Open Access
Artificial Intelligence  |   May 2025
Using Artificial Intelligence for an Efficient Prediction of Outcomes of Deep Anterior Lamellar Keratoplasty (DALK) in Advanced Keratoconus
Author Affiliations & Notes
  • Gairik Kundu
    Department of Cornea and Refractive Surgery, Narayana Nethralaya, Bangalore, India
  • Sharon D'Souza
    Department of Cornea and Refractive Surgery, Narayana Nethralaya, Bangalore, India
  • Durgalaxmi Modak
    Department of Cornea and Refractive Surgery, Narayana Nethralaya, Bangalore, India
  • Srihari Balaraj
    Department of Cornea and Refractive Surgery, Narayana Nethralaya, Bangalore, India
  • Rohit Shetty
    Department of Cornea and Refractive Surgery, Narayana Nethralaya, Bangalore, India
  • Rudy M. M. A. Nuijts
    Department of Ophthalmology, Maastricht University Medical Center, Maastricht, The Netherlands
  • Raghav Narasimhan
    Imaging, Biomechanics and Mathematical Modeling Solutions, Narayana Nethralaya, Bangalore, India
  • Abhijit Sinha Roy
    Imaging, Biomechanics and Mathematical Modeling Solutions, Narayana Nethralaya, Bangalore, India
  • Correspondence: Gairik Kundu, Department of Cornea and Refractive Surgery, Narayana Nethralaya, #121/C, Chord Rd., 1st ‘R’ Block, Rajaji Nagar, Bangalore 560010, India. e-mail: [email protected] 
Translational Vision Science & Technology May 2025, Vol.14, 30. doi:https://doi.org/10.1167/tvst.14.5.30
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Gairik Kundu, Sharon D'Souza, Durgalaxmi Modak, Srihari Balaraj, Rohit Shetty, Rudy M. M. A. Nuijts, Raghav Narasimhan, Abhijit Sinha Roy; Using Artificial Intelligence for an Efficient Prediction of Outcomes of Deep Anterior Lamellar Keratoplasty (DALK) in Advanced Keratoconus. Trans. Vis. Sci. Tech. 2025;14(5):30. https://doi.org/10.1167/tvst.14.5.30.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: To identify and analyze clinical risk factors and imaging parameters influencing the outcomes of deep anterior lamellar keratoplasty (DALK) for advanced keratoconus (KC) using an artificial intelligence (AI) model.

Methods: This study included 250 DALK eyes with a 5-year follow-up for advanced KC. The DALK eyes were classified as having “favorable” or “unfavorable” outcomes based on graft clarity, scarring at the graft–host interface involving the visual axis which was not pre-existing, early suture loosening less than 3 months after the surgery, corneal vascularization reaching up to or into the graft–host junction at any follow up period, persistent corneal edema greater than 3 months after surgery, and change in visual acuity. Clinical risk factors were determined through a detailed clinical evaluation and questionnaire assessment and included the presence of systemic allergy, ocular allergy, or eye rubbing. Immunoglobulin E (IgE) and vitamin D and B12 levels were obtained from blood investigations. A total of 37 tomographic parameters were exported from an OCULUS Pentacam HR. An AI model was then built to assess these risk factors and imaging parameters. The area under the curve (AUC) and other metrics were evaluated.

Results: The AI model classified 92.2% and 89.4% cases as favorable or unfavorable, respectively, based on clinical risk factors and imaging parameters. Systemic allergy, IgE, eye rubbing, and vitamin D had the highest information gains followed by posterior corneal data from the Pentacam HR. The AI model achieved an AUC of 0.957 with sensitivity of 98% and specificity of 85.6%.

Conclusions: Our findings demonstrate the importance of preoperative risk stratification, which can affect surgical outcomes of DALK using AI.

Translational Relevance: Better identification and control of these factors would enable better management and outcomes of DALK for advanced KC.

