Automated Segmentation of Overlapping Cytoplasm in Cervical Smear Images via Contour Fragments

Authors: Youyi Song, Jing Qin, Baiying Lei, Kup-Sze Choi

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using two cervical smear datasets, the performance of our method is extensively evaluated and compared with that of the stateof-the-art approaches; the results show the superiority of the proposed method.
Researcher Affiliation Academia Center for Smart Health, School of Nursing, The Hong Kong Polytechnic University National-Regional Key Technology Engineering Laboratory for Medical Ultrasound, Guangdong Key Laboratory for Biomedical Measurements and Ultrasound Imaging, School of Biomedical Engineering, Shenzhen University, China
Pseudocode No The paper describes its algorithms and methods in detail using text and mathematical equations, but it does not include any structured pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide any statement about making its source code publicly available or a link to a code repository.
Open Datasets Yes This dataset is obtained from the website of ISBI 2015 Overlapping Cervical Cytology Image Segmentation Challenge.
Dataset Splits No All these four parameters are optimized by the cross validation procedure on a small training dataset, and our experimental findings of these parameters are follows.
Hardware Specification Yes All experiments are conducted on a PC with a 2.20 GHz Intel Core i5 CPU and 4.00 GB of RAM, and they are all implemented in MATLAB.
Software Dependencies No The paper states that experiments are 'implemented in MATLAB' but does not specify the version of MATLAB or any other software dependencies with version numbers.
Experiment Setup Yes When using a small value of penalty rate ξ, the role of this energy term is diminished; however, when using a large value of it, it is more likely to lead to turbulence of graph s energy, so that graph s convergence cannot be reached. It is set as 1.5 in both datasets. The value of ω roughly depends on the overlapping degree in the dataset; the higher overlapping degree, the larger value. In our experiments, we set it to 7 in the Pap stain dataset, and to 10 in the H&E dataset. The values of ω1 and ω2 rely on imaging quality and overlapping degree. When images with high imaging quality and low overlapping degree, they both should be set a larger value; otherwise, curvature s role should be stressed greater. They are set as 1 and 0.2 in the Pap dataset, and as 0.7 and 0.4 in the H&E dataset, respectively.