Automatic Histogram Specification for Glioma Grading Using Multicenter Data

Abstract

Multicenter sharing is an effective method to increase the data size for glioma research, but the data inconsistency among different institutions hindered the efficiency. This paper proposes a histogram specification with automatic selection of reference frames for magnetic resonance images to alleviate this problem (HSASR). The selection of reference frames is automatically performed by an optimized grid search strategy with coarse and fine search. The search range is firstly narrowed by coarse search of intraglioma samples, and then the suitable reference frame in histogram is selected by fine search within the sample selected by coarse search. Validation experiments are conducted on two datasets GliomaHPPH2018 and BraTS2017 to perform glioma grading. The results demonstrate the high performance of the proposed method. On the mixed dataset, the average AUC, accuracy, sensitivity, and specificity are 0.9786, 94.13%, 94.64%, and 93.00%, respectively. It is about 15% higher on all indicators compared with those without HSASR and has a slight advantage over the result of a manually selected reference frame by radiologists. Results show that our methods can effectively alleviate multicenter data inconsistencies and lift the performance of the prediction model.

The Automatic Brain Tumor Segmentation Based on MRI Using Optimal Morphology Thresholding Methods

“…The accuracy was 98%. By comparing it to the other methods, it was found that its accuracy was higher than that of the histogram [4], Morphology [18], Neuro Fuzzy Classifier [10] and Fuzzy C-means [8], respectively.…”

“…Primarily, the brain tumors are medically diagnosed. Thus, the algorithms for the automatic segmentation of the images applied to computerized tomography (CT) scan [1], positron emission tomography (PET) scan [2,3] and magnetic resonance imaging (MRI) [4][5][6][7][8][9][10][11][12] for segmenting the images of the brain tumors from the images of CT, PET and MRI, respectively. There are preprocessing and post processing developments.…”

“…The threshold level based on white matters and gray matters is optimal in the case of the segmentation of the brain tumors from the MRI images with different brightness levels only, while the morphological watershed segmentation uses the MRI images of brains with similar brightness levels. Both methods required the MRI images without interferences, quite clear tumors and the color segmentation algorithms with Kmeans in order to convert the colored MRI images into grayscale images and to segment the tumors with the histogram-clustering method that can clearly differentiate the tumors [4]. Then, the color segmentation algorithms with Kmeans and the histogram-clustering method were applied to the MRI images of the 20 patients with the four types of tumors by considering and comparing the segmentation results of the anatomy images [14].…”

HS–GS: A Method for Multicenter MR Image Standardization

“…Experiment 4: We add the comparison method of experiment (3). Radiologist selection and use the classic histogram specification [19], and we also compare the performance of deep learning for glioma grading [38], [39].…”

“…HS can also be used for image standardization. First, the clinician manually selects a set of representative images from the data set and constructs a transfer function based on this set of representative images [19]. All data sets are subjected to histogram matching according to the transfer function, and the images have a highly similar histogram distribution.…”