Introduction
Keratoconus (KC) is an inflammatory disorder causing asymmetrical and progressive corneal ectasia, which typically presents in early adolescence and progresses into the second or third decades of life.1 The major risk factors for KC include eye rubbing, a history of ocular and systemic allergy including atopy, and a family history of KC.2 Early detection of KC is important for better treatment outcomes, as timely intervention with procedures such as corneal cross-linking may avoid more invasive treatments such as corneal transplantation.3 For the advanced KC eye, deep anterior lamellar keratoplasty (DALK) can be performed.46 It is an alternative to full-thickness penetrating keratoplasty for the treatment of KC and has several advantages over penetrating keratoplasty, such as a lower risk of graft rejection and preservation of endothelial cell density.79 This technique can significantly improve the vision-related quality of life in patients, and studies have reported visual outcomes comparable to penetrating keratoplasty.1012 However, DALK is a technically more challenging procedure to perform compared to penetrating keratoplasty and requires a greater level of surgical expertise.13 Although there are various intra- and postoperative factors that can alter the outcomes of DALK, the predictive correlations between demographic and clinical risk factors such as systemic allergy, ocular allergy, and eye rubbing are unknown. Therefore, we aimed to use artificial intelligence (AI) to integrate and stratify various preoperative clinical risk factors and imaging parameters from a Pentacam HR (OCULUS, Wetzlar, Germany), which can predict long-term outcomes of DALK.14 
Methods
This was a retrospective analysis of 250 eyes of 250 KC patients who underwent DALK for advanced KC in a tertiary-care eye hospital between August 2016 and June 2018. The study was approved by the institutional ethics committee and followed the tenets of the Declaration of Helsinki. All of the surgeries were done by trained cornea surgeons. Eyes with intraoperative complications, including corneal perforations requiring conversion to penetrating keratoplasty, were excluded from the study. Detailed ocular and systemic histories were recorded, including history of ocular and systemic allergy, eye rubbing, and family history of KC. Patients were also requested to complete a questionnaire that collected demographic data regarding age, gender, occupation, and educational status. Ocular examination by slit-lamp biomicroscopy was done to evaluate for clinical signs of KC and ocular allergy, including severity and activity of allergy. Corneal vascularization, opacification, and scarring can affect corneal graft outcomes.13 Patients were advised to undergo blood investigations for serum immunoglobulin E (IgE) and vitamin D, as these can influence the outcomes.13 Corneal tomography by the Pentacam HR was done to diagnose and grade the severity of KC. 
All of the eyes included were patients diagnosed with advanced KC (stage 4 on the Amsler–Krumeich KC severity scale) and who underwent DALK. Only eyes with at least 5 years of follow-up after surgery were included. Exclusion criteria were endothelial pathologies, ectatic conditions other than KC (e.g., keratoglobus), pellucid marginal degeneration, corneal scars due to pathologies other than KC, post-refractive surgery ectasia, autoimmune disorders, previous systemic immunosuppression or systemic immunosuppression post-surgery, any ocular surgery prior to the first visit of the patient or during the course of disease follow-up, or the use of topical eye drops other than mast cell stabilizers or antihistamines and lubricating eye drops preoperatively. Elevated IgE is a hallmark of allergic diseases. However, high IgE levels are also found in a number of infectious diseases, such as parasite infections, human immunodeficiency virus infection, Mycobacterium tuberculosis infection, cytomegalovirus infection, Epstein–Barr virus, leprosy, and candidiasis. Inflammatory diseases such as eosinophilic granulomatosis with polyangiitis and Kawasaki disease are also characterized by elevated IgE levels. These systemic factors that can affect IgE levels were excluded.15 
Data Collection
Clinical history, examination findings and diagnostic features from the Pentacam HR were included in the data analyses. The responses to questionnaire parameters that were specifically included in the AI model were as follows: 
  • Eye rubbing—present or absent
  • Systemic allergy/asthma/eczema—present or absent
  • Ocular allergy—present or absent
Ocular allergy was defined as patients having allergy/hypersensitivity that can affect only the eye, including lid, conjunctiva, and/or cornea. Forms of allergy include seasonal or perennial allergic conjunctivitis and the less frequent vernal and atopic keratoconjunctivitis involving only the cornea and no other systemic involvement.16 
