SABOS-Net: Self-supervised attention based network for automatic organ segmentation of head and neck CT images

The segmentation of Organs At Risk (OAR) in Computed Tomography (CT) images is an essential part of the planning phase of radiation treatment to avoid the adverse effects of cancer radiotherapy treatment. Accurate segmentation is a tedious task in the head and neck region due to a large number of small and sensitive organs and the low contrast of CT images. Deep learning-based automatic contouring algorithms can ease this task even when the organs have irregular shapes and size variations. This paper proposes a fully automatic deep learning-based self-supervised 3D Residual UNet architecture with CBAM(Convolution Block Attention Mechanism) for the organ segmentation in head and neck CT images. The Model Genesis structure and image context restoration techniques are used for self-supervision, which can help the network learn image features from unlabeled data, hence solving the annotated medical data scarcity problem in deep networks. A new loss function is applied for training by integrating Focal loss, Tversky loss, and Cross-entropy loss. The proposed model outperforms the state-of-the-art methods in terms of dice similarity coefficient in segmenting the organs. Our self-supervised model could achieve a 4% increase in the dice score of Chiasm, which is a small organ that is present only in a very few CT slices. The proposed model exhibited better accuracy for 5 out of 7 OARs than the recent state-of-the-art models. The proposed model could simultaneously segment all seven organs in an average time of 0.02 s. The source code of this work is made available at https://github.com/seeniafrancis/SABOSNet .     

Residual U-Network for Breast Tumor Segmentation from Magnetic Resonance Images

Breast cancer positions as the most well-known threat and the main source of malignant growth-related morbidity and mortality throughout the world. It is apical of all new cancer incidences analyzed among females. Two features substantially influence the classification accuracy of malignancy and benignity in automated cancer diagnostics. These are the precision of tumor segmentation and appropriateness of extracted attributes required for the diagnosis. In this research, the authors have proposed a ResU-Net (Residual U-Network) model for breast tumor segmentation. The proposed methodology renders augmented, and precise identification of tumor regions and produces accurate breast tumor segmentation in contrast-enhanced MR images. Furthermore, the proposed framework also encompasses the residual network technique, which subsequently enhances the performance and displays the improved training process. Over and above, the performance of ResU-Net has experimentally been analyzed with conventional U-Net, FCN8, FCN32. Algorithm performance is evaluated in the form of dice coefficient and MIoU (Mean Intersection of Union), accuracy, loss, sensitivity, specificity, F1score. Experimental results show that ResU-Net achieved validation accuracy & dice coefficient value of 73.22% & 85.32% respectively on the Rider Breast MRI dataset and outperformed as compared to the other algorithms used in experimentation.     

Pediatric brain extraction from T2-weighted MR images using 3D dual frame U-net and human connectome database

Accurate extraction of brain tissues from magnetic resonance (MR) images is important in neuroradiology. However, brain extraction is more difficult for pediatric brains than for adult brains due to several factors including smaller brain sizes and lower tissue contrasts. In this work, we propose a brain extraction technique that utilizes dual frame (DF) 3D U-net deep learning architecture and the human connectome project (HCP) database for multislice 2D pediatric T2-weighted MR images with diseases. To improve segmentation accuracy in small pediatric brains with detailed boundary regions, DF 3D U-net architecture was used. We pretrained networks with the HCP database to compensate for the limited amount of MR images and manual segmentation masks of pediatric patients. For quantitative analysis, we compared the brain extraction results of brain extraction tool, DF, and conventional 3D U-net using the dice similarity coefficient (DSC), intersection of union (IoU), and boundary F1 (BF) scores; each deep learning architecture was evaluated with and without pretraining using the HCP. This study included 10 patients with diseases and all images were acquired using a PROPELLER MR sequence. Pretraining using the HCP database enhanced segmentation performance of the network, and the skip connections in the DF 3D U-net could enhance the contour similarity of segmentation results. Experimental results showed that the proposed method increased the DSC, IoU, and BF scores by 0.8%, 1.6%, and 1.5%, respectively, compared with those of the conventional 3D U-net without pretraining.     

