Retrieval of Brain Tumors by Adaptive Spatial Pooling and Fisher Vector Representation

Abstract

Content-based image retrieval (CBIR) techniques have currently gained increasing popularity in the medical field because they can use numerous and valuable archived images to support clinical decisions. In this paper, we concentrate on developing a CBIR system for retrieving brain tumors in T1-weighted contrast-enhanced MRI images. Specifically, when the user roughly outlines the tumor region of a query image, brain tumor images in the database of the same pathological type are expected to be returned. We propose a novel feature extraction framework to improve the retrieval performance. The proposed framework consists of three steps. First, we augment the tumor region and use the augmented tumor region as the region of interest to incorporate informative contextual information. Second, the augmented tumor region is split into subregions by an adaptive spatial division method based on intensity orders; within each subregion, we extract raw image patches as local features. Third, we apply the Fisher kernel framework to aggregate the local features of each subregion into a respective single vector representation and concatenate these per-subregion vector representations to obtain an image-level signature. After feature extraction, a closed-form metric learning algorithm is applied to measure the similarity between the query image and database images. Extensive experiments are conducted on a large dataset of 3604 images with three types of brain tumors, namely, meningiomas, gliomas, and pituitary tumors. The mean average precision can reach 94.68%. Experimental results demonstrate the power of the proposed algorithm against some related state-of-the-art methods on the same dataset.

Fully automatic multisegmentation approach for magnetic resonance imaging brain tumor detection using improved region‐growing and quasi‐Monte Carlo‐expectation maximization algorithm

“…The values of dice and sensitivity are close to those reported by Vaidhya et al 14 and Pereira et al 15 and higher than those reported by Demirhan et al 16 Moreover, our technique does not need a training process, unlike the technique developed by Ellwaa, 17 and it assures much lower computational time compared with the method of Pereira et al 15 Finally, we compared our approach with works reported by Zhang and Wu 18 and Dong et al 19 The MRI images used in this stage need a skull isolation process. 20 Table 3 represents a comparison performance in the case of meningioma and glioma images, where results showed that, in the case of meningioma image, the dice and sensitivity were, respectively, 1% and 3% lower than the values reported in Reference 18. In the case of glioma image, the dice and sensitivity were, respectively, 8% and 9% higher than the values reported in both References 18 and 19.…”

Brain Tumor Type Classification via Capsule Networks

“…To test our proposed approach, we have used the data set presented in References [8,21]. This data set contains 3, 064 MRI images of 233 patients diagnosed with one of the aforementioned three brain tumor types.…”

Brain Diagnoses Detection Using Whale Optimization Algorithm Based on Ensemble Learning Classifier

“…The image dataset used in this work is collected from [36]. The total number of images is 3064 with representations from three categories (meningioma, glioma and pituitary).…”