Table 1:
Preparation of sample solutions
| No. | Protein (g) | Water (mL) | Output strip |
|---|---|---|---|
| 1. | 0.00 | 20 | Negative (−) |
| 2. | 1.00 | 20 | Plus-minus (+−) |
| 3. | 3.00 | 20 | Positive 1 (+) |
| 4. | 5.00 | 20 | Positive 2 (++) |
| 5. | 7.30 | 20 | Positive 3 (+++) |
| 6. | 11.60 | 20 | Positive 4 (++++) |
Figure 1:
Protein detection computer program. KNN, K-nearest neighbors.
Figure 2:
Evaluation of the KNN model. KNN, K-nearest neighbors.
Figure 3:
(A) Image of the prototype seen from the outside, (B) prototype components seen from the inside, and (C) shape of the prototype that is ready to be used.
Figure 4:
Distribution of 30 test data.
Table 2:
Training and test data
| Label/class/category | Amount of data | Information protein content (g/L) |
|---|---|---|
| − | 6 | 0 |
| +− | 24 | 0.15 |
| + | 10 | 0.3 |
| ++ | 22 | 1 |
| +++ | 30 | 3 |
| ++++ | 7 | 20 |
Table 3:
Evaluation of the KNN model
| K value | Accuracy (%) | Precision (%) | Recall (%) | F1 score (%) |
|---|---|---|---|---|
| 3 | 96.7 | 97.0 | 96.7 | 96.2 |
| 10 | 86.7 | 75.8 | 86.7 | 80.7 |
| 20 | 76.7 | 60.9 | 76.7 | 67.3 |
Figure 5:
Results of confusion matrix values at K = 3.
Figure 6:
Results of confusion matrix values at K = 10.
Figure 7:
Results of confusion matrix values at K = 20.
Table 4:
Comparison of research results
| No. | Biomarker | Author and year | Color classification | Work principle | Ref. |
|---|---|---|---|---|---|
| 1. | Albumin | Thakur (2021) | RGB, HSV, and Lab | RF algorithm to estimate albumin concentration using a smartphone | [32] |
| 2. | Albumin | Thakur (2022) | RGB, HSV, and Lab | CNN algorithm for classifying Color in detecting albumin using a smartphone. | [41] |
| 3. | Albumin | Kim (2022) | RGB | RGB extraction uses machine learning and iPhone 11 as a means of detecting color in urine. | [42] |
| 4. | Protein | This study (2023) | RGB | Protein detection equipped with a digital color sensor type ELP camera. Image data are classified based on RGB and evaluated using the KNN algorithm |