Table 2.
Performance of the fully convolutional neural network for NFT detection
| Training/Validation | Testing | ||||
|---|---|---|---|---|---|
| Metrics | Total | Total | PART | CTE | PSP |
| Recall, TP/(TP + FN), sensitivity | 0.91 | 0.92 | 0.96 | 0.78 | 0.88 |
| Precision, TP/(TP + FP), PPV | 0.80 | 0.72 | 0.77 | 0.75 | 0.63 |
| F1 score (harmonic mean of precision and recall) | 0.85 | 0.81 | 0.85 | 0.77 | 0.73 |
F1 score harmonic mean is calculated as 2*Precision*Recall/(Precision + Recall); NFT neurofibrillary tangle; PART primary age-related tauopathy; CTE chronic traumatic encephalopathy; PSP progressive supranuclear palsy