• Nu S-Au Găsit Rezultate

View of An Automated Enhanced FCM Based Hippocampus Segmentation


Academic year: 2022

Share "View of An Automated Enhanced FCM Based Hippocampus Segmentation"

Arată mai multe ( pagini)

Text complet


An Automated Enhanced FCM Based Hippocampus Segmentation

S Vijayalakshmi1† and SAVITA2††,

s.vija[email protected] [email protected] Galgotias University, Greater Noida, India


Fragmentation of image in digital image processing is a major issue in medical field. One of the unique characteristic of image is the interaction between adjacent pixels. To extract semantic relationship between the pixels segmentation is used. Lots of algorithms have been developed for segmentation. In recent years, human hippocampus has received the focus of research because of its close link with memory and neurological disorders. To identify and to diagnose neurological disorders, hippocampus is to be segmented. But its small size and complex structure make the segmentation process a very tedious task. In this paper an enhanced Fuzzy C-Means (FCM) method is present to extract hippocampus from human brain MRI. In this proposed method, segmentation of hippocampus is done in two stages: Stage1 involves validation and preprocessing whereas stage2 comprises of segmentation using FCM. To validate and confirm the performance of the presented method with previously available tool and for the comparison between proposed and manual segmented methods Jaccard (j) and dice (d) are calculated.

Key words: Dementia, Hippocampus, Image processing, Magnetic resonance imaging, Median filter, preprocessing, clustering, Segmentation,


1. Introduction

In medical field, identification of disease at the early stage is very important. Various medical imaging modalities like positron emission, magnetic resonance, electromagnetic radiation based X-radiation, computer tomography in which computer system is used with X-ray machine, single photon emission tomography here integration result of CT scan and radioactive tracer are used for identification of disease. These are used to create a detailed picture about inside of the body. MRI is one of the best methods in comparison to CT-SCAN and x-ray because it use magnet, radio waves, that are not harmful for the body .The brain is comprises of three major parts: cerebrum, cerebellum (little brain), brain stem. Brain play a vital role in the human being and is affected by many diseases among which some are very difficult to detect. Neurological disorders are common in this category. Dementia is a neuro disorder caused by damage to brain cells creates difficulties in daily activities, thinking and language. Alzheimer disorder (AD) is the another type of memory related disease dementia and Hippocampus (HC) is the first place affected by AD. HC is located near temporal lobe having the appearance like sea horse and interconnected by cornu ammonies and dentate gyrus. HC is located in both sides of the brain. To measure the HC volume and shape, MRI is used which gives HC in three planes: Axial, coronal and sagittal. Image septation is the procedure for the classification of image pixel into dissimilar groups based on color, texture or intensity. An unsupervised clustering technique is popular approaches for image sepration which can be classified into exclusive, overlapping, hierarchical and probabilistic clustering. Lots of works have been gone on in the scope of segmentation of image by using various automatic and semiautomatic segmentation techniques. In [1] a method for image segmentation is presented in which k-means and Normalized cut algorithm is used. For prospective division of images k-means is applied. Small approximation of k has given better result


