Histopathology Image Analysis and Classification for Cancer Detection Using 2D Autoregressive Model

Performance Evaluation of Cancer Diagnostics Using Autoregressive Features with SVM Classifier: Applications to Brain Cancer Histopathology

“…[1] The two basic purposes of the model design are to find the hidden properties in image sample for analysis or classification, and the second is to generate the same image using model parameters in synthesis experiment. [36] Stochastic model plays an important role in many image processing applications like image compression, analysis, segmentation, restoration and image retrieval.…”

“…While pursuing this study in the stochastic-model-based approach for histopathology image analysis and classification, we met with certain challenges such as to (i) select the proper kind of model as AR, moving average (MA), ARMA, Markov models (MRF) and also model orders, (ii) select the proper method to estimate the model parameters (iii) select proper neighbourhood, checking complexity in estimation process and accuracy in classification and (v) select classifier, which is also very crucial as classifier output (accuracy of classification) is a feature-specific component. [1] …”

“…We represent an image as 2D random field [1] {x The image I can be represented by a difference equation [1]. …”

“…Parameters derived from histopathological image sample with different model order are taken for classification, and the set which gives highest classification accuracy is considered as an optimized model order. [1] In a modelling experiment we encounter two major challenges related to the method of determining the best neighbour set N of the AR model and a method of estimating the parameters of the model. Least square (LS) estimates are consistent and efficient for parameter estimation.…”

“…[1] Human body consists of millions of cells, and mitosis is the most common form of cell division, leading to cell multiplication. Mitosis allows a body to grow.…”