Only those Pentacam scans were used that did not have any blinking or motion artifacts. These scans were automatically classified as “OK” by the Pentacam software. Further, the detected anterior and posterior edges of the corneal scans were manually confirmed so that no missing portions of the detected edges confounded the tomography of the cornea.17 Supplementary Table S1 lists all of the Pentacam HR tomographic and clinical parameters used in the AI model. A total of 40 tomographic and clinical parameters were analyzed. The patient's IgE (IU/mL), vitamin D (ng/mL), and vitamin B12 (pg/mL) were measured from the patient's blood sample and included in the AI analysis along with clinical and diagnostic data. Other clinical parameters included preoperative corneal scar and depth, if present, and the type of DALK surgery done (manual or big bubble).14 
In the clinical setting, we evaluate corneal transplant patients on slit lamp and look for certain clinical features to qualify as favorable or unfavorable outcome. These include features of graft rejection or failure (e.g., loss of graft clarity, edema, scarring) and those that could predispose the graft to a failure or rejection, such as inflammation, vascularization, raised intraocular pressure, poor graft–host junction adherence, recurrence of ectasia, and astigmatism.1820 In this study, unfavorable outcomes were defined as loss of graft clarity, which could be due to rejection or failure; scarring at the graft–host interface involving the visual axis which was not pre-existing; factors that could predispose the eye to early suture loosening less than 3 months after the surgery1822; corneal vascularization reaching up to or into the graft–host junction at any follow-up period21; and persistent corneal edema 3 months after the surgery. 
Another important factor that was considered as an unfavorable outcome was a drop in corrected distance visual acuity (CDVA) by two lines or more on the Snellen distance visual acuity chart after the surgery.2325 Favorable outcomes were defined as maintenance of graft clarity until the last follow-up, no early suture loosening, absence of corneal vascularization, and improvement by one line or more in CDVA.1825 Graft clarity was graded subjectively by two independent cornea surgeons (SD, GK) with more than 10 years of experience. Graft clarity was graded as grade 4 if grafts were optically clear with excellent view of iris details, grade 2 or 3 (borderline) if there was moderate to significant corneal haze with or without good view of iris details, and grade 1 or 0 (failed) for opaque grafts with poor views of iris and anterior segment details.26 
Artificial Intelligence Model
The subjective responses and clinical risk factors from the questionnaire were used as categorical variables in the AI model. The quantitative values for serum IgE, vitamin D, and vitamin B12 were used as continuous variables in this model. Similarly, Pentacam HR preoperative indices were also used as continuous variables. We used the random forest classifier model to train and classify the dataset. The random forest classifier is an ensemble of many decision trees put together and is used to classify or predict a particular target with the given set of features/parameters. It was an efficient AI classifier for the analyses of multiple parameters.27,28 This random forest classifier consisted of 10 trees, and the average of the outcome from the 10 trees was considered (Orange data mining software; University of Ljubljana, Ljubljana, Slovenia). The leave-one-out method was used for cross-validation (Orange data mining software). The clinical classification of outcome (favorable or unfavorable) was used as the target classification for the AI model. This allowed us to assess whether the subjective responses from the questionnaire and Pentacam HR parameters at preoperative visits were able to predict the outcomes of DALK. 
Statistical Analyses
After assessing for normality of distribution with the Kolmogorov–Smirnov test, the medians and interquartile ranges (IQRs) were calculated for quantitative variables, such as IgE, vitamin D, vitamin B12, and Pentacam HR indices; otherwise, mean ± SD was used. The Kruskal–Wallis test was used for comparison of preoperative data. The categorical variables from questionnaires were represented using frequency (contingency) tables. The performance of the AI model was evaluated using several parameters including area under the curve (AUC), sensitivity, specificity, classification accuracy, precision, recall, and F1 score. The Orange3 (version 3.25.0) data mining package was used to build the AI models, and MedCalc 19 (MedCalc Software, Ostend, Belgium) was used for further statistical analyses.27,28 The Orange software uses a rank widget that scores variables according to their correlation with discrete or numeric target variables, based on applicable internal scorers (such as information gain, χ2, and linear regression) and any connected external models that support scoring, such as linear regression, logistic regression, or random forest. Various scoring methods are used to arrive at the top features, such as information gain, gain ratio, and Gini ratio. Figure 1 shows the workflow in the Orange software used while building the AI model. 
Figure 1.
 
