|
| Method name | Technique | Purpose of the application | Mechanism of the application |
|
| AnoGAN | Adversarial learning (generative adversarial neural network (GAN)) | This unsupervised learning method is suitable when the dataset is limited | This method performs anomaly detection by generating a large number of nonlesion images by GAN to detect images with lesions. |
| SG-CNN | Convolutional neural network (CNN) | This method can automatically produce ROI areas independently through a superlabel-guided CNN | This method can improve classification accuracy by generating fine-grained labels and superlabels of the region of interest in medical images whose lesions of interest are not well apparent. |
| U-Net | U-Net structure deep neural network | This method is suitable for the segmentation of lesions when they have abnormal shape and low contrast and are susceptible to transition during classification | The U-Net structure performs semantic segmentation of the osteolytic lesions on the input image by concatenating the convolutional layers in the encoder path with the deconvolutional layers in the decoder path. |
| Seg-UNet | Multilevel Seg-UNet | This method is suitable for segmentation of lesions of interest on the input image when the lesion has abnormal shape and low contrast and the size of the lesion is very small compared to the input image size | The Seg-UNet exploits U-Net structure as well as the global- and patch-based approach in order to improve the classification accuracy. |
| tRTA | Mathematical computation | This method is a manual image processing method for the segmentation of the lesions of interest from dataset that requires trained medical practitioners | The tRTA is a computerized radiographic texture analysis method for the evaluation of ROI through linear regression, BANN temporal analysis technique, and a LDA merging features technique. |
| Morphometry | Manual computation | The method employs the cross-intersect counting approach | In this method, with the use of a morphometric grid that is superimposed onto the region of interest on the radiographic images, computation is performed. |
| ImageJ | Manual image processing | General-purpose image processing software | ImageJ can take the advantage of different plugins and macros for various image processing goals. |
| Osteolytica | Manual image processing | Specifically designed for the measurement of lytic bone lesions | This image processing software is designed for 3D analysis of lesions and requires trained staff. |
| 3D Slicer | Medical image processing software | Manual image processing | 3D Slicer is medical software designed only for research purposes that can perform various image analyses using variety of packages on different anatomical positions. |
| ITK | Open-source medical library | Manual image processing | This medical library is suitable for developers for medical image processing purposes. |
| MITK | Open-source medical library | Manual image processing | This class medical library is based on the ITK library and provides segmentation and registration techniques. It also has a highly customizable workbench. |
|