Classification of mammogram images using the energy probability in frequency domain and most discriminative power coefficients

The purpose of this work is to develop a computer-aided diagnosis (CAD) system to assist radiologists in the classification of mammogram images. The CAD system is composed of three main steps. The first step is image preprocessing and segmentation with the seeded region growing algorithm applied on a localized triangular region to remove only the muscle. In the second step of the CAD system, we proposed a novel features extraction method, which consists of three stages. In the first, the discrete cosine transform (DCT) is applied on all obtained regions of interest and then only the upper left corner (ULC) of DCT coefficients is retained. Second, we have applied the energy probability to the ULCs that is used as a criterion for selecting discriminant information. At the last stage, a new Most Discriminative power coefficient algorithm has been proposed to select the most significant features. In the final step of the CAD, the support vector machines, Naive Bayes, and artificial neural network (ANN) classifiers are used to make an effective classification. The evaluation of the proposed algorithm on the mini-Mammographic Image Analysis Society database shows its efficiency over other recently proposed CAD systems in the literature, whereas an accuracy of 100% can be achieved using ANN with a small number of features.     

Brain Tumor Detection and Classification Using PSO and Convolutional Neural Network

Tumor detection has been an active research topic in recent years due to the high mortality rate. Computer vision (CV) and image processing techniques have recently become popular for detecting tumors in MRI images. The automated detection process is simpler and takes less time than manual processing. In addition, the difference in the expanding shape of brain tumor tissues complicates and complicates tumor detection for clinicians. We proposed a new framework for tumor detection as well as tumor classification into relevant categories in this paper. For tumor segmentation, the proposed framework employs the Particle Swarm Optimization (PSO) algorithm, and for classification, the convolutional neural network (CNN) algorithm. Popular preprocessing techniques such as noise removal, image sharpening, and skull stripping are used at the start of the segmentation process. Then, PSO-based segmentation is applied. In the classification step, two pre-trained CNN models, alexnet and inception-V3, are used and trained using transfer learning. Using a serial approach, features are extracted from both trained models and fused features for final classification. For classification, a variety of machine learning classifiers are used. Average dice values on datasets BRATS-2018 and BRATS-2017 are 98.11 percent and 98.25 percent, respectively, whereas average jaccard values are 96.30 percent and 96.57% (Segmentation Results). The results were extended on the same datasets for classification and achieved 99.0% accuracy, sensitivity of 0.99, specificity of 0.99, and precision of 0.99. Finally, the proposed method is compared to state-of-the-art existing methods and outperforms them.     

A rapid automatic brain tumor detection method for MRI images using modified minimum error thresholding technique

This proposed work is aimed to develop a rapid automatic method to detect the brain tumor from T2‐weighted MRI brain images using the principle of modified minimum error thresholding (MET) method. Initially, modified MET method is applied to produce well segmented and sub‐structural clarity for MRI brain images. Further, using FCM clustering the appearance of tumor area is refined. The obtained results are compared with corresponding ground truth images. The quantitative measures of results were compared with the results of those conventional methods using the metrics predictive accuracy (PA), dice coefficient (DC), and processing time. The PA and DC values of the proposed method attained maximum value and processing time is minimum while compared to conventional FCM and k‐means clustering techniques. This proposed method is more efficient and faster than the existing segmentation methods in detecting the tumor region from T2‐weighted MRI brain images. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 77–85, 2015     

Brain Tumor Auto-Segmentation on Multimodal Imaging Modalities Using Deep Neural Network

