Application of surface-enhanced Raman scattering combined with artificial intelligence for multiplex detection of...

Chemotherapy remains a cornerstone in the treatment of multiple solid and gynecologic malignancies, including endometrial cancer, ovarian cancer, and breast cancer. Carboplatin, epirubicin, gemcitabine, paclitaxel, topotecan, and docetaxel are widely used in combination regimens; however, there is still a lack of sensitive, rapid, and multiplex technologies for real-time monitoring of chemotherapeutic drug concentrations. In this study, we developed a carboxyl-polyethylene glycol (PEG)-modified magnetite (Fe₃O₄@PEG) nanoplatform combined with artificial intelligence (AI)-driven spectral analysis to achieve ultra-sensitive and multiplex detection of six chemotherapeutic drugs. Fe₃O₄ nanoparticles were synthesized by co-precipitation and subsequently functionalized with carboxyl-PEG to enhance drug capture efficiency and resistance to matrix interference. Using a confocal Raman imaging system, SERS spectra of single and mixed drug systems were collected, and convolutional neural networks (CNNs) were employed for automated feature extraction and classification to enable qualitative and quantitative analysis. The constructed SERS platform achieved 10⁻⁷-10⁻⁸ milligrams per milliliter (mg/mL) detection limits for all six drugs in serum matrices, with classification accuracies exceeding 95% in multiplex mixtures. For improved quantitative accuracy and normalization, the 2078 cm⁻¹ Raman peak of deuterated methanol (CD₃OD) was employed as an internal standard to correct for potential fluctuations in laser intensity and sample concentration. Within the simulated therapeutic concentration range, quantitative regression yielded R² ≥ 0.98. This AI-SERS platform offers advantages of multiplex detection, trace-level sensitivity, and strong anti-matrix interference, holding great promise for pharmacokinetic monitoring and individualized chemotherapy optimization across a broad range of solid tumors.