compared to Greater approximation. For Greater approximation of k, parting becomes tangible and also numerous varieties into image spots at distinctive place. N-cuts algorithm gives better result for greater approximation of k. Segmentation of Medical Image is a difficult task due to the presence of partial volume effect (PVE). For PVE, FCM is the best algorithm but it also faces a great challenge in efficacy. For segmentation of partial volume effective medical image, a two phase’s algorithm is presented in [2-4]: At first step, to deal with PVE effective medical image, improved FCM based on histogram HisFCM is utilized. HisFCM works in two stages: first Stage retrieves interval to compute cluster centroid. On second stage segmentation is done using improved FCM. At the end, comparison between fcm, fcm_s, EnFcm and HisFCM has been done by applying them on four images breast, head, tumor, and brain. HisFCM gives the best result in comparison to all others. Edge based image segmentation process is depend on the intensity in homogeneity of the picture at the edge, but they are unable to give accurate segmentation result. So, a new approach is presented by authors of boundary detection to fragment image without destroying image structure and boundary [5-7]. Before segmentation of hippocampus from human brain MRI morphological operations are applied to make the image free from noise and then k-means clustering is used to segment hippocampus. This method works on all the images which have weak edges. In [8-10, 11] segmentation of image is done by using hierarchical clustering method but takes large storage complexity for large data. Mean shift and hierarchical clustering (MSHC) are used to segment the images maintaining desirable discontinuity of image. This segmentation technique can be applied on color as well as medical images. To find out abnormality in human skin tissue, Haematoxylin and eosin (H&E) stained image, soft clustering FCM is used. This segmentation technique can find out the ROI from high resolution histological images [12-14]. Segmentation is most widely used in computer vision, pattern recognition, image processing.

A Various technique have been developed in the area of segmentation of color image but not providing sufficient result. So, to get a desired result, for segmentation of color image enhanced watershed with region growing technique is implemented. N-cut eigenvectors are used to compare the original and converted image after which watershed algorithm is applied. Segmentation is based on seed selection region growing in which identification of segmented object is easy. The proposed method and the previous method are compared by evaluating the Liu's factor value. After comparison it is found that Liu's factor value of proposed method is lower than the existing work [15]. For segmentation of hippocampus from T2w sagittal image a prior shape-based model is used. To get location of hippocampus Union of manual segmented result is calculated and ROI of hippocampus is set which lies in the range 180< p > 165. Geometric mean filter is applied to make the strong edge of hippocampus and differentiate boundary of hippocampus from other objects by using 3*3 structuring element. Image is converted into binary image because on binary image extraction of HC is easy by, using K means clustering which is taken for further segmentation. Connected components have been used to segment hippocampus from binary image. To identify and to diagnose brain abnormalities, 3D visualization of brain is utilized. MRI segmentation is the first and most important step in this procedure. Various Image segmentation techniques have been developed to segment the brain but found that unsupervised FCM and support vector machine give the accurate segmented result which helps the expert [16].

We have presented a novel method here to extract HC from human brain MRI. The remainder part of this research paper has been structured as mentioned here: part II describes methods and material used.

Section III elaborates result and discussion. Finally, we have concluded with the section IV.



Data collected from Penn hippocampus atlas resource. PHA is medical resource that contains high resolution normalized segmented MRI and histology of human hippocampus. PHA was narrated by yushkevich [17].


The proposed technique workflow is:-

Input Image


Construct HC maxima

Feature selection

Compare input image & HC maxima


Fuzzy C-Means

Segmented Hippocampus


This process comprises of two phases namely validation and segmentation. During validation input slice is checked to determine whether HC is presented or not, which can be done only by comparing some image characteristic. Here we have taken mean, standard deviation and entropy by trial and error method as well as with the historical knowledge of manual segmented result priorly available which is done as follows manually segmented images are required for creating the mathematical model only once. But after the consumption of the model no manual intervention is required in the segmentation process and it can be applied to any data set with the same resolution. At the first stage, the manual segmented results available in PHA are analyzed and the structure of maximum hippocampus (HC maxima) is obtained, which is a common shape created by union between the manual segmented hippocampus from slices numbering penn002l_01070 to penn002l_01100 which is utilized for validation. Mi(H) is a manual segmented HC in the ith slice, then HC maxima say M for the whole atlas is given by:-

𝑀 = 𝑛𝑖=1𝑀𝑖(𝐻) (1)

The HC maxima as created by equation (1) from penn002l_01070 to penn002l_01100 are shown in fig 2(a).

A sample input slice from penn002l_01070 to penn002l_01100 is given in in figure 2(b).


Figure 2. (a) HC maxima (b) input slice

Computed features like mean, standard deviation, and entropy for HC in the manual segmented result and shown in Table1. which can be used for validation. To check if HC is present in the input image from any of the slice in the atlas, the respective HC maxima region is first cropped from the input slice.