Due to the difficulties of brain tumor segmentation, this paper proposes a strategy for extracting brain tumors from three-dimensional Magnetic Resonance Image (MRI) and Computed Tomography (CT) scans utilizing 3D U-Net Design and ResNet50, taken after by conventional classification strategies. In this inquire, the ResNet50 picked up accuracy with 98.96%, and the 3D U-Net scored 97.99% among the different methods of deep learning. It is to be mentioned that traditional Convolutional Neural Network (CNN) gives 97.90% accuracy on top of the 3D MRI. In expansion, the image fusion approach combines the multimodal images and makes a fused image to extricate more highlights from the medical images. Other than that, we have identified the loss function by utilizing several dice measurements approach and received Dice Result on top of a specific test case. The average mean score of dice coefficient and soft dice loss for three test cases was 0.0980. At the same time, for two test cases, the sensitivity and specification were recorded to be 0.0211 and 0.5867 using patch level predictions. On the other hand, a software integration pipeline was integrated to deploy the concentrated model into the webserver for accessing it from the software system using the Representational state transfer (REST) API. Eventually, the suggested models were validated through the Area Under the Curve–Receiver Characteristic Operator (AUC–ROC) curve and Confusion Matrix and compared with the existing research articles to understand the underlying problem. Through Comparative Analysis, we have extracted meaningful insights regarding brain tumour segmentation and figured out potential gaps. Nevertheless, the proposed model can be adjustable in daily life and the healthcare domain to identify the infected regions and cancer of the brain through various imaging modalities.     

Detection and classification of microcalcification from digital mammograms with firefly algorithm,extreme learning machine and non-linear regression models: A comparison

In this study, abnormalities in medical images are analysed using three classifiers, and the results are compared. Breast cancer remains a major public health problem among women worldwide. Recently, many algorithms have evolved for the investigation of breast cancer diagnosis through medical imaging. A computer-aided microcalcification detection method is proposed to categorise the nature of breast cancer as either benign or malignant from input mammogram images. The standard mammogram image corpus, the Mammogram Image Analysis Society database is utilised, and feature extraction is performed using five different wavelet families at level 4 and level 6 decomposition. The work is accomplished through firefly algorithm (FA), extreme learning machine (ELM) and least-square-based non-linear regression (LSNLR) classifiers. The performance of the classifiers is compared by benchmark metrics, such as total error rate, specificity, sensitivity, area under the receiver operating characteristic curve, precision, F1 score and the Matthews correlation coefficient. As validation of the classifier results, a kappa analysis is included to determine the agreement among classifiers. The LSNLR classifier attains a 3% to 7% improvement in average accuracy compared with the average classification accuracy of the FA (86.75%) and ELM (90.836%) classifiers.     

Multi-Scale 3D U-Nets: An approach to automatic segmentation of brain tumor

Gliomas segmentation is a critical and challenging task in surgery and treatment, and it is also the basis for subsequent evaluation of gliomas. Magnetic resonance imaging is extensively employed in diagnosing brain and nervous system abnormalities. However, brain tumor segmentation remains a challenging task, because differentiating brain tumors from normal tissues is difficult, tumor boundaries are often ambiguous and there is a high degree of variability in the shape, location, and extent of the patient. It is therefore desired to devise effective image segmentation architectures. In the past few decades, many algorithms for automatic segmentation of brain tumors have been proposed. Methods based on deep learning have achieved favorable performance for brain tumor segmentation. In this article, we propose a Multi-Scale 3D U-Nets architecture, which uses several U-net blocks to capture long-distance spatial information at different resolutions. We upsample feature maps at different resolutions to extract and utilize sufficient features, and we hypothesize that semantically similar features are easier to learn and process. In order to reduce the computational cost, we use 3D depthwise separable convolution instead of some standard 3D convolution. On BraTS 2015 testing set, we obtained dice scores of 0.85, 0.72, and 0.61 for the whole tumor, tumor core, and enhancing tumor, respectively. Our segmentation performance was competitive compared to other state-of-the-art methods.     

