Granulometry-Based Descriptor for Pathological Tissue Discrimination in Histopathological Images

Esteban, A. E.; Colomer, Adrián; Naranjo, Valery; Sales, María Á.

doi:10.1109/icip.2018.8451805

Cited by 3 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Cyan is used because it is the channel in which the differences between the four tissue components can be better differentiated. In addition, we compute the hematoxylin and eosin colour images applying the colour deconvolution method proposed in [ 29 ], which was also implemented in other studies of the state of the art related with this research field, such as [ 10 , 30 ]. The colour deconvolution method allows for separating the contributions of each stain using the Optical Density (OD) parameter Equation ( 5 ), where A is the absorbance and

is the concentration of a certain stain s :

…”

Section: Methodsmentioning

confidence: 99%

“…In other studies, such as [ 11 , 13 , 32 ], the authors also applied GLCM-based techniques but on histological regions, instead of gland units. On the other hand, we use Local Binary Patterns (LBP) to extract local intensity changes of the gland candidates, unlike other works which used LBP to segment different tissue structures [ 9 , 30 ] or discriminate between cancerous and non-cancerous patches [ 7 , 10 ]. Note that, in this case, we also use the cyan, hematoxylin and eosin channels to compute a total of 186 textural features.…”

Section: Methodsmentioning

confidence: 99%

“…Machine-learning classifiers . In order to address the classification stage through hand-driven learning methods, we make use of two different classifiers widely utilised in the literature [ 6 , 9 , 10 , 15 ]. Specifically, we optimise a Support Vector Machine (SVM) classifier, which allows for finding the optimal hyperplane h that separates two regions of the input space maximising the distance between two support vectors , one of each class [ 38 ].…”

Section: Methodsmentioning

confidence: 99%

“…For this reason, there are a lot of studies in the state of the art whose goal is to provide automatic models capable of reporting an initial indicative diagnosis from histopathological images. In most of these studies, the authors tried to perform computer-aided prognosis to discern between all Gleason grades [ 4 , 5 , 6 , 7 ] or simply between cancer and non-cancer [ 8 , 9 , 10 ]. Nevertheless, we propose for the first time a gland classification system exclusively focused on patterns corresponding to moderately differentiated carcinomas, i.e., patterns associated with Gleason grades 2 and 3, with the aim of helping the pathologists to reduce the workload and the subjectivity level when they try to diagnose this type of heterogeneous structures.…”

Section: Introductionmentioning

confidence: 99%

“…They applied a k-nearest neighbour (K-NN) classifier using textural features based on Local Binary Patterns (LBPs) and Gabor Filters. In a more recent study [ 10 ], the authors used 45 whole-slide images from which they composed sub-regions of three different dimensions:

and

pixels. The best accuracy (0.98) was achieved by applying the surfacic granulometry descriptor on the

images and implementing an SVM classifier with a linear kernel.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

First-Stage Prostate Cancer Identification on Histopathological Images: Hand-Driven versus Automatic Learning

García

Colomer

Naranjo

2019

Entropy

View full text Add to dashboard Cite

Analysis of histopathological image supposes the most reliable procedure to identify prostate cancer. Most studies try to develop computer aid-systems to face the Gleason grading problem. On the contrary, we delve into the discrimination between healthy and cancerous tissues in its earliest stage, only focusing on the information contained in the automatically segmented gland candidates. We propose a hand-driven learning approach, in which we perform an exhaustive hand-crafted feature extraction stage combining in a novel way descriptors of morphology, texture, fractals and contextual information of the candidates under study. Then, we carry out an in-depth statistical analysis to select the most relevant features that constitute the inputs to the optimised machine-learning classifiers. Additionally, we apply for the first time on prostate segmented glands, deep-learning algorithms modifying the popular VGG19 neural network. We fine-tuned the last convolutional block of the architecture to provide the model specific knowledge about the gland images. The hand-driven learning approach, using a nonlinear Support Vector Machine, reports a slight outperforming over the rest of experiments with a final multi-class accuracy of 0.876 ± 0.026 in the discrimination between false glands (artefacts), benign glands and Gleason grade 3 glands.

show abstract

is the concentration of a certain stain s :

…”

Section: Methodsmentioning

confidence: 99%