Single-frame vs 30-frame Absorbance Comparison

Dataset: 2026-06-26 TRITC on-chip urea assay (TRITC bandpass filter ~530–590 nm, 4× objective + 1.5× zoom, 0.987 µm/px). Cropped images: 352 × 1686 px. Green-channel absorbance A = −log₁₀(T).

This report compares absorbance results computed from a single frame (middle frame of 30) versus a 30-frame average for each concentration, dark image, and flat-field image. The goal is to determine whether the 30-frame average provides a meaningful improvement over using a single frame.

1. Calibration Curves

Solid circles and lines: 30-frame average. Dashed lines and triangles: single frame. One curve per channel section (S1–S8); deeper sections have steeper slopes.

Figure 1. Calibration curves for both methods. Fits are nearly identical; R² values agree to ≥ 3 decimal places across all sections.

2. Fit Parameter Comparison

Section	Depth (µm)	Equation (30f)	R² (30f)	Equation (1f)	R² (1f)	Slope Δ%
S1	200	A = 0.00461·c + 0.1137	0.99756	A = 0.00462·c + 0.1138	0.99752	0.27
S2	150	A = 0.00398·c + 0.0919	0.99562	A = 0.00398·c + 0.0918	0.99550	0.02
S3	120	A = 0.00348·c + 0.0757	0.99560	A = 0.00349·c + 0.0757	0.99563	0.11
S4	90	A = 0.00290·c + 0.0589	0.99714	A = 0.00290·c + 0.0589	0.99715	0.06
S5	70	A = 0.00250·c + 0.0477	0.99600	A = 0.00250·c + 0.0477	0.99590	0.20
S6	50	A = 0.00191·c + 0.0334	0.99644	A = 0.00192·c + 0.0333	0.99592	0.24
S7	40	A = 0.00156·c + 0.0270	0.99707	A = 0.00155·c + 0.0270	0.99709	-0.16
S8	30	A = 0.00119·c + 0.0208	0.99670	A = 0.00119·c + 0.0208	0.99685	0.05

Table 1. Linear fit equations and R² for both methods. Slope differences are ≤ 0.3% in all sections.

3. ROI Mean Absorbance Difference (single − 30-frame)

Section	S1	S2	S3	S4	S5	S6	S7	S8
Concentration (mM)
0.0	0.00007	-0.00011	0.00008	0.00002	-0.00004	0.00003	-0.00007	0.00007
0.5	0.00013	0.00004	-0.00000	-0.00003	-0.00004	-0.00003	0.00000	-0.00004
1.0	-0.00007	-0.00006	0.00002	-0.00001	-0.00002	-0.00015	0.00007	-0.00005
2.0	0.00005	-0.00007	-0.00001	-0.00005	0.00000	-0.00007	0.00000	0.00004
5.0	0.00009	-0.00008	0.00012	0.00001	0.00008	-0.00011	-0.00000	-0.00010
10.0	0.00002	0.00001	0.00004	0.00006	0.00010	0.00016	0.00001	-0.00006
15.0	0.00035	0.00002	0.00008	-0.00002	0.00001	0.00007	-0.00007	0.00004
20.0	0.00025	-0.00009	0.00011	0.00003	0.00008	-0.00005	-0.00003	0.00001

Table 2. Absolute difference in mean ROI absorbance (AU). Maximum difference across all sections and concentrations: 0.00035 AU.

4. Pixel-level Noise Comparison

Figure 2. Ratio of pixel-level standard deviation (single frame / 30-frame average) within each section ROI, averaged across all concentrations. The red dashed line shows the theoretical √30 ≈ 5.48× expected for Poisson-limited noise. Observed ratios of 3–4.5× indicate the camera noise is sub-Poisson or partially correlated.

Section	Std dev (30f)	Std dev (1f)	Noise ratio (1f/30f)
S1	0.0028	0.0097	3.3929
S2	0.0023	0.0069	3.0361
S3	0.0021	0.0068	3.2799
S4	0.0018	0.0070	3.9025
S5	0.0024	0.0072	3.0762
S6	0.0017	0.0072	4.2873
S7	0.0016	0.0073	4.5537
S8	0.0018	0.0074	4.2452

Table 3. Mean pixel standard deviation and noise ratio per section.

5. Conclusions

Calibration accuracy: single-frame and 30-frame fit slopes agree within 0.3%; R² values are essentially identical (differ by < 0.001).
ROI mean absorbance: maximum difference is 0.00035 AU — well within the noise floor of either method.
Pixel-level noise: the 30-frame average reduces pixel noise by 3–4.5×, but this does not affect section-mean values meaningfully.
Recommendation: a single corrected frame per concentration is sufficient for the urea assay at this ROI size. The 30-frame average provides no practical benefit for calibration or concentration determination.