Rapid brain tissue segmentation process by modified FCM algorithm with CUDA enabled GPU machine

The proposed work introduces a modified method of fuzzy c means (FCM) algorithm using bias field correction and partial supervision techniques. The proposed method is named as bias corrected partial supervision FCM (BCPSFCM). The modified membership function takes the advantage of available knowledge from labeled patterns with the bias field correction. The experiment is tested on internet brain segmentation repository with their gold standard. The performance of the method is compared with three existing methods and 12 state of the art methods using dice coefficient, sensitivity, specificity, and accuracy. Accuracy of the proposed method reached upto 98%, 98%, and 99% of GM, WM, and CSF segmentation but required additional computation power from graphics processing unit (GPU). Further parallel BCPSFCM is proposed with the help of compute unified device architecture enabled GPU machine and the processing time is reduced up to 49 times than the serial implementation.     

A comprehensive review: Segmentation of MRI images—brain tumor

Segmentation of tumors in human brain aims to classify different abnormal tissues (necrotic core, edema, active cells) from normal tissues (cerebrospinal fluid, gray matter, white matter) of the brain. In existence, detection of abnormal tissues is easy for studying brain tumor, but reproducibility, characterization of abnormalities and accuracy are complicated in the process of segmentation. The magnetic resonance imaging (MRI)‐based segmentation of tumors in brain images is more enhancing and attracting in current years of research studies. It is due to non‐invasive examination and good contrast prone to soft tissues of images obtained from MRI modality. Medical approval of different segmentation techniques depends on the benchmark and simplicity of the method. This article incorporates both fully‐automatic and semi‐automatic methods for segmentation. The outlook study of this article is to provide the summary of most significant segmentation methods of tumors in brain using MRI. © 2016 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 26, 295–304, 2016     

Breast cancer diagnosis from mammographic images using optimized feature selection and neural network architecture

Breast cancer is one of the deadly diseases in women that have raised the mortality rate of women. An accurate and early detection of breast cancer using mammogram images is still a complex task. Hence, this article proposes a novel breast cancer detection model, which included five major phases: (a) preprocessing, (b) segmentation, (c) feature extraction, (d) feature selection, and (e) classification. The input mammogram image is initially preprocessed using contrast limited adaptive histogram equalization (CLAHE) and median filtering. The preprocessed image is then subjected to segmentation via the region growing algorithm. Subsequently, geometric features, texture features and gradient features are extracted from the segmented image. Since the length of the feature vector is large, it is essential to select the optimal features. Here, the selection of optimal features is done by a hybrid optimization algorithm. Once the optimal features are selected, they are subjected to the classification process involving the neural network (NN) classifier. As a novelty, the weight of NN is selected optimally to enhance the accuracy of diagnosis (benign and malignant). The optimal feature selection as well as the weight optimization of NN is accomplished by merging the Lion algorithm (LA) and particle swarm optimization (PSO), named as velocity updated lion algorithm (VU‐LA). Finally, a performance‐based evaluation is carried out between VU‐LA and the existing models like, whale optimization algorithm (WOA), gray wolf optimization (GWO), firefly (FF), PSO, and LA.     

An assessment of digital stereotaxis in the National Health Service Breast Screening Programme

A woman may be invited back for further assessment if an abnormality is found on her mammogram. A stereotactic attachment is used to determine where to place the biopsy device. Digital equipment has the advantage of providing images almost instantaneously and utilising software to determine the location of the sampling device. These devices should make the procedure quicker and hence less traumatic for the women. However, digital devices have poorer spatial and contrast resolution, which could adversely affect the accuracy of sampling device placement. Although the dose received during a normal screening mammogram is well known, the dose for the stereo procedure is unknown, partly because only a small part of the breast is directly irradiated. However, the lady may undergo multiple exposures. For a prospective survey of doses and technique arm of the study a proforma was completed for 134 women. The dose women received during a stereotactic procedure was estimated.     

