Deep-Learning-Assisted Discriminative Detection of Vitamin B12 and Vitamin B9 by Fluorescent MoSe2 Quantum Dots

A facile and environmentally mindful approach for the synthesis of MoSe₂ QDs was developed via the hydrothermal method from bulk MoSe₂. In this, the exfoliation of MoSe₂ was enhanced with the aid of an intercalation agent (KOH), which could reduce the exfoliation time and increase the exfoliation efficiency to form MoSe₂ QDs. We found that MoSe₂ QDs display blue emission that is suitable for different applications. This fluorescence property of MoSe₂ QDs was harnessed to fabricate a dual-modal sensor for the detection of both vitamin B₁₂ (VB₁₂) and vitamin B₉ (VB₉), employing fluorescence quenching. We performed a detailed study on the fluorescence quenching mechanism of both analytes. The predominant quenching mechanism for VB₁₂ is via Förster resonance energy transfer. In contrast, the recognition of VB₉ primarily relies on the inner filter effect. We applied an emerging and captivating approach to pattern recognition, the deep-learning method, which enables machines to "learn" patterns through training, eliminating the need for explicit programming of recognition methods. This attribute endows deep-learning with immense potential in the realm of sensing data analysis. Here, analyzing the array-based sensing data, the deep-learning technique, "convolution neural networks", has achieved 93% accuracy in determining the contribution of VB₁₂ and VB₉.