- Multi-scale and multi-directional feature extraction using LDH
- Pattern classification via MLP
- Sagittal planes of 3D US volumes collected with motorized transducer from 13 volunteers
97% and 94% accuracy on train and test data, respectively. For more details, please refer to the paper.
- docker
- numpy
- opencv
- No GPU!
sh build.shsh run.sh [data folder] [data file]
ex: sh run.sh /home/deepinfer/image_data img1.jpg
Single image with arbitrary size and .jpg .png .bmp formats OR numpy array. The numpy array can be 2D (I, J) (one image) or 3D (N, I, J) (multiple images) where N indicates the image index.
Numpy array of shape (1, 2) saved as output.npy in [data folder] in the format of [P of non-target, P of target]. In the case of 3D input, the output is in the form of (N, 1, 2) where N is the number of images in the 3D volume.
If you use this code and docker image, please cite:
- Pesteie, M., Abolmaesumi, P., Ashab, H. A. D., Lessoway, V. A., Massey, S., Gunka, V., & Rohling, R. N. (2015). Real-time ultrasound image classification for spine anesthesia using local directional Hadamard features. International journal of computer assisted radiology and surgery, 10(6), 901-912.
- Mehrtash, A., Pesteie, M., Hetherington, J., Behringer, P. A., Kapur, T., Wells, W. M., … & Abolmaesumi, P. (2017, March). DeepInfer: open-source deep learning deployment toolkit for image-guided therapy. In SPIE Medical Imaging (pp. 101351K-101351K). International Society for Optics and Photonics.