The hand held device captures dual-spectrum autofluorescent images and use devised image analysis methods for extracting features, which can best distinguish cancer from normal lesions. Embedded inside the system is also a deep learning model which can automatic detect the suspect regions from buccal areas.

1: The design of the dual spectrum autofluorescent image analyzer is based on cancer optical properties to select the spectrum images which can reveal the best distinguishing between tumor and normal tissues. 

2: Several features which demonstrated the best distinguish capability between tumor and normal tissues are selected. The doctor can analyze these features between the target image and the cohort data sets in the system, as one reference to the diagnosis.

3: A deep learning model which can perform automatic analysis of buccal areas to identify the suspect regions is designed inside the system. The deep learning model combine collaboration of detection branch and segmentation branch to to achieve a high accurate segmentation. Furthermore, data augmentation is also performed during the training, to improve its capability. Such a design reveals as a breakthrough in oral cancer analysis, The system is also revealed to have a high potential to be used for oral screening, and to reduce the misdetection rate. 

4: The system demonstrates as an integration with flexibilities, of the handheld image capture device, a control and display panel, and a cloud platform. The components can be used as a standalone portable system, It can be also connected to the cloud platform, for cohort data analysis to support the diagnosis.

鱗狀上皮細胞病變或附屬器官細胞惡性腫瘤檢驗等等相關檢測或輔助治療應用。