Stereoscopic Projection of Cataract Surgery Video and Observer Perception of Intraocular Spatial Relationships

To evaluate whether stereoscopic projection of cataract surgery video changes ophthalmologists' perception of intraocular spatial relationships compared with nonstereoscopic projection, and whether such changes are scene-dependent.

Prospective, paired, randomized, observer-masked comparative study, conducted in July 2006. A stereoscopic system was assembled from a conventional surgical stereomicroscope by coupling 2 synchronized cameras to its assistant optical paths and reprojecting the images through orthogonal polarizing filters and matching eyewear. Three lens procedures, 2 phacoemulsifications and 1 anterior vitrectomy with scleral-fixated intraocular lens (IOL) implantation, were edited into 22 scenes (11 situations, each in stereoscopic and nonstereoscopic form) shown in a single randomized order to 63 ophthalmologists. Responses were recorded on visual analog scales or as binary answers and compared between modes and against a surgeon-defined perceptual reference; paired mean differences, 95% CIs, effect sizes, numerical P values and Holm-adjusted P values are reported.

All 11 paired comparisons were based on 63 observations. Five of 11 situations differed between modes, and all 5 survived Holm correction. Stereoscopic projection moved responses toward the reference in deep-chamber capsulorhexis (mean difference 15.3 mm, 95% CI 8.2 to 22.3; P = 0.00005), in the subluxated lens (P = 0.0008), and for the vitrector tip relative to the IOL, where correct answers rose from 23.8% to 49.2% (paired risk difference 25.4%, 95% CI 13.0 to 37.8; P = 0.0004). In shallow-chamber capsulorhexis and the collapsed posterior capsule, the nonstereoscopic mean lay closer to the reference. Stereopsis was identified in 85.2% of stereoscopic scenes.

Stereoscopic projection produced scene-dependent changes rather than uniform improvement, clearest when the judgment lay along the depth axis and monocular cues were degraded by transparency or overlap. These exploratory findings are specific to the scenes and system tested and do not establish an effect on learning or surgical performance.