FIGURE 2.

Hidden Markov model for the underlying true pathological states (CIN categorization). We defined four states: normal (state 1), cervical intraepithelial neoplasia 1 (CIN1, state 2), CIN2 (state 3), and CIN3/cancer (state 4). The arrows in this figure specify possible transitions among these states; all transitions between adjacent states were allowed, except for the transition from CIN3/cancer to CIN2 (i.e., CIN3/cancer was the absorbing state). The transition parameter ${\lambda }_{\mathit{ij}}$ represents an instantaneous rate of transition from the true pathological state i to the true pathological state j. $e_{\mathit{rs}}$ denotes the probability of observing the state s conditionally on the true state r ($e_{11}$ = 1−$e_{12}$, $e_{22}$ = 1−$e_{21}$−$e_{23}$, $e_{33}$ = 1−$e_{32}-e_{34}$, $e_{44}$ = 1−$e_{43}$)