Label-Free Prediction of EGFR Mutation Status Using Fluorescence Lifetime Imaging and Deep Learning in Lung...

Area of Science:

• Computational pathology
• Biophotonics
• Molecular diagnostics

Background:

• Accurate epidermal growth factor receptor (EGFR) mutation prediction is crucial for non-small cell lung cancer (NSCLC) targeted therapy.
• Current molecular techniques (PCR, NGS) are costly, time-consuming, destructive, and tissue-intensive.
• There is a clinical need for noninvasive, rapid, and tissue-sparing methods for EGFR mutation identification.

Purpose of the Study:

• To develop a deep learning (DL)-based, label-free approach for predicting EGFR mutations using fluorescence lifetime imaging (FLIM).
• To assess the accuracy and efficiency of this nondestructive method compared to conventional techniques.
• To enable faster and more accessible EGFR mutation prediction for NSCLC patients.

Main Methods:

• Utilized deep learning (DL) algorithms applied to label-free fluorescence lifetime imaging (FLIM) data.
• Developed a nondestructive imaging technique that eliminates the need for tissue staining.
• Validated the method on formalin-fixed, unstained lung adenocarcinoma tissue samples.

Main Results:

• Achieved a state-of-the-art area under the receiver operating characteristic curve (AUC) of 0.966 for EGFR mutation prediction.
• Accurately classified common EGFR mutations (exon 19 deletion, exon 21 point mutation) in NSCLC.
• Demonstrated the potential to significantly reduce clinical overhead from traditional sequencing and staining.

Conclusions:

• Label-free FLIM combined with DL offers a rapid, accurate, and nondestructive method for EGFR mutation prediction.
• This approach can overcome limitations of conventional molecular testing in NSCLC diagnostics.
• The developed strategy holds promise for accelerating targeted therapy selection and improving patient outcomes.