Compute mean, Standard deviation and entropy for the cropped HC maxima are compared with the priorly computed values of HC maxima. If the computed values lie within the expected range then the input slice comprises of HC and taken for further processing otherwise it is discarded.

The value of mean (Mmean) for the HC maxima is calculated

𝑀𝑚𝑒𝑎𝑛 = 𝑖𝑚 =1 𝑗𝑛=1𝑝𝑚𝑛 (2)

Where pmn denotes the intensity of the pixel and i,j denotes the dimension. Next feature is the Standard deviation (Sdeviation) of the HC maxima which is calculated as:-

𝑆𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 = 𝑝𝑚𝑛−𝑚𝑚𝑒𝑎𝑛

𝑗 𝑛 =1 𝑖𝑚 =1




To measure the image texture feature, entropy which gives information about the spatial arrangement of intensity is calculated:

𝐸𝑒𝑛𝑡𝑟𝑜𝑝𝑦 = 𝑠𝑢𝑚 ℎ.∗ 𝑙𝑜𝑔2 ℎ 4 Where h denote histogram counts.

Table 1. Features of HC maxima

Feature Min Value Max Value

Mean 186.61 195.33

Standard deviation

50.059 56.581

Entropy 6.3999 6.8427

On the next step, the image is preprocessed to make the HC more clear. To reduce the noise from the image Gaussian low pass filter is applied can be expressed as the mathematical equation:

LoG x, y = − 1

πσ2 1 −x2+ y22 e

x2+y2 2



and to make HC more prominent by using median filter which is non linear filter that moves on image pixel by pixel to preserve the edge with the parameter 15 is applied[18-21]. The bottom hat operation is applied here to make the HC boundary more distinguishable

𝐵𝑜𝑡𝑡𝑜𝑚 𝐻𝑎𝑡 𝐵𝐻 = 𝐹𝐼∙ 𝐷𝑠 − 𝐹𝐼 (6)

Where FI is median filtered image and Ds is the disk-shaped data structuring element.

On the next step to segment hippocampus from the preprocessed image a modified FCM is appled. The most efficient algorithm for classification of data is FCM given by J.C. DUNN in 1973 and later J.C.BEZDEK improved it[22-23]. This is used in pattern recognition, machine learning and image processing. According to FCM algorithm nature one part of data can belongs to more than one clusters.

In modified FCM, first we fixed the number of the clusters i.e three clusters within the HC and outside the HC. To find the distance between the data and the cluster center in the place of Euclidean distance we used mahalanobis distance(m.d) which is suitable to find out the cluster for ellipsoid type shapes and given by P.C. mahalanobis and the equation for m.d

𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑚𝑎 ℎ𝑎𝑙𝑎𝑛𝑜𝑏𝑠𝑖𝑠 = 𝑢𝑖 − 𝑘𝑗 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑠𝑒 𝑐𝑜𝑣−1( 𝑢𝑖 − 𝑘𝑗 2 (7)

where T is the metrix transpose and C is the covariance matrix of µi and kj. Now the focus is on minimizing the objective function:

𝐻𝑛 = 𝑋𝑖𝑗


𝑗 =1 𝑁


𝑢𝑖− 𝑘𝑗 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑠𝑒 𝑐𝑜𝑣−1( 𝑢𝑖− 𝑘𝑗 2 (8)

Where I≤n<∞ , N is real number which is usually more than 1, xij is membership degree of ui in group j, ui is the ith of d- dimensional measured data, kj is d-dimensional center of the group. This |∗| show the similarity between center and measure data. FCM can be carried out by using an iterative optimization process of the target function Hn, updation of xij and group center kj