Workflow of the AI model using the Orange software.
Figure 1.
 
Workflow of the AI model using the Orange software.
The basic steps were as follows: 
  • 1. Data extraction and uploading were performed in the File widget.
  • 2. The Select Columns widget was used to manually compose the data domain. Orange distinguishes among ordinary attributes, class attributes, and meta-attributes. For example, to build a classification model, the domain would be composed of a set of attributes and a discrete class attribute.
  • 3. The Test & Score widget first shows a table with different classifier performance measures, such as classification accuracy and AUC. It then outputs evaluation results, which can be used by other widgets for analyzing the performance of classifiers, such as receiver operating characteristics analysis or confusion matrix.
  • 4. The Rank widget scores variables according to their correlation with discrete or numeric target variables, based on applicable internal scorers (such as information gain, χ2, and linear regression) and any connected external models that support scoring, such as linear regression, logistic regression, or random forest.
  • 5. The Confusion Matrix gives the number/proportion of instances between the predicted and actual classes. The selection of the elements in the matrix feeds the corresponding instances into the output signal. This way, one can observe which specific instances were misclassified and how.
  • 6. The Prediction widget receives a dataset and one or more predictors (predictive model) and outputs the predictions.
Results
A total of 250 eyes of 250 patients were analyzed, of which 135 were males and 115 were females, with an average age of 28 ± 3.9 years. The median follow-up was 5.9 years (range, 5.2–6.7). All eyes studied were Indian–Asian. The demographic and preoperative variables are listed in Table 1. After assessing the quality of the Pentacam images, a total of 250 eyes were included in the study. Out of the 250 DALK eyes, 176 eyes were operated on using the big-bubble technique, and 74 eyes underwent manual dissection.14 The mean CDVA improved from 1.22 ± 0.26 logMAR preoperatively to 0.46 ± 0.19 logMAR postoperatively in the favorable outcomes group (P < 0.05), and CDVA changed from 0.96 ± 0.55 logMAR preoperatively to 1.13 ± 0.16 logMAR postoperatively in the unfavorable outcomes group (P > 0.05). Tables 2 and 3 show preoperative comparisons of the clinical risk factors and tomographic parameters between the unfavorable and favorable outcomes group. Among the clinical parameters, vitamin B12, vitamin D, and IgE were similar between the two groups preoperatively, but the two groups had significant differences with respect to the presence or absence of systemic allergies and eye rubbing (P < 0.05). All Pentacam HR indices (Table 3) were similar between the two groups preoperatively (P > 0.05). 
Table 1.
 
Demographic and Preoperative Features of Eyes Included in the AI Model
Table 1.
 
Demographic and Preoperative Features of Eyes Included in the AI Model
Table 2.
 
Preoperative Differences in Clinical Risk Factors Between the Groups Clinically Classified as Unfavorable and Favorable
Table 2.
 
Preoperative Differences in Clinical Risk Factors Between the Groups Clinically Classified as Unfavorable and Favorable
Table 3.
 
Preoperative Differences in Tomographic Parameters Between the Groups Clinically Classified as Unfavorable and Favorable
Table 3.
 
Preoperative Differences in Tomographic Parameters Between the Groups Clinically Classified as Unfavorable and Favorable
The AI model accurately classified 92.2% and 89.4% of the eyes in the favorable and unfavorable outcomes groups, respectively. Only 7.8% and 10.6% were classified otherwise in each group, respectively. The AUC, sensitivity, specificity, classification accuracy, precision, recall, and F1 score were 0.957, 98%, 86%, 91.4%, 91.3%, 0.91, and 0.92, respectively. The top 10 parameters identified by the AI model in order of decreasing importance were as follows: 
  • 1. Systemic allergy
  • 2. IgE
  • 3. Eye rubbing
  • 4. Vitamin D
  • 5. Defocus (corneal back)
  • 6. Ocular allergy
  • 7. Spherical aberration (corneal back)
  • 8. Central keratoconus index (CKI)
  • 9. Maximum curvature of anterior surface
  • 10. Minimum corneal thickness
Supplementary Table S2 gives the overall performance of the random forest AI model and the contribution of each of the top 10 parameters to the performance of the AI model. Further, we performed post hoc analyses of the four predicted groups based on clinical classification and AI classification (Tables 4 and 5). The medians and IQRs of clinical risk factors are summarized in Table 4. The IgE levels were the lowest in group 4 (“favorable” according to both clinical and AI classification). Similarly, group 1 (“unfavorable” according to both clinical classification and AI classification) showed the highest IgE levels. The vitamin D levels had the lowest value in group 1 and highest in group 4. Group 1 had a higher proportion of systemic allergy and eye rubbing as compared to other groups. Eyes in group 2 (“unfavorable” with clinical classification but “favorable” with AI classification) tended to have a lower serum IgE and higher vitamin D with a lower proportion of systemic allergies, eye rubbing, and ocular allergy as compared to group 3 (“favorable” with clinical classification but “unfavorable” with AI classification) 
Table 4.
 