Automatic segmentation system for liver tumors based on the multilevel thresholding and electromagnetism optimization algorithm

In this article, we propose an automated segmentation system for liver tumors using magnetic resonance imaging and computed tomography. The proposed system is based on the algorithm of multilevel thresholding with electromagnetism optimization (EMO). The system starts with visualizing a patient's digital communication in medicine (DICOM) abdominal data set in three views. Two-stage active contour segmentation methods that integrate region-based local and global techniques using the active geodesic contour technique are proposed to segment the liver. To increase the accuracy and speed of segmentation for liver images, we identify the optimal threshold of the image segmentation method based on EMO with Otsu and Kapur algorithms. EMO offers interesting search capabilities while keeping a low computational cost. The proposed system was tested using a set of five DICOM data sets. All images were of the same size and stored in JPEG format (512 × 512 pixels). Experimental results illustrate that the proposed system outperforms state-of-the-art methods such as the watershed algorithm. The average sensitivity, specificity, and accuracy of the segmented liver using the active contour model were 97.05%, 99.88%, and 98.47%, respectively. Moreover, the average sensitivity, specificity, and accuracy of the segmented liver tumor results were 94.15%, 99.57%, and 96.86%, respectively.     

Proposed Framework for Detection of Breast Tumors

Computer vision is one of the significant trends in computer science. It plays as a vital role in many applications, especially in the medical field. Early detection and segmentation of different tumors is a big challenge in the medical world. The proposed framework uses ultrasound images from Kaggle, applying five diverse models to denoise the images, using the best possible noise-free image as input to the U-Net model for segmentation of the tumor, and then using the Convolution Neural Network (CNN) model to classify whether the tumor is benign, malignant, or normal. The main challenge faced by the framework in the segmentation is the speckle noise. It’s is a multiplicative and negative issue in breast ultrasound imaging, because of this noise, the image resolution and contrast become reduced, which affects the diagnostic value of this imaging modality. As result, speckle noise reduction is very vital for the segmentation process. The framework uses five models such as Generative Adversarial Denoising Network (DGAN-Net), Denoising U-Shaped Net (D-U-NET), Batch Renormalization U-Net (Br-U-NET), Generative Adversarial Network (GAN), and Nonlocal Neutrosophic of Wiener Filtering (NLNWF) for reducing the speckle noise from the breast ultrasound images then choose the best image according to peak signal to noise ratio (PSNR) for each level of speckle-noise. The five used methods have been compared with classical filters such as Bilateral, Frost, Kuan, and Lee and they proved their efficiency according to PSNR in different levels of noise. The five diverse models are achieved PSNR results for speckle noise at level (0.1, 0.25, 0.5, 0.75), (33.354, 29.415, 27.218, 24.115), (31.424, 28.353, 27.246, 24.244), (32.243, 28.42, 27.744, 24.893), (31.234, 28.212, 26.983, 23.234) and (33.013, 29.491, 28.556, 25.011) for DGAN, Br-U-NET, D-U-NET, GAN and NLNWF respectively. According to the value of PSNR and level of speckle noise, the best image passed for segmentation using U-Net and classification using CNN to detect tumor type. The experiments proved the quality of U-Net and CNN in segmentation and classification respectively, since they achieved 95.11 and 95.13 in segmentation and 95.55 and 95.67 in classification as dice score and accuracy respectively.     

Anatomical region segmentation method from dermoscopic images of pigmented skin lesions

Melanoma tumor can cause a serious life threatening problem in humans, if left untreated for a long time without early diagnosis. For early diagnosis of melanoma, it is more significant to develop novel methods based on biophysics analyses, molecular targets recognitions, and new image analysis criteria. In this article, anatomical region segmentation and diameter identification is proposed to detect melanoma from dermoscopic images. Four main steps of the proposed system are as follows: In the first step, the preprocessing is performed to smooth the melanoma extraction process. The region segmentation is done in the second step using watershed segmentation and Sobel operator. In the third step, the postprocessing procedures like as morphological open, canny edge detection also applied to improve the region of interest. Finally, the melanoma region is identified using color symmetry features. The proposed method is tested with two data sets to prove the performance proposed method. The proposed method achieved an accuracy of 95.31% and specificity of 98.3%, which is better than other methods. Experimental results show that the effectiveness of the proposed method and illustrate viability of real-time clinical applications.     