𝑋𝑖𝑗 = 1

𝑢𝑖 − 𝑘𝑗 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑠𝑒 − 𝑐𝑜𝑣1−1 𝑢𝑖− 𝑘𝑗 𝑢𝑖− 𝑘𝑗 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑠𝑒 − 𝑐𝑜𝑣2−1 𝑢𝑖 − 𝑘𝑗

2 𝑞−1 𝑐𝑘=1


\Where 𝐾𝑗 = 𝑥𝑖𝑗

𝑛.𝑢𝑖 𝑁𝑖=1

𝑥𝑖𝑗𝑛 𝑁 𝑖=1

, 𝐶𝑜𝑣1−1

𝑎𝑛𝑑 𝐶𝑜𝑣2−1

are inverse of covariance matrix of µi and ki and µi and kj.

This iteration break off when maximum {|xij(M+1) –xijM |}<€, where € is a conclusion which will be decided between 0 and 1 where M is series of repetitive steps.

By historical analysis of data set we have determined that HC is enclosed within the window boundary with the following values of ROWmin=82, ROWmax=146, COLUMNmin=72, COLUMNmax =216 [16-19]. To get HC, we have to get rid of other objects within the cluster so discard all the objects which has pixel value outside the ROWmin=82, ROWmax=146, COLUMNmin=72, COLUMNmax

=216 values and then rest of the object is HC.


3. Result and Discussion

A novel method using FCM has been proposed here to extract HC efficiently in MRI of human brain.

The segmented results of some sample slices are presented in table 2. In data set total number of slice is 130 and HC Is available from slice number Penn002L_01_Seg_JP_2008002 to Penn002L_01_Seg_JP_2008117and the proposed method is suitable to extract HC from all the available slices in which HC is available.

Table 2. Segmented results of sample slices

The performance analysis is given in the table 3, which shows that the proposed technique provides result closer to the gold standard as well as proved to be better than the existing method. This technique also reduces the time complexity and can assist the physician for diagnosing the AD.

Slice No.

Manual segmentation

Proposed segmentation



Penn002 L _01078

Penn002L _01084Penn002L _01096Penn002L _01099Penn002L _01100


Table 3. Performance Analysis

For quantitative analysis with the existing method we have computed the average values of performance metrics such as Jaccard coefficient(J), Sorensen-Dice or Dice similarity coefficient, sensitivity(S) which is used to correctly identified the specific disease , specificity(Sp) also known as selectivity, predictive accuracy (PA), between the hand segmented images, and the images segmented by the our presented method and software ITK-SNAP tool [26-27]. The calculated values of Jaccard, Sorensen Dice, Selectivity, Specificity, and predictive accuracy for our presented method and that of the ITK-SNAP against the gold standard given in Table 3.

The snapshots of this active contour evolution are shown in Figure 3 Slic


Jac car d

Dic e

Sen sitiv it y

Spe cific it y

Predi ctive accu racy



Pen n00 2L_

010 78

0.99 86 0.9

99 2

0.99 90 0.7

29 2

99.8 614 0.0

004 0.0 010

Pen n00 2L_

010 84


99 91


99 95

0.99 94

0 . 9 2 8


90 84

0.0 003

0.0 006

Pen n00 2L_

010 96

0.99 91


99 95

0.99 92

0.9 01 4

99.9 1

0.0 001

0.0 008

Pen n00 2L_

010 99


80 29

0.8 90


0.84 81

0.9 97 7

99.1 762

0.0 563

0.1 519

Pen n00 2L_

011 00

0.88 92

0.9 41


0.91 02

0.9 98 9

99.4 845

0.0 236

0.0 898


Pro pos ed met hod

0.9 37 78

0.96 604

0.9 51 18

0.91 104


668 1

0.0 161 4


0.8 67


0.92 88 0.8

98 8

0.99 84 99.

