Figure 4
Examples with tumor correctly located
Source publication
Computer Aided Diagnosis (CAD) was approved to automate breast cancer detection with mammograms in 1998. But due to the great variability in tumor sizes and shapes, and underlying breast tissue structures, pattern recognition algorithms have a difficult time adapting to different situations. In this paper, a marker-controlled watershed segmentation...
Context in source publication
Context 1
... ill-defined masses and normal. The first six categories are considered abnormal. The mammograms were originally digitized with 50 micron pixel edge but then reduced to 200 micron pixel edge. The size of each image is 1024x1024 and the bit depth per pixel is 8. Other information, such as breast tissue type, severity of abnormality, and location of abnormalities, is also provided. In this study, we selected circumscribed, spiculated and miscellaneous masses as types of breast tumors that were to be identified. A total of 48 images were used with 22 being circumscribed (CIRC), 19 being speculated (SPIC), and 7 being miscellaneous (MISC). A couple images have more than one tumor; the total number of tumors in these images was 50. Table 1 showed the results of the study. 87% of the circumscribed masses, 94% of the spiculated masses, and 85% of the miscellaneous masses were correctly identified. The overall detection rate was 90%. Figure 4 showed three pairs of mammograms with tumor locations correctly identified. For each pair of images, the one on the left was the original mammogram with tumor location marked in red circle; the one on the right showed the watershed segmentation result. Regions of different colors represented different segments. As can be seen, the proposed watershed segmentation is promising in locating tumor candidates. In some cases, the segmented region matched the mass tumor very well, e.g. mdb010. In other cases, the segmented regions were either bigger, e.g. mdb181, or smaller, e.g. mdb267, than actual mass tumor sizes. These were generally because background markers were either not restrictive enough or too restrictive. An improvement could be made in locating background markers more accurately. Another issue shown was that there were multiple false positive regions located in each pair. As we indicated in Section 2, the primary goal for this work was to locate all real tumors. False positives will be removed later. That is part of our ongoing work. In this paper, a marker-controlled watershed segmentation for breast tumor candidates detection was investigated. Instead of applying watershed segmentation directly on mammograms, we studied a morphological approach to clean up images and then determined foreground and background markers, which addressed the issue of over-segmentation and made the watershed segmentation result more reliable. The experiment with MIAS showed a 90% detection rate for mass tumors. Future work will be directed toward finding ways to better locate foreground and background markers in order to improve the detection rate. Another focus is to remove false positives from the results. The goal of segmentation step is to find all mass candidates even with some false positives. A natural step after segmentation will be to remove false positives as much as possible. Finally, we would like to test our algorithm on a larger data set to generate stronger ...
Citations
... Its marker region is corrected by AMS. Taking full consideration of the advantages of classical segmentation algorithms, such as the level set method [5], morphological snake (MS) [9] and MW [31], we find that AMSMW has higher segmentation precision and is 3-4 times faster than other existing methods. ...
... In terms of traditional segmentation methods, we implemented some related and classical methods that include level set [5], MS [9], MW [31], and FSMW [30]. For the MS method, we set its initialization position and radius to be the centre of the RROI and 70% of the smallest length and width of the RROI. ...
... Quantitative results of different segmentation methodsBold indicates the best results in the current column MW: marked watershed[31]; Level set[5]; MS: morphological snake[9]; AMS: adaptive morphological snake; FSMW[30] ACC (%) TPR (%) FPR (%) DSC (%) ...
Background
Breast cancer is one of the most serious diseases threatening women’s health. Early screening based on ultrasound can help to detect and treat tumours in the early stage. However, due to the lack of radiologists with professional skills, ultrasound-based breast cancer screening has not been widely used in rural areas. Computer-aided diagnosis (CAD) technology can effectively alleviate this problem. Since breast ultrasound (BUS) images have low resolution and speckle noise, lesion segmentation, which is an important step in CAD systems, is challenging.
Results
Two datasets were used for evaluation. Dataset A comprises 500 BUS images from local hospitals, while dataset B comprises 205 open-source BUS images. The experimental results show that the proposed method outperformed its related classic segmentation methods and the state-of-the-art deep learning model RDAU-NET. Its accuracy (Acc), Dice similarity coefficient (DSC) and Jaccard index (JI) reached 96.25%, 78.4% and 65.34% on dataset A, and its Acc, DSC and sensitivity reached 97.96%, 86.25% and 88.79% on dataset B, respectively.
Conclusions
We proposed an adaptive morphological snake based on marked watershed (AMSMW) algorithm for BUS image segmentation. It was proven to be robust, efficient and effective. In addition, it was found to be more sensitive to malignant lesions than benign lesions.
