Table 1 Overview of the performance of our model, the pathologists, and previous models.
Model / Pathologist | TP | FP | FN | TN | Sensitivity | Specificity | MCC |
---|---|---|---|---|---|---|---|
Our model | 263 | 12 | 32 | 849 | 0.8915(0.8503–0.9246) | 0.9861(0.9758–0.9928) | 0.8986(0.8686–0.9269) |
Pathologist S.-C.H. | 289 | 2 | 6 | 859 | 0.9797(0.9563–0.9925) | 0.9977(0.9916–0.9997) | 0.9818(0.9681–0.9932) |
Pathologist 1 | 279 | 2 | 16 | 859 | 0.9458(0.9134–0.9687) | 0.9977(0.9916–0.9997) | 0.9589(0.9392–0.9767) |
Pathologist 2 | 286 | 11 | 9 | 850 | 0.9695(0.9429–0.9860) | 0.9872(0.9773–0.9936) | 0.9546(0.9340–0.9731) |
Pathologist 3 | 275 | 2 | 20 | 859 | 0.9322(0.8972–0.9581) | 0.9977(0.9916–0.9997) | 0.9497(0.9284–0.9690) |
Pathologist 1with partial AI assistance | 289 | 2 | 6 | 859 | 0.9797(0.9563–0.9925) | 0.9977(0.9916–0.9997) | 0.9818(0.9682–0.9932) |
Pathologist 2with partial AI assistance | 291 | 2 | 4 | 859 | 0.9864(0.9656–0.9963) | 0.9977(0.9916–0.9997) | 0.9863(0.9742–0.9956) |
Pathologist 3with partial AI assistance | 288 | 2 | 7 | 859 | 0.9763(0.9517–0.9904) | 0.9977(0.9916–0.9997) | 0.9795(0.9648–0.9912) |
Hu et al. | 159 | 11 | 21 | 1025 | 0.8833(0.8272–0.9263) | 0.9894(0.9811–0.9947) | 0.8937(0.8566–0.9283) |
Wang et al. | 5217 | 391 | 82 | 9544 | 0.9845 (0.9808–0.9877) | 0.9606 (0.9566–0.9644) | 0.9334(0.9275–0.9391) |