Esotropia, a convergent misalignment of the visual axes1, disrupts binocular integration during critical developmental periods, often leading to irreversible deficits in stereopsis and contrast sensitivity (CS) despite surgical realignment.1–4

The disruption of binocular vision is associated with decreased quality of life in patients with esotropia.5 The critical period for susceptibility of human stereopsis is defined as beginning soon after birth, peaking at 3.5 months, and followed by a continued gradual improvement until at least 4.6 years of age.6Re-alignment of the visual axes in esotropia contributes to the restoration of binocular vision and recovery of stereopsis, which depends on factors such as the timing of the esotropia onset, the duration of misalignment and unequal visual acuity.7,8 However, the reacquisition of high-grade binocularity and stereopsis are rare.9

While strabismus surgery corrects ocular alignment, residual deficits in high spatial frequency (SF) contrast sensitivity, a key determinant of fine-detail vision, and suboptimal stereopsis persist in up to 70 % of patients, compromising functional outcomes such as reading proficiency and depth perception.4–6 The neural basis for these deficits lies in maladaptive cortical plasticity: prolonged misalignment during sensitive developmental windows induces interocular suppression, degrading thalamocortical pathways responsible for high-SF processing and binocular summation.7–9 Traditional post-surgical therapies, including patching and prism adaptation, often fail to address these neuroadaptations, underscoring the need for interventions that directly target binocular plasticity.10,11 The treatment of binocular vision was underused over the past few decades.10–12 Using electronic equipment such as tablets, computers, or mobile phones, perceptual learning shows its advantages of persistent amelioration in psychophysical sensitivity, perceptual discrimination and detection capabilities following visual training.13–15

Perceptual learning (PL), a paradigm leveraging repetitive, task-specific visual training to drive neural reorganization, has emerged as a promising strategy to rehabilitate binocular function in strabismus.12–14 By exploiting Hebbian mechanisms of synaptic potentiation, PL enhances signal-to-noise ratios in visual cortices, promoting recovery of suppressed neural pathways.15,16 Recent advances in digital therapeutics have enabled home-based PL programs utilizing Gabor patches and dichoptic stimuli to selectively train spatial frequency channels and stereoacuity.17,18 While robust evidence supports PL efficacy in intermittent exotropia and amblyopia, demonstrating 0.2–0.3 logCS gains at 12–18 cycles per degree (cpd) and stereoacuity improvements of 60–100 arcseconds,16–19 its application to esotropia remains unexplored. This gap is critical, as esotropia’s distinct pathophysiology (e.g., stronger suppression scotomas, earlier onset) may demand tailored PL protocols.20,23

Current literature emphasizes the clinical relevance of high-SF contrast sensitivity (12–18 cpd), which correlates with reading speed and face recognition, skills pivotal to pediatric development.24,25 Post-surgical deficits at these frequencies, attributed to residual cortical suppression, are rarely addressed by conventional therapies.26 Furthermore, stereopsis recovery in esotropia remains suboptimal, with only 20–30 % of surgically aligned patients achieving functional stereoacuity (<200 arcseconds).27 Emerging neuroimaging data suggest that PL-induced improvements in amblyopia correlate with increased V1/V2 activation and interhemispheric coherence, providing a mechanistic framework to optimize training paradigms.28,29 However, no studies have yet translated these insights to esotropia or evaluated PL’s synergistic potential with surgical alignment.

This study investigates the hypothesis that home-based PL enhances medium and high-SF contrast sensitivity and stereopsis in pediatric esotropia patients following surgical realignment. By bridging the gap between neuroplasticity principles and clinical practice, this work advances a paradigm shift in post-surgical esotropia management, positioning PL as a tool to optimize functional vision restoration.

The cohort comprised 44 patients (22 male, 22 female) with a mean age of 4.34 ± 1.68 years at surgery. Esotropia subtypes included infantile (25 %, n = 11), partially accommodative (50 %, n = 22), nonaccommodative (20.45 %, n = 9), and consecutive esotropia (4.55 %, n = 2) (Table 1). Preoperative esodeviation averaged 45.57 ± 19.54 PD, with 52.27 % (n = 23) exhibiting superior oblique palsy (SOP) and 38.64 % (n = 17) demonstrating V-pattern strabismus. Surgical interventions predominantly involved bilateral medial rectus (MR) recessions (81.82 %, n = 36). Perceptual learning (PL) commenced at a median interval of 95 days postoperatively, with a mean training duration of 172.15 ± 86.03 days. The mean ± SD age at the initial point of perceptual learning was 5.39 ± 1.54 years. Compliance exceeded 80 % in 86.36 % (n = 38) of participants.

Successful alignment (≤10 PD) was maintained in 88.64 % (n = 39) at final follow-up. Five patients (11.36 %) developed consecutive exotropia (15–40 PD) at a median onset of 6.26 months post-PL, requiring reoperation in two cases. No adverse events related to PL were reported.

