1：Experimental Design and Method Selection:

The study used ex-vivo fluorescence lifetime imaging (FLI) with a 375 nm pulsed laser for excitation. Data analysis involved double exponential fitting of fluorescence decays, principal component analysis (PCA) for dimensionality reduction, and support vector machine (SVM) for classification.

2：Sample Selection and Data Sources:

Fresh human cervical tissue samples were obtained from GSVM Medical College, Kanpur, India, including 10 EH samples (25 sites), 11 CIN-I samples (20 sites), and 1 CIN-II sample (2 sites). Samples were stored in ice, thawed to room temperature, and rinsed with saline before experiments within 4 hours of biopsy.

3：List of Experimental Equipment and Materials:

PicoQuant picosecond pulsed diode laser (375 nm wavelength, pulse width 48 ps, repetition rate 40 MHz, average power

4：5 mW), LaVision ICCD camera, PDL 800-B driver, achromatic lens (Thorlab f = 75 mm), 450 nm long pass filter, Nikon AF Nikkor 50 mm f/8 D camera lens, high rate imager (HRI), high rate delay generator (HDG), and 'Davis' user interface software. Experimental Procedures and Operational Workflow:

Tissue samples were excited with the laser, and fluorescence signals were collected through the filter and imaged onto the ICCD. FLI were captured at delay steps of 100 ps over

5：7 nanoseconds with a gate width of 200 ps and CCD acquisition time of 3000 ms per step. Data Analysis Methods:

Fluorescence decays were fitted to a double exponential function using MATLAB's 'fit' function. PCA was applied to average decay profiles using correlation matrix and singular value decomposition in MATLAB. SVM classification used holdout cross-validation with training and validation sets, employing radial basis function (RBF) kernel for non-linear classification.