Chamidu Atupelage

Computational grading of hepatocellular carcinoma using multifractal feature description

Cancer grading has become an important topic in the field of image interpretation-based computer aided diagnosis systems. This paper proposes a novel feature descriptor to observe the characteristics of histopathological textures in a discriminative manner. The proposed feature descriptor utilizes fractal geometric analysis with four multifractal measures to construct an eight dimensional feature space. The proposed method employed a bag-of-feature-based classification model to discriminate a set of hepatocellular carcinoma images into five categories according to Edmondson and Steiners grading system. Three feature selection methods were utilized to obtain the most discriminative features of codeword dictionary (codebook). Furthermore, we incorporated four other textural feature descriptors: Gabor-filters, LM-filters, local binary patterns, and Haralick, to obtain a benchmark of the accuracy of the classification. Two experiments were performed: (i) classifying non-neoplastic tissues and tumors and (ii) grading the hepatocellular carcinoma images into five classes. Experimental results indicated the significance of the multifractal features for describing the histopathological image texture because it outperformed other four feature descriptors. We graded a given ROI image by defining a threshold-based majority-voting rule and obtained an average correct classification rate around 95% for five classes classification.

Computational hepatocellular carcinoma tumor grading based on cell nuclei classification

Abstract. Hepatocellular carcinoma (HCC) is the most common histological type of primary liver cancer. HCC is graded according to the malignancy of the tissues. It is important to diagnose low-grade HCC tumors because these tissues have good prognosis. Image interpretation-based computer-aided diagnosis (CAD) systems have been developed to automate the HCC grading process. Generally, the HCC grade is determined by the characteristics of liver cell nuclei. Therefore, it is preferable that CAD systems utilize only liver cell nuclei for HCC grading. This paper proposes an automated HCC diagnosing method. In particular, it defines a pipeline-path that excludes nonliver cell nuclei in two consequent pipeline-modules and utilizes the liver cell nuclear features for HCC grading. The significance of excluding the nonliver cell nuclei for HCC grading is experimentally evaluated. Four categories of liver cell nuclear features were utilized for classifying the HCC tumors. Results indicated that nuclear texture is the dominant feature for HCC grading and others contribute to increase the classification accuracy. The proposed method was employed to classify a set of regions of interest selected from HCC whole slide images into five classes and resulted in a 95.97% correct classification rate.

Multifractal Feature Descriptor for Histopathology

Background: Histologic image analysis plays an important role in cancer diagnosis. It describes the structure of the body tissues and abnormal structure gives the suspicion of the cancer or some other diseases. Observing the structural changes of these chaotic textures from the human eye is challenging process. However, the challenge can be defeat by forming mathematical descriptor to represent the histologic texture and classify the structural changes via a sophisticated computational method. Objective: In this paper, we propose a texture descriptor to observe the histologic texture into highly discriminative feature space. Methods: Fractal dimension describes the self-similar structures in different and more accurate manner than topological dimension. Further, the fractal phenomenon has been extended to natural structures (images) as multifractal dimension. We exploited the multifractal analysis to represent the histologic texture, which derive more discriminative feature space for classification. Results: We utilized a set of histologic images (belongs to liver and prostate specimens) to assess the discriminative power of the multifractal features. The experiment was organized to classify the given histologic texture as cancer and non-cancer. The results show the discrimination capability of multifractal features by achieving approximately 95% of correct classification rate. Conclusion: Multifractal features are more effective to describe the histologic texture. The proposed feature descriptor showed high classification rate for both liver and prostate data sample datasets.

Significance of Cytological Findings of Neuroblastomas: Rosette Arrangement and Neuropil Structure

Objectives: Improve treatment outcomes and clarify the biological characteristics of neuroblastoma, development of an international histological classification started several years ago. Aiming at the establishment of a cytology criteria corresponding to the new histological classification, we investigated a criteria comparing lesions related to neuroblastoma on referring to the morphological indices of neuroblastoma reported in the international classification. Methods: Several tumor specimens were investigated: 37 cases of neuroblastoma (undifferentiated type: 3, poorly differentiated type: 34), 3 cases of ganglioneuroblastoma (mixed type: 2, nodular type: 1), and one case of ganglioneuroma. Stamp cytology samples were prepared from cut surfaces of the tumors and then stained to the Papanicolaou method. Results: In neuroblastoma of the undifferentiated type, tumor cells contained a small oval nucleus with a high N/C ratio, showing a bare nucleus, and the nucleolus was distinct: no rosette formation or neutrophil was observed. In the poorly differentiated type, tumor cells showed a round-oval bare nucleus were scattered: rosette arrangement was observed in the background neuroblasts containing a bare nucleus. In ganglioneuroblastoma, immature neuroblasts showed a round-oval nucleus and large ganglion-like cells possessed a distinct nucleolus similar to poorly differentiated-type epithelial adenocarcinoma. Conclusion: In neuroblastoma, neutrophils were stained light green, and a partial Homer-Wright-type rosette arrangement was observed in the background. In the poorly differentiated type, tumor cells were generally large compared to those observed in the undifferentiated type. In ganglioneuroblastoma, cytological diagnosis can be relatively easily made when differentiated mature ganglion like cells are observed. In the case of surgery, a histological diagnosis of nervous system tumors is often performed using frozen sections, however tissue is usually damaged during freezing. Thus, cytology is more advantageous for diagnosis. The diagnostic accuracy can be improved utilizing the cytological characteristics of neuroblastic tumors.

MULTIFRACTAL COMPUTATION FOR NUCLEAR CLASSIFICATION AND HEPATOCELLULAR CARCINOMA GRADING

Hepatocellular carcinoma (HCC) is graded mainly based on the characteristics of liver cell nuclei. This paper pro- poses a textural feature descriptor and a novel computa- tional method for classifying liver cell nuclei and grading the HCC histological images. The proposed textural fea- ture descriptor observes local and spatial characteristic s of the texture patterns by using multifractal computation. The textural features are utilized for nuclear segmentation, fi ber region detection, and liver cell nuclei classification. Fou r categories of nuclear features are computed such as texture, geometry, spatial distribution, and surrounding texture, for HCC classification. Significance of liver cell nuclei classi - fication method is evaluated by classifying non-neoplastic and tumor tissues. Furthermore, characteristics of the liv er cell nuclei were utilized for grading a set of HCC images into four classes and obtained 97.77% classification accu- racy.

Multifractal feature descriptor for diagnosing liver and prostate cancers in H&E stained histologic images

Histologic imaging plays an important role in discriminating cancerous tissues of several body organs. However, the human histopathological examinations may be subjective and error prone, because of the complexity of the appearances of the histologic texture. These limitations can be overcome by adopting quantitative computational methods with human histopathological examination routines. This study proposes a new feature descriptor to characterize texture of histologic images. The proposed method derives a discriminative feature space by observing the self-similarity characteristics of the texture based on fractal geometry. The merit of utilizing fractal geometry to describe the histologic texture is assessed by a classification experiment. The experimental results indicate that the proposed feature descriptor can classify cancer and non-cancer tissues of histologic images of liver and prostate images around 95% of correct classification rate.