Cervical cell segmentation
Song et.al (2017) proposed a frame work of deep learning technique and
deformation model to segment cervical cell accurately from overlapping clumps.
This frame work not only helps in shape correlation of overlapping cells but
also by the proposed method it is helpful in utilizing the cell structure
contextual information, the proposed convolutional neural network (CNN) learns
features at different scales. Therefore, contexts of multiple scales are
integrated into local decisions to correctly classify each pixel into
background, cytoplasm, or nucleus pixel.
cells labeling method is proposed to get accurate splitting of detected
overlapping cytoplasm. The classi?ed cytoplasm pixels are labeled with the
corresponding nucleus labels generated by CNN. Gaussian kernels ?tting is used
to estimate the shape of the cell, therefore the dependence of color information
is obtained by the shape cues. Two different data sets have been
used in the proposed framework and the segmentation results are compared with
the state-of-the-art approaches. Additionally, the
proposed method is also effective in segmenting abnormal cells, even for images
having large number of overlapped cells and high degrees of overlapping.
Xue et.al (2010) analyzed and separated clinically a definite area from uterine
cervix image in a huge database created for the research of cervical cancer.
Boundary Marketing tool segmentation (BMT), cervigram segmentation tool (CST),
Multi-observer segmentation Evaluation system (MOSES) are the method followed
in medical image processing. Cervicographic image processing algorithm is used to
design for Cervicographic image processing. The proposed algorithm is nominal
in cost in late years. Similar to colposcopy, Cervicographic is based on the
aceto whitening, a transient phenomenon observed when a diluted acetic acid
solution is applied to the surface of the cervix.