A New Hybrid Method for Gland Segmentation in Histology Images

Wang, Liyang, Zhou, Yu orcid iconORCID: 0000-0002-8071-6572 and Matuszewski, Bogdan orcid iconORCID: 0000-0001-7195-2509 (2019) A New Hybrid Method for Gland Segmentation in Histology Images. In: Workshop on Visual Computing and Machine Learning for Biomedical Applications (ViMaBI), co-located with CAIP 2019, 3-5 September 2019, Salerno, Italy.

[thumbnail of Author Accepted Manuscript]
PDF (Author Accepted Manuscript) - Accepted Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.


Official URL: http://dx.doi.org/10.1007/978-3-030-29930-9_2


Gland segmentation has become an important task in biomedical image analysis. An accurate gland segmentation could be instrumental in designing of personalised treatments, potentially leading to improved patient survival rate. Different gland instance segmentation architectures have been tested in the work reported here. A hybrid method that combines two-level classification has been described. The proposed method achieved very good image-level classification results with 100% classification accuracy on the available test data. Therefore, the overall performance of the proposed hybrid method highly depends on the results of the pixel-level classification. Diverse image features reflecting various morphological gland structures visible in histology images have been tested in order to improve the performance of the gland instance segmentation. Based on the reported experimental results, the hybrid approach, which combines two-level classification, achieved overall the best results among the tested methods.

Repository Staff Only: item control page