Deep learning based Image Processing
Objective
The application of deep learning is increasing day by day, especially in the field of Biomedical engineering and image processing. AI can be used in different areas, including detecting any region of biomedical interest including cells or tissue, segmenting out those areas, determining the severity of diseases, tracking any moving cells like red blood cells, or enhancing any raw biomedical image. In our Neurophotonics Lab, we leverage the power of deep learning to find out pathological lesions and automate the processes of detection, diagnosis, and biomedical decision making.
1. Biomedical Image Segmentation
In the case of biomedical image segmentation, we collect data from ophthalmologists, prepare a dataset by labeling them using leveling tools. We train SOTA neural networks with those images with labeled data to automate glands segmentation. Other applications include finding the shape of red blood cells and pointing coordinates.
2. Biomedical Image classification
Similarly, those images with grades assigned to them are trained to predict the assessment score of those glands images. Besides, we use knowledge transfer technology to improve the accuracy of the network. Other applications include classifying healthy vs. affected tissue, cancer cell detection, and finding the density of abnormal cells.
3. Image enhancement with GAN
Generative Adversarial Network is another advanced technology used to remove reflected areas from images. Other usage includes making image super-resolution(SR), speed up confocal line scan by SR based sparse scan, deblurring image.
