Brain Cancer Detection Using a Fusion of Convolutional Neural Networks and Random Forests in Deep Vision Systems

This paper has achieved high levels of accuracy through the combination of CNNs and RF and been highly praised, especially for medical image recognition studies. This paper concentrates on how these methodologies can be combined to take the advantages of one in feature extraction and the other in classification. CNNs are well known for learning probably hierarchical features from image data while RFs on the other hand provide stable an robust classification. Thus, through introducing these aspects, it is possible to improve the efficiency and effectiveness of vision systems to a considerable extent. In this work, several types of fusion approaches are discussed in detail and the comparative investigation of them is carried out to assess the efficiency of the approach in brain cancer detection based on MRI data. Combining analysis and semantically bottom-up and top-down approaches to image fusion we discuss advantages and disadvantages of various approaches in the aspects of classification, identification and segmentation. By performing numerous experiments and analyses on benchmarks, the authors prove that the applied CNN-RF fusion concept enhances the accuracy and robustness of the corresponding classification models and is highly scalable and easy to implement for complicated vision problems. We see our results indicating the promising nature of the presented hybrid approach as a way to improve modern deep vision systems most notably the framework of medical imaging.