides robust tools for future radiomic biomarker validation.

Deep Learning Initialized and Gradient Enhanced Level-Set Based Segmentation for Liver Tumor From CT Images

Liver and liver tumor segmentation provides vital biomarkers for surgical planning and hepatic diagnosis. In this paper, we propose and validate a novel level-set method integrating an enhanced edge indicator and an automatically derived initial curve for CT based liver tumor segmentation. At the preprocessing step, the CT image intensity values were truncated to lie in a fixed range to enhance the image contrast surrounding liver and liver tumor. To remove non-liver tissues for subsequent tumor segmentation, liver was firstly segmented using two convolutional neural networks in a coarse-to-fine manner. A 2D slice-based U-net was used to roughly localize the liver and a 3D patch-based fully convolutional network was used to refine the liver segmentation as well as to roughly localize the liver tumor. A novel level-set method was then presented to further refine the tumor segmentation. Specifically, the probabilistic distribution of the liver tumor was estimated using unsupervised fuzzy c-means clustering, which was then utilized to enhance the edge-detector used in level-set. Effectiveness of the proposed pipeline was validated on two publicly-available datasets. Experimental results identified the superior segmentation performance of the proposed pipeline over state-of-the-art methods.

Automated Prostate Gland Segmentation Based on an Unsupervised Fuzzy C-Means Clustering Technique Using Multispectral T1w and T2w MR Imaging

Prostate imaging analysis is difficult in diagnosis, therapy, and staging of prostate cancer. In clinical practice, Magnetic Resonance Imaging (MRI) is increasingly used thanks to its morphologic and functional capabilities. However, manual detection and delineation of prostate gland on multispectral MRI data is currently a time-expensive and operator-dependent procedure. Efficient computer-assisted segmentation approaches are not yet able to address these issues, but rather have the potential to do so. In this paper, a novel automatic prostate MR image segmentation method based on the Fuzzy C-Means (FCM) clustering algorithm, which enables multispectral T1-weighted (T1w) and T2-weighted (T2w) MRI anatomical data processing, is proposed. This approach, using an unsupervised Machine Learning technique, helps to segment the prostate gland effectively. A total of 21 patients with suspicion of prostate cancer were enrolled in this study. Volume-based metrics, spatial overlap-based metrics and spatial distance-based metrics were used to quantitatively evaluate the accuracy of the obtained segmentation results with respect to the gold-standard boundaries delineated manually by an expert radiologist. The proposed multispectral segmentation method was compared with the same processing pipeline applied on either T2w or T1w MR images alone. The multispectral approach considerably outperforms the monoparametric ones, achieving an average Dice Similarity Coefficient 90.77 ± 1.75, with respect to 81.90 ± 6.49 and 82.55 ± 4.93 by processing T2w and T1w imaging alone, respectively. Combining T2w and T1w MR image structural information significantly enhances prostate gland segmentation by exploiting the uniform gray appearance of the prostate on T1w MRI.

Gamma Knife treatment planning: MR brain tumor segmentation and volume measurement based on unsupervised Fuzzy C-Means clustering

Nowadays, radiation treatment is beginning to intensively use MRI thanks to its greater ability to discriminate healthy and diseased soft-tissues. Leksell Gamma Knife® is a radio-surgical device, used to treat different brain lesions, which are often inaccessible for conventional surgery, such as benign or malignant tumors. Currently, the target to be treated with radiation therapy is contoured with slice-by-slice manual segmentation on MR datasets. This approach makes the segmentation procedure time consuming and operator-dependent. The repeatability of the tumor boundary delineation may be ensured only by using automatic or semiautomatic methods, supporting clinicians in the treatment planning phase. This article proposes a semiautomatic segmentation method, based on the unsupervised Fuzzy C-Means clustering algorithm. Our approach helps segment the target and automatically calculates the lesion volume. To evaluate the performance of the proposed approach, segmentation tests on 15 MR datasets were performed, using both area-based and distance-based metrics, obtaining the following average values: Similarity Index = 95.59%, Jaccard Index = 91.86%, Sensitivity = 97.39%, Specificity = 94.30%, Mean Absolute Distance = 0.246[pixels], Maximum Distance = 1.050[pixels], and Hausdorff Distance = 1.365[pixels]. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 213–225, 2015

A fully automatic 2D segmentation method for uterine fibroid in MRgFUS treatment evaluation

