A Multimode Microfiber Specklegram Biosensor for Measurement of CEACAM5 through AI Diagnosis

Area of Science:

• Biomedical Engineering
• Optoelectronics
• Artificial Intelligence

Background:

• Carcinoembryonic antigen (CEACAM5) is a critical tumor biomarker for cancer monitoring.
• Accurate detection of CEACAM5 is essential for assessing therapeutic efficacy and disease progression.

Purpose of the Study:

• To develop a novel biosensor for sensitive and accurate CEACAM5 detection.
• To utilize fiber specklegrams and 2D-CNNs for enhanced biomarker analysis.

Main Methods:

• A tapered multimode fiber (MMF) biosensor modified with CEA antibodies was fabricated.
• Specklegrams generated by the MMF were analyzed using a zero mean normalized cross-correlation (ZNCC) function.
• A 2D-CNN was employed to process specklegram data and improve detection accuracy.

Main Results:

• The biosensor demonstrated high sensitivity to CEACAM5 concentrations from 1 to 1000 ng/mL.
• A surface sensitivity of 0.0012 (ng/mL)^-1 was achieved within the 1-50 ng/mL range.
• The 2D-CNN model achieved a maximum detection error of 0.358% for CEACAM5 monitoring.

Conclusions:

• The proposed fiber specklegram biosensing scheme provides a sensitive and accurate method for CEACAM5 detection.
• This approach has significant potential for postoperative patient monitoring and cancer management.