Clinical Risk Parameters
Table 4.
 
Clinical Risk Parameters
Table 5.
 
Tomographic Parameters
Table 5.
 
Tomographic Parameters
Table 5 shows the medians and IQRs of the Pentacam HR parameters of the four predicted groups. Group 1 had a higher Kmax value as compared to group 4. Group 1 also showed a lower Belin–Ambrósio enhanced ectasia total deviation display (BAD-D) and lower posterior corneal surface Zernike coefficients as compared to group 4. Interestingly, group 2 (“unfavorable” with clinical classification but “favorable” with AI classification) had a lower Kmax value and lower Pentacam HR–derived indices, including BAD-D, index surface variance (ISV), index vertical asymmetry (IVA), index height decentration (IHD), and index height asymmetry (IHA), as well as lower posterior surface Zernike coefficients compared to group 3 (“favorable” with clinical classification but “unfavorable” with AI classification). Figure 2 shows the plot of receiver operating characteristics curve of the random forest model predictions. 
Figure 2.
 
Receiver operating characteristics (ROC) curve for the AI model.
Figure 2.
 
Receiver operating characteristics (ROC) curve for the AI model.
Discussion
Although various risk factors have been implicated in the progression of KC and corneal graft outcomes,28 a number of risk factors still are not clearly understood. Using AI to profile risk factors leading to a favorable or an unfavorable outcome in a large sample has not been attempted. Systemic allergy status was an important factor picked up by the AI analyses affecting outcomes following DALK. Allergic disease has been continuously suggested to be one of the important risk factors for onset and progression of KC.29 There has been some debate around whether ocular allergy alone or other allergic systemic diseases might promote the development or progression of KC. Large-scale studies and meta-analyses have shown that systemic allergy has a significant association with KC.30,31 Pre-existing allergy can also affect KC patients after DALK, as shown in studies where uncontrolled ocular allergy predisposed the eyes to corneal vascularization following DALK.21 
In our study, ocular allergy was one of top 10 parameters detected by the AI. IgE was also a parameter identified by the AI analyses. IgE and its relationship with allergy are known.28,32 It is a common mediator in risk factors implicated in progression of KC such as allergy, atopy, and eye rubbing. Studies have shown serum that IgE is significantly elevated in KC patients who have allergy.3335 Interestingly, elevated serum IgE can also be associated with a subset of KC patients without signs of ocular allergy. Thus, these patients can have subclinical allergy and inflammation on the ocular surface which, in turn, can have an impact on corneal graft outcomes after DALK. Corneal vascularization after DALK was related to inadequate control of allergy status and early discontinuation of topical steroids and allergy management.21 In both penetrating keratoplasty and deep anterior lamellar keratoplasty, the management of inflammation due to allergy in the perioperative period has been shown to be crucial for achieving successful outcomes after corneal transplantation; thus, long-standing allergy leading to limbal stem cell deficiency should be considered prior to surgery.36 It is important to control the ocular surface and systemic allergy adequately, which may lead to reduced eye rubbing and associated inflammation of the ocular surface, thereby also attaining better treatment outcomes after DALK for advanced KC.37,38 Systemic involvement invariably may require a multicentric approach to optimize outcomes.37 