We set four inspection time points, preoperative (within one week before surgery), postoperative (initial point before training), 3 months after training, and final visit after training. After measurements were taken at five spatial frequencies, the mean contrast sensitivity values were recorded in Table 2, and the logCS curves were shown in Fig. 1. We can divide the five different spatial frequencies into three categories, low frequency (1.5 cpd, 3 cpd), medium frequency (6 cpd) and high frequency (12 cpd, 18cpd). As shown in the line chart, the logCS curves of 4 groups were overlapped at 1.5 cpd and 3 cpd (Fig. 1). We compared the logCS values between different groups at low spatial frequencies, none of these were statistically significant (P > 0.05). Medium SF contrast sensitivity (6 cpd) and high SF contrast sensitivity (12–18 cpd) demonstrated progressive improvement post-training (Table 1, Fig. 1A):

• •

  6 cpd: Baseline logCS improved from 2.02 ± 0.45 postoperatively to 2.15 ± 0.37 at final follow-up (P < 0.0001, Wilcoxon matched-pairs signed rank test; Cohen’s d = 0.32).

• •

  12 cpd: Baseline logCS improved from 1.66 ± 0.53 postoperatively to 1.89 ± 0.41 at final follow-up (P < 0.0001; Cohen’s d = 0.62).

• •

  18 cpd: logCS increased from 1.18 ± 0.53 postoperatively to 1.52 ± 0.41 (P < 0.0001; Cohen’s d = 0.78). Repeated-measures ANOVA confirmed a significant time × SF interaction (F[2, 86] = 9.24, P = 0.003), with post-hoc tests revealing improvements localized to 12–18 cpd (P < 0.01 after Bonferroni correction). Low SFs (1.5–3 cpd) showed no statistically significant changes (P > 0.05).

Fig. 1.

Line chart of contrast sensitivity at five spatial frequencies.

Stereoacuity improved across successive evaluations (Fig. 2):

• •

  Postoperative: 63.64 % (n = 28) had no stereopsis, while 13.64 % (n = 6) achieved good stereoacuity (≤100 arcseconds).

• •

  3-month PL: Proportions shifted to 40.91 % (n = 18) with no stereopsis and 27.27 % (n = 12) with good stereoacuity (χ² = 10.69, P = 0.0135).

• •

  Final follow-up: Only 18.18 % (n = 8) lacked stereopsis, whereas 47.73 % (n = 21) attained good stereoacuity (χ² = 18.67, P = 0.0003 vs. postoperative baseline). Stereoacuity gains correlated with PL duration (r = −0.51, P = 0.002; Spearman’s test).

Fig. 2.

Stereoacuity after perceptual learning.

This study provides the first evidence that home-based perceptual learning (PL) enhances medium to high SF contrast sensitivity (6–18 cpd) and promotes stereopsis recovery in pediatric esotropia patients following surgical alignment. Our findings advance the understanding of post-surgical neurorehabilitation by demonstrating that targeted binocular training can address residual cortical deficits, offering a novel adjuvant to traditional strabismus management.

The best timing for surgical intervention in children with esotropia, especially infantile esotropia, remains controversial. However, numerous studies have reached a consensus that early surgical correction to minimize the duration of constant esotropia could promote probability of developing stereopsis, facilitate treatment of amblyopia, restore normal appearance, and improve general health-related quality of life.1,3,7,21–24

The contrast sensitivity test provides a detailed description of an individual's spatial resolution of the visual system, the results of which can be abnormal in a variety of diseases, including anisometropia, amblyopia, strabismus, glaucoma, optic neuropathy, and brain lesions.25–28 The preoperative logCS values of patients in our results at low spatial frequencies (1.5 cpd, 3 cpd) were within the normal range of the general population,29–31 while a significant decline at medium spatial frequencies (6 cpd) and at high spatial frequencies (12 cpd, 18 cpd). Hence, we can conclude that esotropia may prevailingly affects the contrast sensitivity at medium and high spatial frequencies in children. Our findings showed that logCS was increased at medium to high spatial frequencies after strabismus surgery (6cpd, 12 cpd, 18 cpd), the values of which could be further improved by perceptual learning both at the time point of 3 months after training and final visit after training. And based off of that, perceptual learning can further improve medium and high-SF contrast sensitivity in children, showing the potential clinical application of perceptual learning in esotropia.