PurposeMagnetic Resonance guided Focused UltraSound (MRgFUS) represents a non-invasive surgical approach that uses thermal ablation to treat uterine fibroids. After the MRgFUS treatment, an operator must manually segment the treated fibroid areas to evaluate the NonPerfused Volume (NPV). This manual approach is operator-dependent, introducing issues of result reproducibility, which could lead to errors in the subsequent follow-up phase. Moreover, manual segmentation is time-consuming, and can have a negative impact on the optimization of both machine-time and operator-time. MethodTo address these issues, in this paper a novel fully automatic method based on the unsupervised Fuzzy C-Means clustering and iterative optimal threshold selection algorithms for uterus and fibroid segmentation is proposed. The developed method could be used to enhance the current manual methodology performed by healthcare operators for post-operative NPV evaluation in uterine fibroid MRgFUS treatments. ResultsThe proposed method was tested on 15 MR datasets of 15 different patients with uterine fibroids and evaluated using area-based and distance-based metrics. A comparison of extracted volume was also performed. Average values for fibroid (ROT) segmentation are SDI=88.67%, JI=80.70%, SE=89.79%, SP=88.73%, MAD=2.200 [pixels], MAXD=6.233 [pixels] and HD=2.988 [pixels]. Moreover, to make a quantitative evaluation of this method, our experimental results were compared with similar literature approaches. ConclusionsThe proposed method provides a practical approach for the automatic evaluation of the boundary and volume of ablated fibroid regions, without any external user input. The achieved segmentation results show the validity and the effectiveness of the proposed solution.

Detection of Fusarium in single wheat kernels using spectral imaging

Fusarium head blight (FHB) is a harmful fungal disease that occurs in small grains. Non-destructive detection of this disease is traditionally done using spectroscopy or image processing. In this paper the combination of these two in the form of spectral imaging is evaluated. Transmission spectral images are recorded, both in the visible and near-infrared range from FHB infected wheat kernels. These images are analyzed, using light absorption, the relation between two wavelength bands, unsupervised fuzzy c-means clustering and supervised partial least squares regression. The reference method for training and validation is TaqMan real-time PCR. Results show that nearinfrared spectral images perform much better than spectral images in the visible range. Kernels with more than 6000 pg Fusarium DNA could clearly be identified. Above 100 pg it was possible to predict the amount of Fusarium with a Q2 of 0.8. This was both for Partial Least Squares regression (PLS) and a simple wavelength ratio. Also fuzzy c-means clustering shows a relation between amount of Fusarium and spectra. Fusarium head blight (FHB) is a harmful fungal disease that occurs in small grains. Non-destructive detection of this disease is traditionally done using spectroscopy or image processing. In this paper the combination of these two in the form of spectral imaging is evaluated. Transmission spectral images are recorded, both in the visible and near-infrared range from FHB infected wheat kernels. These images are analyzed, using light absorption, the relation between two wavelength bands, unsupervised fuzzy c-means clustering and supervised partial least squares regression. The reference method for training and validation is TaqMan real-time PCR. Results show that near-infrared spectral images perform much better than spectral images in the visible range. Kernels with more than 6000 pg Fusarium DNA could clearly be identified. Above 100 pg it was possible to predict the amount of Fusarium with a Q(2) of 0.8. This was both for Partial Least Squares regression (PLS) and a simple wavelength ratio. Also fuzzy c-means clustering shows a relation between amount of Fusarium and spectra.

Delineating productivity zones on claypan soil fields using apparent soil electrical conductivity

Efficient and cost-effective methods are needed for delineating sub-field productivity zones to improve soil and crop site-specific management. This investigation was conducted to answer the question of whether apparent soil electrical conductivity (EC a) and elevation could be used to delineate productivity zones (SPZ) for claypan soil fields that would agree with productivity zones delineated from yield map data (YPZ). Ten and seven years of combine-monitored yield maps were available for two Missouri claypan soil fields, designated Field 1 and Field 2, respectively. The fields were generally cropped in corn and soybean. Soil EC a data were collected with a non-contact, electromagnetic induction-based EC a sensor (Geonics EM38) and a coulter-based sensor (Veris model 3100). Elevation data were collected using a real-time kinematic GPS. Unsupervised fuzzy c-means clustering was independently used both on yield data to delineate three YPZ and on combinations of EC a and/or elevation data to delineate three SPZ. Outcomes of YPZ and SPZ were matched and agreement calculated with an overall accuracy statistic and a statistical index called the Kappa coefficient. Best performing combinations of EC a and elevation variables gave 60–70% agreement between YPZ and SPZ. We consider this level of agreement promising, especially considering that there were many other yield-limiting factors unrelated to EC a and elevation. Generally, multiple variables of EC a and elevation were better than a single variable for generating SPZ. The specific combinations of EC a and/or elevation variables that gave highest agreement between YPZ and SPZ were field specific. Based on these findings, we conclude EC a and elevation measurements can be reliably used for creating productivity zones on claypan soil fields.