Eye rubbing also affected graft outcomes in our study. Eye rubbing and its association with KC development and progression are known. Studies have shown a history of eye rubbing in KC ranging from 50% to 80% of the patients.3943 Mazharian et al.44 showed that cessation of eye rubbing halted progression in over 80% of eyes studied with KC. The frequency and force of rubbing are important factors that can influence corneal eye rubbing related changes, which make the corneal shape susceptible to changes in force in addition to intraocular pressure in KC eyes, and can influence graft outcomes.45,46 There have been eyes where persistent eye rubbing led to an increase in corneal curvature and reduction in uncorrected visual acuity after DALK.47 
Serum vitamin D level was picked up by the AI analyses as an important contributor to graft outcomes. Low vitamin D levels have been shown to contribute to the development and progression of KC.28,48 Studies have shown that individuals with both progressive and non-progressive KC have significantly lower serum vitamin D levels than those without KC.48 A study on impact of vitamin D supplementation on systemic biomarkers of collagen degradation analyzed by enzyme-linked immunosorbent assays showed stabilization of KC progression in 60% of patients (72% of eyes) after 12 months with vitamin D supplementation.49 Animal studies have shown that vitamin D deficiency can lead to delayed or prolonged corneal wound healing, higher risks for ocular surface infections, and subsequent stromal opacification, resulting in visual loss. Also, vitamin D deficiency could play a role in wound healing after DALK.50 Patients with KC, particularly those with progressive disease, may have lower serum levels of vitamin D than the normal population51 and may have a decreased expression of vitamin D receptor in the epithelium over the ectatic zones of cornea in KC.52 
We also evaluated imaging parameters from the Pentacam HR preoperatively. The corneal higher order aberrations from the corneal back surface were identified by the AI analyses to influence outcomes. Interestingly, preoperatively there were no significant differences among the Pentacam HR parameters between the two clinical groups (Table 3). However, the AI model identified defocus and spherical aberration from the corneal back surface, along with CKI, Kmax, and thinnest corneal thickness (TCT), as important factors affecting postoperative outcomes. A study by Kemer Atik et al.53 investigated the effect of KC stage according to the Amsler–Krumeich classification system (utilizing preoperative keratometry and pachymetry) on postoperative DALK outcomes. Interestingly, these were not found useful in predicting postoperative outcomes. In another study, greater preoperative corneal higher order aberrations were associated with poor visual acuity after DALK.54 Thus, aberrations in the posterior corneal surface (which is preserved in DALK) can affect visual quality outcomes after DALK. 
A limitation of the present study was that it was performed entirely on an Indian–Asian population. Although the grading process was as per clinical norms, determining graft clarity and the grading used were subjective and hence could have varied from one clinician to another. Another limitation was that the study did not quantify changes in the levels of IgE, vitamin D, and vitamin B12 at follow-ups, as this was only a single time point preoperative measurement. We also did not study whether the effect of interventions for high IgE or low vitamin D via systemic supplementation could impact outcomes after DALK. Although the more common topographical factors are intuitive for most clinicians treating KC, some risk factors can be missed, and this study more precisely identified Pentacam indices that are critical for a favorable outcome after DALK. This study is unique, as it predicted corneal graft outcomes using preoperative risk factors and corneal tomography only. Stratifying these factors with AI could help in better evaluating patients undergoing DALK for KC and predicting the need for additional interventions to optimize outcomes. 
