Skip to main content
. Author manuscript; available in PMC: 2023 Apr 21.
Published in final edited form as: Pattern Recognit. 2022 Jul 22;132:108919. doi: 10.1016/j.patcog.2022.108919

Table 4.

Performance of the loose model by 1) using four priors but without capturing spatiotemporal relationships and 2) by directly using the differences of sequential priors.

Model MLO+CC view AUC
LRP-NET (prior-4 → prior-3 → prior-2 → prior-1) 0.67 (referent)
Loose model using four priors (prior-4, 3, 2, 1) 0.63 (p = 0.03)
Loose model using differences of sequential priors (| prior-3 - prior-4 |, |prior-2 - prior-3 |, |prior-1 - prior-2 |) 0.57 (p < 0.01)