Methods
The proposed method consists of two steps. In the first step, contrast limited adaptive histogram equalization (CLAHE) and a side window filter (SWF) are used to preprocess BUS images. Lesion contours can be effectively highlighted, and the influence of noise can be eliminated to a great extent. In the second step, we propose adaptive morphological snake (AMS). It can adjust the working parameters adaptively according to the size of the lesion. Its segmentation results are combined with those of the morphological method. Then, we determine the marked area and obtain candidate contours with a marked watershed (MW). Finally, the best lesion contour is chosen by the maximum average radial derivative (ARD).
... The greatest difficulty in this process is image segmentation, which requires to segment tumor and normal tissue. Existing image segmentation methods are mainly based on boundary detection (such as Sobel, Canny, LoG and etc), active contour model (ACM) [5], threshold classification [6], snake model [7], watershed [8], Markov Random Field [9] and etc. They are all used to extract appropriate ROI referred to the differences of image boundaries. ...
Ultrasound image diagnosis of breast tumors has been widely used in recent years. However, there are some problems of it, for instance, poor quality, intense noise and uneven echo distribution, which has created a huge obstacle to diagnosis. To overcome these problems, we propose a novel method, a breast cancer classification with ultrasound images based on SLIC (BCCUI). We first utilize the Region of Interest (ROI) extraction based on Simple Linear Iterative Clustering (SLIC) algorithm and region growing algorithm to extract the ROI at the super-pixel level. Next, the features of ROI are extracted. Furthermore, the Support Vector Machine (SVM) classifier is applied. The calculation states that the accuracy of this segment algorithm is up to 88.00% and the sensitivity of the algorithm is up to 92.05%, which proves that the classifier presents in this paper has certain research meaning and applied worthiness.
... [15] evaluated CAD influence inspects execution in recognizing early lung disease on chest radiographs. [16] depicted CAD segmentation based on marker-controlled watershed procedures to find possible masses of tumor in the breast. ...
A vital necessity for clinical determination and treatment is an opportunity to prepare a procedure that is universally adaptable. Computer aided diagnosis (CAD) of various medical conditions has seen a tremendous growth in recent years. The frameworks combined with expanding capacity, the coliseum of CAD is touching new spaces. The goal of proposed work is to build an easy to understand multifunctional GUI Device for CAD that performs intelligent preparing of lung CT images. Functions implemented are to achieve region of interest (ROI) segmentation for nodule detection. The nodule extraction from ROI is implemented by morphological operations, reducing the complexity and making the system suitable for real-time applications. In addition, an interactive 3D viewer and performance measure tool that quantifies and measures the nodules is integrated. The results are validated through clinical expert. This serves as a foundation to determine, the decision of treatment and the prospect of recovery.
... (2) Because compared with the traditional opening and closing operator, opening by reconstruction and closing by reconstruction are less destructive and can maintain the object shape better (Lewis and Dong, 2012). Thus, the marker image is derived based on the morphological operations, including opening by reconstruction and closing by reconstruction, from the original RGB images. ...
... (2) Because compared with the traditional opening and closing operator, opening by reconstruction and closing by reconstruction are less destructive and can maintain the object shape better (Lewis and Dong, 2012). Thus, the marker image is derived based on the morphological operations, including opening by reconstruction and closing by reconstruction, from the original RGB images. ...
Indoor scenes have the characteristics of abundant semantic categories, illumination changes, occlusions and overlaps among objects, which poses great challenges for indoor semantic segmentation. Therefore, we in this paper develop a method based on higher-order Markov random field model for indoor semantic segmentation from RGB-D images. Instead of directly using RGB-D images, we first train and perform RefineNet model only using RGB information for generating the high-level semantic information. Then, the spatial location relationship from depth channel and the spectral information from color channels are integrated as a prior for a marker-controlled watershed algorithm to obtain the robust and accurate visual homogenous regions. Finally, higher-order Markov random field model encodes the short-range context among the adjacent pixels and the long-range context within each visual homogenous region for refining the semantic segmentations. To evaluate the effectiveness and robustness of the proposed method, experiments were conducted on the public SUN RGB-D dataset. Experimental results indicate that compared with using RGB information alone, the proposed method remarkably improves the semantic segmentation results, especially at object boundaries.
... Recent studies for tumor segmentation have been successfully applied to regionbased techniques for tumor segmentation. Lewis et. al [11] employ Watersheds to automatically segment tumor candidate regions, achieving an overall detection rate for mass tumors of 90%. However, the used metric of analysis was based only on tumor location, not on the quality of segmentation. ...
... The results were analyzed for all images with lesions from MiniMIAS, corresponding to 57 images divided into circumscribed, spiculated and those with illdefined margins. Six state-of-the-art works were compared to the Fuzzy GrowCut technique: BEMD [21], BMCS [30], MCW [11], Topographic Approach [13], and Wavelet Analysis [20]. ...