Acknowledgments
Disclosure: G. Kundu, None; S. D'Souza, None; D. Modak, None; S. Balaraj, None; R. Shetty, AI diagnostics in corneal tomography (P); R.M.M.A. Nuijts, None; R. Narasimhan, None; A.S. Roy, AI diagnostics in corneal tomography (P) 
References
Shetty R, Deshmukh R, Ghosh A, Sethu S, Jayadev C. Altered tear inflammatory profile in Indian keratoconus patients – the 2015 Col Rangachari Award paper. Indian J Ophthalmol. 2017; 65(11): 1105–1108. [PubMed]
Hashemi H, Heydarian S, Hooshmand E, et al. The prevalence and risk factors for keratoconus: a systematic review and meta-analysis. Cornea. 2020; 39(2): 263–270. [CrossRef] [PubMed]
Bui AD, Truong A, Pasricha ND, Indaram M. Keratoconus diagnosis and treatment: recent advances and future directions. Clin Ophthalmol. 2023; 17: 2705–2718. [CrossRef] [PubMed]
Melles GR, Lander F, Rietveld FJ, Remeijer L, Beekhuis WH, Binder PS. A new surgical technique for deep stromal, anterior lamellar keratoplasty. Br J Ophthalmol. 1999; 83(3): 327–333. [CrossRef] [PubMed]
Coombes AG, Kirwan JF, Rostron CK. Deep lamellar keratoplasty with lyophilised tissue in the management of keratoconus. Br J Ophthalmol. 2001; 85(7): 788–791. [CrossRef] [PubMed]
Reddy JC, Modiwala Z, Mathew M. Lamellar keratoplasty in keratoconus. In: Das S, ed. Keratoconus. Singapore: Springer Nature; 2022: 205–220.
Keane M, Coster D, Ziaei M, Williams K. Deep anterior lamellar keratoplasty versus penetrating keratoplasty for treating keratoconus. Cochrane Database Syst Rev. 2014; 2014(7): CD009700. [PubMed]
Chen G, Tzekov R, Li W, Jiang F, Mao S, Tong Y. Deep anterior lamellar keratoplasty versus penetrating keratoplasty: a meta-analysis of randomized controlled trials. Cornea. 2016; 35(2): 169–174. [CrossRef] [PubMed]
Ang M, Mehta JS. Deep anterior lamellar keratoplasty as an alternative to penetrating keratoplasty. Ophthalmology. 2011; 118(11): 2306–2307. [CrossRef] [PubMed]
Watson SL, Ramsay A, Dart JK, Bunce C, Craig E. Comparison of deep lamellar keratoplasty and penetrating keratoplasty in patients with keratoconus. Ophthalmology. 2004; 111(9): 1676–1682. [CrossRef] [PubMed]
Reinhart WJ, Musch DC, Jacobs DS, Lee WB, Kaufman SC, Shtein RM. Deep anterior lamellar keratoplasty as an alternative to penetrating keratoplasty: a report by the American Academy of Ophthalmology. Ophthalmology. 2011; 118(1): 209–218. [CrossRef] [PubMed]
Yildiz E, Toklu M, Turan Vural E. Vision-related quality of life before and after deep anterior lamellar keratoplasty. Eye Contact Lens. 2018; 44(3): 144–148. [CrossRef] [PubMed]
Nanavaty MA, Vijjan KS, Yvon C. Deep anterior lamellar keratoplasty: a surgeon's guide. J Curr Ophthalmol. 2018; 30(4): 297–310. [CrossRef] [PubMed]
Pellegrini M, Yu AC, Busin M. Deep anterior lamellar keratoplasty for keratoconus: elements for success. Saudi J Ophthalmol. 2022; 36(1): 36–41. [CrossRef] [PubMed]
Kostova P, Papochieva V, Miteva D, et al. Elevated IgE levels—an allergy or an underlying inborn error of immunity in children with recurrent infections? Antibodies (Basel). 2023; 12(4): 70. [CrossRef] [PubMed]
Friedlaender MH . Ocular allergy. Curr Opin Allergy Clin Immunol. 2011; 11(5): 477–82. [CrossRef] [PubMed]
Matalia H, Narasimhan R, Chinnappaiah N, Kumar V, Sinha Roy A. An interesting case of data gaps in measurement of corneal curvature with Scheimpflug tomography. J Refract Surg. 2020; 36(5): 350–351. [CrossRef] [PubMed]
El Sayed SH, Ismail MM, El Sawy MF, El Hagaa AA, Abdel Aziz MS. The visual outcome and complications in deep anterior lamellar keratoplasty for keratoconus. Menoufia Med J. 2016; 29: 587–92
Huang T, Zhang X, Wang Y, Zhang H, Huand A, Gao N. Outcomes of deep anterior lamellar keratoplasty using the big-bubble technique in various corneal diseases. Am J Ophthalmol. 2012; 154(2): 282–289.e1. [CrossRef] [PubMed]
