Hidden Markov Model-Based Behavioral Classification Reveals Visual Function Recovery After Retinal Organoid Transplantation in Mice

Abstract

Retinal degenerative diseases cause irreversible vision loss due to photoreceptor degeneration. Retinal organoid transplantation offers a promising strategy for restoring vision, but assessing functional recovery remains challenging. Standard visual function tests provide binary or coarse measures that do not fully capture how visual input influences natural behavior. Here, we applied a Hidden Markov Model (HMM)-based behavioral analysis to the visual cliff test to track locomotor state transitions in wild-type (WT) and rd1-2 J (RD) mice, evaluating depth perception and its recovery following photoreceptor transplantation. WT mice exhibited a strong cliff avoidance response, while RD mice showed no response, confirming the model's sensitivity to depth perception. Over repeated trials, WT mice rapidly habituated, shifting from three behavioral states (Resting, Exploring, and Navigating) to just two (Resting and Navigating). Transplanted RD mice began responding to the cliff at 2 weeks posttransplantation, coinciding with early synapse formation between grafted photoreceptors and host bipolar cells. The avoidance response became robust by 16 weeks but disappeared by 18 weeks, accompanied by state collapse, a hallmark of habituation never observed in untreated RD mice. These findings demonstrate that behavioral state-based analysis provides a sensitive and dynamic measure of functional vision recovery, capturing not only the emergence of depth perception but also its integration into adaptive behavior. This approach may help refine clinical evaluations of vision restoration therapies, bridging the gap between sensory recovery and real-world functional outcomes.