422 4

0.0 361


Figure 3: - ITK-SNAP snapshots for HC segmentation

For quantitative analysis we compute the jaccard, Dice, FPR (false positive rate), sensitivity, specificity and FNR (false negative rate) which are computed as follows: -

𝐽𝑎𝑐𝑐𝑎𝑟𝑑 𝐽𝐷 =𝐴∩𝐵

𝐴∪𝐵 𝐷𝑖𝑐𝑒 𝐷 = 2 𝐴∩𝐵

𝐴 + 𝐵

𝑺𝒆𝒏𝒔𝒊𝒕𝒊𝒗𝒊𝒕𝒚 𝑺𝑵 = 𝑻𝑷

𝑻𝑷+𝑭𝑵 𝑺𝒑𝒆𝒄𝒊𝒇𝒊𝒄𝒊𝒕𝒚(𝑺𝑷) = 𝑻𝑵


Any misclassification done by proposed method can be computed by using the quantitative function false positive rate (FPR ) and false negative rate :



The predictive accuracy is calculated as: -

𝑃𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑣𝑒 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 𝑃𝐴 = 100 ∗ 𝑇𝑃 + 𝑇𝑁 𝑇𝑃 + 𝐹𝑃 + 𝑇𝑁 + 𝐹𝑁

The calculated average values of quantitative analysis Jaccard similarity (intersection over union) and Sorensen dice are 0.94 and 0.97 respectively which is near to 1 that shows the accuracy of our method.


Our proposed technique works well for the data set available in PHA. In future this method can be enhanced to provide 3-D HC.


[1] P. Srinivasa Rao. etal, Image Segmentation using Clustering Algorithms. International Journal of Computer Applications (0975 –8887)Volume 120 –No.14, June 2015,pp.36-38.

[2] K.Somasundaram, S.Vijayalakshmi, A Fcm Based Approach For Automated Segmentation Of Breast Masses In Mammograms. International Journal of Computational Science and Information Technology, ISSN: 2320-7442 (Online); 2320 - 8457 (Print), Feb-2016

[3] K.Somasundaram* and S.Vijayalakshmi, A Semi-Automatic Atlas-Based Method for Segmentation of the Hippocampus from the mid slices of brain MRI. International Journal of Computer Application, ISSN- 2250-1797 , Oct-2012 , ICV=6.1.


[4] S. Vijayalakshmi, Savita, A Semi-Automated K-Means Clustering Based Hippocampus Segmentation. International Journal of Electrical Electronics & Computer Science Engineering Special Issue - NCCT- 2018 | E-ISSN : 2348-2273 | P-ISSN : 2454-1222 February, 2018

[5] S.Vijayalakshmi and savita, Medical Image Segmentation Using Various Techniques: A Survey.

International Journal of Recent Trends in Engineering and Research, Volume 3, Issue 2; February 2017.Issn 2455-1457,Pp 120-130

[6] C. Chijindu, H. C. Inyiama, and G. Uzedhe, Medical image segmentation methodologies – a classified overview. IEEE African Journal of Computing & ICT, 2012,vol. 5, no. 5. pp. 100-108.

[7] K.Somasundaram, S.Vijayalakshmi, A Fcm Based Approach For Automated Segmentation Of Breast Masses In Mammograms. International Journal of Computational Science and Information Technology, ISSN: 2320-7442 (Online); 2320 - 8457 (Print), Feb-2016

[8] S. Vijayalakshmi, Savita, A Semi-Automated K-Means Clustering Based Hippocampus Segmentation. International Journal of Electrical Electronics & Computer Science Engineering Special Issue - NCCT- 2018 | E-ISSN : 2348-2273 | P-ISSN : 2454-1222 February, 2018.

[9] Kishore Gunna, Implementation and Comparison of Different Segmentation Techniques for Medical Images. International Journal of Computer Applications (0975 – 8887) January 2016.Volume 134.

