Automatic classification of brain magnetic resonance images with hypercolumn deep features and machine learning

Abstract

Brain tumours are life-threatening and their early detection is very important in a patient's life. At the present time, magnetic resonance imaging is one of the methods used for detecting brain tumours. Expert decision support systems serve specialist physicians to make more accurate diagnoses by minimizing the errors arising from their subjective opinions in real clinical settings. The model proposed in this study detects important keypoints and then extracts hypercolumn deep features of these keypoints from some convolutional layers of VGG16. Finally, Random Forest and Logistic Regression classi ers are fed with a set of these features. Random Forest classi er offered the best performance with 94.84% accuracy, 95.48% sensitivity, 4.52% false negative rate, 94.19% speci city, and 94.35% precision in 5-fold crossvalidation in this study. Consequently, it is thought that the proposed model could contribute to eld experts by integrating it into computer-aided brain magnetic resonance imaging diagnosis systems.

Enhancing X-ray-Based Wrist Fracture Diagnosis Using HyperColumn-Convolutional Block Attention Module

“…As depicted in Figure 4 , the HyperColumn technique upsamples the feature maps of the downsampled layers and stacks them to form a HyperColumn feature map. This map retains both low-level and high-level information, where lower-level information conveys fine-grained detail, and higher-level information encapsulates more abstract global features, offering the advantage of capturing both intricate details and broader patterns [ 40 , 41 ].…”

“…As d Figure 4, the HyperColumn technique upsamples the feature maps of the dow layers and stacks them to form a HyperColumn feature map. This map retains level and high-level information, where lower-level information conveys fine-g tail, and higher-level information encapsulates more abstract global features, o advantage of capturing both intricate details and broader patterns [40,41]. Nonetheless, there is a risk that less useful features may overwhelm the m ones for classification when features extracted at various resolutions are comb out consideration of their relative importance.…”