Rapid and sensitive acute leukemia classification and diagnosis platform using deep learning-assisted SERS detection

Area of Science:

• Biomedical Diagnostics
• Computational Biology
• Spectroscopy

Background:

• Accurate and rapid diagnosis of acute leukemia (AL) is crucial for effective patient management.
• Current diagnostic methods may require significant time and sample volumes.
• Subtyping and genetic abnormality identification in AL are essential for treatment stratification.

Purpose of the Study:

• To develop a combined deep learning and surface-enhanced Raman scattering (DL-SERS) strategy for rapid and sensitive identification of acute leukemia.
• To evaluate the performance of the DL-SERS approach in classifying AL subtypes and genetic abnormalities.
• To explore the versatility of the DL-SERS platform for diagnosing other neurological diseases.

Main Methods:

• Collection and analysis of over 390 cerebrospinal fluid (CSF) samples, including healthy controls, AL patients, and other disease groups.
• Utilizing surface-enhanced Raman scattering (SERS) for sensitive detection within 5 minutes using minimal CSF volume (0.5 μL).
• Implementing an integrated feature fusion approach (1D spectra and 2D images) with a transformer model for classification.

Main Results:

• Achieved rapid and sensitive detection of acute leukemia using DL-SERS.
• Demonstrated exceptional classification performance in terms of accuracy, sensitivity, specificity, and reliability for AL diagnosis.
• Showcased the platform's versatility by extending its application to the classification of meningitis diseases.

Conclusions:

• The developed DL-SERS classification strategy provides a powerful tool for the rapid and sensitive identification of acute leukemia.
• This approach offers a potential auxiliary in vitro diagnostic tool for clinical settings.
• The DL-SERS platform exhibits significant potential for broader applications in disease diagnostics.