[10] S.Vijayalakshmi and savita, Hippocampus segmentation techniques: A survey. International Journal of Recent Trends in Engineering & Research (IJRTER) Volume 03, Issue 08; August - 2017 [ISSN: 2455- 1457], pp.223-233

[11] Kewal Krishan .et al, Color Image Segmentation Using Improved Region Growing and K-Means Method. ISSN (e): 2250-3021, ISSN (p): 2278-8719 Vol. 04, Issue 05 (May. 2014), ||V4|| PP 43-46

[12] K.Somasundaram and S.Vijayalakshmi, A Simple Approach For Relatively Automated

Hippocampus Segmentation From Sagittal View of Brain MRI. International Journal on Bioinformatics & Biosciences, pp.23-30. 2016

[13] K. Somasundaram1and T. Genish2, THE KNOWLEDGE BASED SEGMENTATION OF


[14] P. Yuehao , W. Huang, Z. Lin, W. Zhu, J. Zhou, J. Wong, Z. Ding, Brain Tumor Grading Based on Neural Networks and Convolutional Neural Networks. 978-1-4244-9270-1/15/$31.00 ©2015 IEEE.

[15] A. Nyma, M. Kang, Y. KeunKwon, C. HongKim, and J. MyonKim, A Hybrid Technique for Medical Image Segmentation. Hindawi Publishing Corporation Journal of Biomedicine and Biotechnology Volume 2012, Article ID 830252,7pagesdoi:10.1155/2012/830252.

[16] S.Vijayalakshmi, savita, A Review and Proposal of an Innovative Approach for Medical Image Segmentation. IEEE, ISBN:978-1-5090-6471-7/17/$31.00 ©2017 IEEE 1167

[17] Paul A. Yushkevich, .et al, Dawn Mechanic-Hamilton, Simon Glynn, Stephen Pickup, Weixia Liu, James


[18] C. Gee, Murray Grossman, & John A. Detre (2008), A high-resolution computational atlas of the human hippocampus from postmortem magnetic resonance imaging at 9.4 T , Neuroimage vol.44, pp.385-398.

[19] K.Somasundaram and S.Vijayalakshmi, A Simple Automatic Tool For Segmenting the Right Hippocampus from Mid-Slices of Brain MRI. International Journal of Scientific Research and Publications, Sep-2012, ISSN 2250-3153.

[20] D. Kaushik, S, Utkarsha , P. Singha, and V. Singh, Medical Image Segmentation using Genetic Algorithm. Int. Journal of Computer Applications (0975 – 8887) , November 2013,Volume 81 .

[21] M. Rastgarpour, and J. Shanbehzadeh, The problems, applications and growing interest in automatic segmentation of medical images from the year 2000 till 2011. Int. Journal of Computer Theory and Engineering, 2013, vol. 5.

[22] T. Schoenemann, F. Kahl, S. Masnou, and D. Cremers, A linear framework for region-based image segmentation and inpainting involving curvature penalization.Int. Journal of Computer Vision, vol.

99, 2012, pp. 53-68.

[23] S.Vijayalakshmi, Savita, Fuzzy C-Means Based Automated Technique for Hippocampus

Segmentation. International Journal of Computer Sciences and Engineering(IJCSE), Vol.6(4), May 2018.

[24] Rupesh Mandal, et al, Histological Image Segmentation using Fuzzy C-Means. international Journal of Computer Applications (0975 –8887) International Conference on Computing &

Communication (ICCC2016),pp.1-4.

[25] K.Somasundaram and S.Vijayalakshmi, A Semi-Automatic Atlas-Based Method for Segmentation

of the Hippocampus from the mid slices of brain MRI. International Journal of Computer Application, ISSN- 2250-1797, Oct-2012 , ICV=6.1.

[26] Rafael C. Gonzelez and Richard E Woods, 2002, Digital Image Processing, Second edition, Pearson Education

[27] ITK-SNAP 2.4, http://www.itksnap.org/pmwiki/pmwiki.php/

Dr. S.Vijayalakshmi, Professor, Galgotias University Vijayalakshmi S was born in the year 1975. She received the B.Sc. degree in

Computer Science from Bharathidasan University, Tiruchirappalli, India in 1995, the MCA degree from the same University in 1998 and the M.Phil.

degree from the same University in the year 2006. She received her doctorate in 2014. She has been working as a Professor, Galgotias University, Greater Noida, Uttar Pradesh, India. she has 19 years of teaching experience and 10 years of research experience. She has published one patent, more than fifty research papers in the area of image processing especially in medical imaging and have published few book chapters in the area of internet of things and image processing and authored books under process.