According to the World Health Organization, breast cancer is the most common cancer in women worldwide, becoming one of the most fatal types of cancer. Mammography image analysis is still the most effective imaging technology for breast cancer diagnosis, which is based on texture and shape analysis of mammary lesions. The GrowCut algorithm is a general-purpose segmentation method based on cellular automata, able to perform relatively accurate segmentation through the adequate selection of internal and external seed points. This chapter shows an adaptive semisupervised version of the GrowCut algorithm, based on the modification of the automaton evolution rule by adding a Gaussian fuzzy membership function in order to model nondefined borders. In this proposal, manual selection of seed points of the suspicious lesion is changed by a semiautomatic stage, where just the internal points are selected by using a differential evolution algorithm. We evaluated the proposal using 59 lesion images obtained from MiniMIAS database. The results were compared with the semisupervised state-of-the-art approaches bidimensional empirical mode decomposition, breast mass contour segmentation, wavelet analysis, topographic approach, and marker-controlled watershed (MCW). The results show that fuzzy GrowCut achieves better results for circumscribed, spiculated lesions, and ill-defined lesions, considering the similarity between segmentation results and ground-truth images.
... Through the image samples and the features of nuclei, the watershed algorithm can divide the squamous epithelium image into different small regions. A previous study [21][22][23] also exploited the watershed method that has been widely used in medical image segmentation. Watershed segment method is based on topological theory. ...
The issue of an automated approach for detecting cervical cancer is proposed to improve the accuracy of recognition. Firstly, the cervical cancer histology source images are needed to use image preprocessing for reducing the impact brought by noise of images as well as the impact on subsequent precise feature extraction brought by irrelevant background. Secondly, the images are grouped into ten vertical images and the information of texture feature is extracted by Grey Level Co-occurrence Matrix (GLCM). GLCM is an effective tool to analyze the features of texture. The textures of different diseases in the source image of Cervical Cancer Histology (such as contrast, correlation, entropy, uniformity and energy, etc.) can all be obtained in this way. Thirdly, the image is segmented by using K-means clustering and Marker-controlled watershed Algorithm. And each vertical image is divided into three layers to calculate the areas of different layers. Based on GLCM and lesion area features, the tissues are investigated with segmentation by using Support Vector Machine (SVM) method. Finally, the results show that it is effective and feasible to recognize cervical cancer by automated approach and verified by experiment.
... Recent studies for tumor segmentation have been successfully applied to region-based techniques for tumor segmentation. Lewis et. al [9] employ Watersheds to automatically segment tumor candidate regions, achieving an overall detection rate for mass tumors of 90%. However, the used metric of analysis was based only on tumor location, not on the quality of segmentation. ...
... Jaccard index (J) indicates how similar the segmentation is when compared to its ground truth. The measure is given by the Equation 9: ...
... In addition, the concept of morphological reconstruction and marker extraction can be used to eliminate over-segmentation that results from the application of watershed transformation on a gradient image. Morphological watershed segmentation is widely described in the literature (Haris et al., 1998;Pesaresi and Benediktsson, 2001;Mukhopadhyay and Chanda, 2003;Nguyen et al., 2003;Nallaperumal et al., 2007;Parvati et al., 2008;Gonzalez and Ballarin, 2009;Hamarneh and Li, 2009;Han et al., 2012;Lewis and Dong, 2012). ...
Watershed transformation is an effective segmentation algorithm that originates from the mathematical morphology field. This algorithm is widely used in medical image segmentation because it produces complete division even under poor contrast. However, over-segmentation is its most significant limitation. Therefore, this article proposes a combination of watershed transformation and the expectation-maximization (EM) algorithm to segment MR brain images efficiently. The EM algorithm is used to form clusters. Then, the brightest cluster is considered and converted into a binary image. A Sobel operator applied on the binary image generates the initial gradient image. Morphological reconstruction is applied to find the foreground and background markers. The final gradient image is obtained using the minima imposition technique on the initial gradient magnitude along with markers. In addition, watershed segmentation applied on the final gradient magnitude generates effective gray matter and cerebrospinal fluid segmentation. The results are compared with simple marker controlled watershed segmentation, watershed segmentation combined with Otsu multilevel thresholding, and local binary fitting energy model for validation. © 2016 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 26, 225–232, 2016
... Recent studies for tumor segmentation have been successfully applied to region-based techniques for tumor segmentation. Lewis and Dong (2012) uses Watershed to automatically segment tumor candidate regions, achieving an overall detection rate for mass tumors of 90%. However, the metric of analysis that was used was based only on the location of the tumor and not on the quality of segmentation. ...
... Our proposal was compared to six state-of-the-art works: BEMD ( Jai-Andaloussi et al., 2013 ), BMCS ( Berber, Alpkocak, Balci, & Dicle, 2013 ), LBI ( Sharma & Khanna, 2013 ), MCW ( Lewis & Dong, 2012 ), Topographic Approach ( Hong & Sohn, 2010 ) and Wavelet Analysis ( Pereira, Ramos, & Do Nascimento, 2014 ). Each technique was implemented using the parameters provided by each article. ...
