Dense Convolutional Neural Network for Identification of Raman Spectra

Deep Learning-Assisted Intelligent Artificial Vision Platform Based on Dual-Luminescence Eu(III)-Functionalized HOF for the Diagnosis of Breast and Ovarian Cancer

Developing an advanced analytical method to detect spermine (Spm) and N-acetylneuraminic acid (NANA), the biomarkers of breast and ovarian cancers, respectively, is critical for the early diagnosis of the two cancers, which is very meaningful for women's health. Here, a deep learning-assisted artificial vision platform based on a dual-emission ratiometric fluorescence sensor is first constructed to monitor Spm and NANA. The ratiometric fluorescence sensor (Eu@TCBP-HOF, 1) can selectively detect Spm with high sensitivity based on "Turn-on" mode. After adding Spm, the new ratiometric fluorescence sensor (1-Spm, named 2) shows high sensitivity for NANA with "Turn-off" mode. Moreover, the fluorescence sensors can achieve an obvious fluorescence color response to Spm and NANA. Even in real saliva and serum samples, 1 and 2 still show high sensitivity and color responsiveness with limit of detection (LODs) of 0.5 μM for Spm and 0.96 μM for NANA. In virtue of different fluorescence responses, the DenseNet algorithm of deep learning assists the fluorescence sensors, which can simulate the human visual systems to identify fluorescence images and distinguish the concentration of Spm and NANA within 1 s with over 99% recognition accuracy. The intelligent artificial vision platform developed in this work may provide a prospective analytical method for the early diagnosis of female malignant tumors.