SAVITA, Research Scholar, Galgotias University

SAVITA was born in the year 1988. She received the Bachelor Of Art degree from Hindu Girl college, Sonipat in year 2007. She received the MCA degree from SBIT College, Sonipat in year 2010. She has 2 years of teaching experience. She is currently pursuing Ph.D. in School of computing science & engg., Galgotias University, Greater Noida since 2016. She has published 6 research paper in the area of image processing and 3 book chapters.



Using a UNET-based deep neural network, Dong et al., (2017) developed a completely automated brain tumor identification and segmentation approach.. They demonstrate that

Two different objective functions are used- image entropy is used as objective function for tumor segmentation and accuracy function is used as objective function for

The analysis of data in microarray has provide the result in the terms of accuracy and robustness.In this paper presents theprotracted K-means NeT an efficient technique

These investigations consist of four stages, image preprocessing, segmentation of image using k-mean clustering, detection, and tracking of breast cancer location and

In coming recent time unverified approach has get experts interest in not only for their outstanding performance but also the mechanically generate features increase in the amount

Pre-processing step plays an important role before images segmentation. First, the image is converted to grayscale. A homo-morphic filter is applied to it. A

(2014), detected bone cancer from the MRI scan imagery utilizing K-Means Clustering techniques for segmentation and clustering and for detection he used Man Pixel Intensity method

These factors should be envisaged by Romanian society when trying to make both the minority and the majority cooperate in order to avoid the transformation of the Roma

In this study, K-Means based Morphological segmentation and deep learning based VGGNet 16 classification is carried out to recognize the outcomes of colorectal

This article presents GATT, principles of commercial multilateralism (democratic principles), GATT evolution to WTO (causes that compromised GATT system, the importance of

discrepancy (or supervised) methods – compare the results with a reference image or gold standard or ground truth; measures as the number of misclassified voxels, position of

Finally, an algorithm for clustering by graph flow simulation, TribeMCL [21], is used to cluster the interaction network and then to reconvert the identified clusters from

The number of vacancies for the doctoral field of Medicine, Dental Medicine and Pharmacy for the academic year 2022/2023, financed from the state budget, are distributed to

“Alexandru Ioan Cuza”, University of Iași Department of Computer Science...

The calculation names the picture dependent on the diagram cuts method and fuses the shape earlier utilizing a procedure dependent on level-sets.. The strategy requires

Brain tumor segmentation using convolutional neural networks in MRI images. The multimodal brain tumor image segmentation benchmark

Here, an effectual segmentation is utilized to segment hand and tongue images with a algorithm called Morphological with random walker segmentation for accurate

Brain Tumor Segmentation and Classification based on Deep Learning-Based Inception Networks..

By learning from the labeled hippocampus region, a model mapping is built on original image and the segmented result to effectively segments the left and right side

In this paper, identifying the immune cells, that can find the tumor cells and for the segmentation of brain image is used Intuitionistic Fuzzy Clustering Mean

Multi Model Clustering Segmentation and Intensive Pragmatic Blossoms (Ipb) Classification Method based Medical Image Retrieval SystemR. 2 Head &amp; Associate Professor, Department

Diabetic Foot Ulcer (DFT) is a problem worldwide and prevention is crucial. We hypothesized that the inability of the skin to respond to skin is involved in DFU pathogenesis and

Background: This research article developed a hybrid model called Hierarchical K-means clustering with Fuzzy c and Super Rule tree (HKMFSRT-Model) to introduce an asynchronous