With the rise in computing power, deep-learning based computer-aided diagnosis systems have gained interest in medical research community. Our advanced AI system processes the images to assist doctors in order to identify whether the patients have nodules in lungs. Meanwhile, we utilized the weakly supervised learning based patching network to extend the receptive field on the convolutional kernel, which improved the performance on the small nodule detection with various locations in CXR. The weakly supervised learning mechanism also achieves the way of soft-annotation to reduce physician effort in medical image annotation.

"With radiologist’manually labeled images as target examples for training, we developed the novel patching model to identify potential hot spot areas of nodules. Meanwhile, we utilized a novel weakly supervised learning framework to extend the receptive field on the convolutional kernel. The final result was accepted by the international conference of MIDL2022 (July 5-7) to present results in Switzerland.
 
 Our team was chosen as the Top 10 Global Outstanding Projects by the UNESCO Scientific Program Committees in 2022, the only team from Asia country."

"Lung cancer is a major cause of death in Taiwanworldwide. As early detection can improve outcome, regular screening is of great interest. Regular X-ray screening with smart medical imaging methods is beneficial. 
 
 Deep learning-based computer aided CXR diagnosis systems have proven their ability of identifying nodules in radiographiesthus may assist physicians in clinical practice. 
 
 The aim of this research is to develop an AI detector for the task of pulmonary nodule detection for further integrate as an industrial level of medical device. In summary, the presented AI method has great potential to be smart medical deviceto help physicians during clinical routine."