Image segmentation methods Comparison with MRI

16 State of the art image segmentation methods are compared with 117 MRI LV images

Comparisons of state of the art methods for Left Ventricle Segmentation
16 state of the art methods are compared:
The 16 state-of-the-art segmentation methods are SDD thresholding method

(zhenzhou_threshold_selection_updated.m), Adaptive thresholding (adaptivethreshold.m and

averagefilter.m), Active contour without edge(region_seg.m), Region-based active contour

(local_AC_MS.m and local_AC_UM.m), GM(CMF_Cut.m), Level set (drlse_edge.m), Cross entropy

(minCE1.m), Fuzzy entropy (entropy_fuzzy_threshold1.m), ISO (isodata1.m), Maximum entropy

(maxentropie.m), Otsu (otsu.m), EM (EMSeg.m), K-means (kmeans.m), Soft (fth.m and

im_expand.m), Fuzzy C means ( fcmthresh.m and trapmf_mat.m) and Local thresholding

(niblack.m) respectively.

117 images are used to test the image segmentation methods:
The Cardiac MR images and the benchmark manual contours are from Medical Image

Computing and Computer Assisted Intervention (MICCAI) 2009. In this MATLAB cabinet, the

117 images are from Online_SC_HF_I9，Online_SC_HF_I10, Online_SC_HF_I11,

Online_SC_HYP12, Training_SC_HF_I4，Training_SC_HF_NI3, Training_SC_HF_NI4 and

Training_SC_HF_NI36. The data of the region of interest of these 117 images are integrated

into 117_ROI.mat and the benchmark manual contours corresponding to these images are

integrated into 117_Manual.mat.

APD and DICE are the measures to evaluate the segmentation accuracy. The results will be

writted into 117frame_data.xlsx.

Please Run Demo.m and QuantitativeComparison.m for the comparisons.

The codes are written by ZhenZhou Wang and JingJing Xiong, Please contact

[e-mail] for more information.

For the LV segmentation, please kindly refer to my past publications:
[1] Robust and automatic diagnosis of the intraventricular mechanical dyssynchrony for the

left ventricle in cardiac magnetic resonance images, International Journal of Computer

Assisted Radiology and Surgery,2017
[2] Segmentation of the Left Ventricle in Short-axis Sequences by Combining Deformation

Flow and Optical Flow, IET Image Processing, 2016
[3] A New Approach for Automatic Identification of the Ventricular Boundary in Cine

Magnetic Resonance Images, Journal of Medical Imaging and Health Informatics, 2016
[4] An Efficient and Robust Method for Automatically Identifying the Left Ventricular

Boundary in Cine Magnetic Resonance Images, IEEE Transactions on Automation Science and

Engineering, 2016