Associate researcher  Artificial Intelligence (AI) Core Laboratory of Chang Gung Hospital •   September 2019 - Present

1. Deep learning module establishment, optimization and verification
Using the Python programming language with Tensorflow and Keras to construct different deep learning networks by setting parameters, then GPU computing is used to build prediction model. Classification of the input samples through the prediction model and output the heatmap of results to verify whether the model has learned areas of image features.
2. Programming and Docker operation
First, modifying and writing python scripts as well as go language (code of existing web pages) to create two web pages on the server. Then, the connection between the Docker prediction model software and the web pages are used to establish two web pages data distribution services, and the input of the web pages will be connected to the Docker prediction model software. Finally, the heatmap and prediction results will be displayed on the web pages, showing the areas where the prediction model software has a high probability of identifying image features.
3. Server workstation establishment and maintenance
Management of related softwares and drivers, as well as all sudo and general users permissions, such as Nvidia-GPU graphics card drivers and matching Cuda versions.
4. mongoDB establishment and maintenance, medical image collection and management
Using regular expression to search keywords to batch anonymize the applied medical images, then mongoDB will connect medical information with anonymized information to obtain the required information.