The selective improvement in medium and high-SF contrast sensitivity aligns with prior work in amblyopia, where PL enhances parvocellular pathway integrity through cortical reorganization.24,29,30 However, unlike amblyopia, esotropia is characterized by stronger interocular suppression and earlier onset, which may explain why low SFs (1.5–3 cpd) showed no significant gains. The observed 0.3–0.4 logCS improvements at 12–18 cpd (Cohen’s d = 0.62–0.78) surpass those reported for intermittent exotropia (0.2 logCS),16 suggesting esotropia’s unique neural adaptability to PL. These gains likely reflect reduced cortical suppression and restored binocular summation, as high SFs depend heavily on synchronized V1/V2 activity.1,20

Stereoacuity improvements, 47.73 % achieving ≤100 arcseconds post-PL versus 13.64 % post-surgery, mirror outcomes in surgically aligned intermittent exotropia,18,19 yet exceed rates seen in non-PL esotropia cohorts (20–30 %).20 The correlation between PL duration and stereoacuity gains (r = −0.51, P = 0.002) underscores the dose-dependent nature of neural plasticity, consistent with fMRI studies linking PL to increased interhemispheric coherence.

Visual perceptual learning through practice or training in perceptual tasks, such as the detection of visual gratings, stimulus orientation judgment, texture discrimination, retinal location recognition, and hyperacuity and vernier tasks, can progressively promotes perceptual performance. Perceptual learning may reflect plasticity in a complex set of brain networks.14,32–34 A number of studies have described the importance of binocular vision and supported the effectiveness of perceptual learning for intermittent exotropia,16–19 and this is the first study to evaluate the effectiveness of perceptual learning for esotropia after strabismus surgery, highlighting another therapeutic approach to treating esotropia. Notably, there existed five cases (11.26 %) with consecutive exotropia at the final follow-up. The incidence of consecutive exotropia secondarily to esotropia is estimated to 4 %∼27 %.35,36 Early onset of esotropia, preoperative angle of distance deviation, presence of postoperative adduction deficit, amblyopia and hyperopia are the risk factors for the development of consecutive exotropia.37–39 Since our study lack of control groups, we cannot elucidate the relationship between perceptual learning and consecutive exotropia, and further research is needed.

This study has several limitations that should be acknowledged. First, its retrospective design and the inclusion of four distinct types of esotropia may introduce inherent biases and constrain the ability to establish causal inferences. Second, the relatively small sample size limits the statistical power and precision of our findings, potentially increasing the risk of Type II errors (false negatives) and affecting the generalizability of the results across the broader esotropia population. Heterogeneity in age, preoperative visual function, and compliance patterns further underscores the need for caution in interpreting outcomes. Despite these limitations, we wish to emphasize that the primary findings remain consistent and clinically informative. To address these constraints in future work, we strongly advocate for prospective, randomized controlled trials (RCTs) with larger cohorts. Such studies would help validate the efficacy of perceptual learning protocols, establish evidence-based guidelines for post-surgical visual rehabilitation, and optimize personalized treatment strategies for esotropia patients.

During the follow-up period, we observed suboptimal adherence to the prescribed training regimen among some pediatric participants. This lack of engagement could be attributed to a combination of factors, including a lack of intrinsic motivation towards the tasks, the perceived repetitiveness of the exercises, or a mismatch between the stimulus characteristics and the child's individual preferences and cognitive profile. These observations underscore a critical challenge in perceptual learning interventions: sustaining children's motivation and active participation over time. In response, we propose that future iterations of perceptual learning programs necessitate a fundamental optimization. This can be achieved through the integration of Artificial Intelligence (AI)-driven adaptive algorithms that dynamically personalize task difficulty and content based on real-time performance feedback, thereby maintaining an appropriate challenge level and reducing monotony. Concurrently, Virtual Reality (VR)-based immersive environments can be leveraged to transcend mere entertainment; they offer a platform to create engaging, context-rich, and rewarding experiences that align with pediatric interests. By fully harnessing the synergies between AI and VR, we can significantly enrich the training content and context, address the root causes of poor compliance and foster a more profound and sustained engagement throughout the rehabilitation process.

In conclusion, perceptual learning can improve the contrast sensitivity at medium SF (6cpd) and high SFs (12 cpd, 18 cpd) and promote the recovery of stereoscopic vision after esotropia surgery, displaying great application prospects for the treatment of esotropia. Our results challenge the conventional focus on ocular alignment as the endpoint of esotropia management. By addressing residual cortical dysfunction, PL bridges the gap between surgical success and functional vision restoration. The 86 % compliance rate highlights the feasibility of home-based digital therapy in children, overcoming scalability barriers of clinic-based regimens. Notably, PL’s benefits emerged despite heterogeneous esotropia subtypes, suggesting broad applicability across etiologies.

This study establishes perceptual learning as a viable strategy to augment surgical outcomes in esotropia, targeting neural deficits refractory to alignment alone. By restoring medium-to- high-SF contrast sensitivity and stereopsis, PL empowers patients to achieve not only ocular normality but functional visual proficiency. These findings advocate for integrating PL into post-surgical protocols, heralding a paradigm shift toward comprehensive, plasticity-driven strabismus care.