Deep Learning in DXA Image Segmentation

Many existing techniques to acquire dual-energy X-ray absorptiometry (DXA) images are unable to accurately distinguish between bone and soft tissue. For the most part, this failure stems from bone shape variability, noise and low contrast in DXA images, inconsistent X-ray beam penetration producing shadowing effects, and person-to-person variations. This work explores the feasibility of using state-of-the-art deep learning semantic segmentation models, fully convolutional networks (FCNs), SegNet, and U-Net to distinguish femur bone from soft tissue. We investigated the performance of deep learning algorithms with reference to some of our previously applied conventional image segmentation techniques (i.e., a decision-tree-based method using a pixel label decision tree [PLDT] and another method using Otsu’s thresholding) for femur DXA images, and we measured accuracy based on the average Jaccard index, sensitivity, and specificity. Deep learning models using SegNet, U-Net, and an FCN achieved average segmentation accuracies of 95.8%, 95.1%, and 97.6%, respectively, compared to PLDT (91.4%) and Otsu’s thresholding (72.6%). Thus we conclude that an FCN outperforms other deep learning and conventional techniques when segmenting femur bone from soft tissue in DXA images. Accurate femur segmentation improves bone mineral density computation, which in turn enhances the diagnosing of osteoporosis.     

Mammographic mass segmentation using multichannel and multiscale fully convolutional networks

Breast cancer is one of the leading causes of death among women worldwide. Mammographic mass segmentation is an important task in mammogram analysis. This process, however, poses a prominent challenge considering that masses can be obscured in images and appear with irregular shapes and low image contrast. In this study, a multichannel, multiscale fully convolutional network is proposed and evaluated for mass segmentation in mammograms. To reduce the impact of surrounding unrelated structures, preprocessed images with a salient mass appearance are obtained as the second input channel of the network. Furthermore, to jointly conduct fine boundary delineation and global mass localization, we incorporate more crucial context information by learning multiscale features from different resolution levels. The performance of our segmentation approach is compared with that of several traditional and deep-learning-based methods on the popular DDSM and INbreast datasets. The evaluation indices consist of the Dice similarity coefficient, area overlap measure, area undersegmentation measure, area oversegmentation measure, and Hausdorff distance. The mean values of the Dice similarity coefficient and Hausdorff distance of our proposed segmentation method are 0.915 ± 0.031 and 6.257 ± 3.380, respectively, on DDSM and 0.918 ± 0.038 and 2.572 ± 0.956, respectively, on INbreast, which are superior to those of the existing methods. The experimental results verify that our proposed multichannel, multiscale fully convolutional network can reliably segment masses in mammograms.     

MRI brain segmentation in combination of clustering methods with Markov random field

Medical image segmentation is a preliminary stage of inclusion in identification tools. The correct segmentation of brain Magnetic Resonance Imaging (MRI) images is crucial for an accurate detection of the disease diagnosis. Due to in‐homogeneity, low distinction and noise the segmentation of the brain MRI images is treated as the most challenging task. In this article, we proposed hybrid segmentation, by combining the clustering methods with Hidden Markov Random Field (HMRF) technique. This aims to decrease the computational load and improves the runtime of segmentation method, as MRF methodology is used in post‐processing the images. Its evaluation has performed on real imaging data, resulting in the classification of brain tissues with dice similarity metric. These results indicate the improvement in performance of the proposed method with various noise levels, compared with existing algorithms. In implementation, selection of clustering method provides better results in the segmentation of MRI brain images.