Karimian F, Feizi S. Deep anterior lamellar keratoplasty: indications, surgical techniques and complications. Middle East Afr J Ophthalmol. 2010; 17(1): 28–37. [CrossRef] [PubMed]
Pellegrini M, Scorcia V, Giannaccare G, et al. Corneal neovascularisation following deep anterior lamellar keratoplasty for corneal ectasia: incidence, timing and risk factors. Br J Ophthalmol. 2022; 106(10): 1363–1367. [CrossRef] [PubMed]
Kodavoor SK, Rathi N, Dandapani R. Complications in deep anterior lamellar keratoplasty – a retrospective interventional analysis in a large series. Indian J Ophthalmol. 2022; 70(10): 3501–3507. [CrossRef] [PubMed]
Sarnicola V, Toro P, Sarnicola C, Sarnicola E, Ruggiero A. Long-term graft survival in deep anterior lamellar keratoplasty. Cornea. 2012; 31(6): 621–626. [CrossRef] [PubMed]
Arundhati A, Chew MC, Lim L, et al. Comparative study of long-term graft survival between penetrating keratoplasty and deep anterior lamellar keratoplasty. Am J Ophthalmol. 2021; 224: 207–216. [CrossRef] [PubMed]
Panda A, Bageshwar LM, Ray M, Singh JP, Kumar A. Deep lamellar keratoplasty versus penetrating keratoplasty for corneal lesions. Cornea. 1999; 18(2): 172–175. [CrossRef] [PubMed]
Ardjomand N, McAlister JC, Rogers NJ, Tan PH, George AJ, Larkin DF. Modulation of costimulation by CD28 and CD154 alters the kinetics and cellular characteristics of corneal allograft rejection. Invest Ophthalmol Vis Sci. 2003; 44(9): 3899–3905. [CrossRef] [PubMed]
Shetty R, Kundu G, Narasimhan R, et al. Artificial intelligence efficiently identifies regional differences in the progression of tomographic parameters of keratoconic corneas. J Refract Surg. 2021; 37(4): 240–248. [CrossRef] [PubMed]
Kundu G, Shetty N, Shetty R, et al. Artificial intelligence-based stratification of demographic, ocular surface high-risk factors in progression of keratoconus. Indian J Ophthalmol. 2023; 71(5): 1882–1888. [CrossRef] [PubMed]
Woodward MA, Blachley TS, Stein JD. The association between sociodemographic factors, common systemic diseases, and keratoconus: an analysis of a nationwide heath care claims database. Ophthalmology. 2016; 123(3): 457–465.e2. [CrossRef] [PubMed]
Seth I, Bulloch G, Vine M, et al. The association between keratoconus and allergic eye diseases: a systematic review and meta-analysis. Clin Exp Ophthalmol. 2023; 51(4): O1–O16. [CrossRef] [PubMed]
Leonardi A, Bogacka E, Fauquert JL, et al. Ocular allergy: recognizing and diagnosing hypersensitivity disorders of the ocular surface. Allergy. 2012; 67(11): 1327–1337. [CrossRef] [PubMed]
Kittaka H, Tominaga M. The molecular and cellular mechanisms of itch and the involvement of TRP channels in the peripheral sensory nervous system and skin. Allergol Int. 2017; 66(1): 22–30. [CrossRef] [PubMed]
Rahi A, Davies P, Ruben M, Lobascher D, Menon J. Keratoconus and coexisting atopic disease. Br J Ophthalmol. 1977; 61(12): 761–764. [CrossRef] [PubMed]
Kemp EG, Lewis CJ. Immunoglobulin patterns in keratoconus with particular reference to total and specific IgE levels. Br J Ophthalmol. 1982; 66(11): 717–720. [CrossRef] [PubMed]
Cingu AK, Cinar Y, Turkcu FM, et al. Effects of vernal and allergic conjunctivitis on severity of keratoconus. Int J Ophthalmol. 2013; 6(3): 370–374. [PubMed]
Feizi S, Javadi MA, Alemzadeh-Ansari M, Arabi A, Shahraki T, Kheirkhah A. Management of corneal complications in vernal keratoconjunctivitis: a review. Ocul Surf. 2021; 19: 282–289. [CrossRef] [PubMed]
Ahuja P, Dadachanji Z, Shetty R, et al. Relevance of IgE, allergy and eye rubbing in the pathogenesis and management of keratoconus. Indian J Ophthalmol. 2020; 68(10): 2067–2074. [PubMed]
Patel P, Shetty R, Ghosh A, et al. Ocular surface immune trafficking and eye rubbing in keratoconus, and its impact on treatment outcomes. In Proceedings of the CXL Experts’ Meeting 2018. Dietikon, Switzerland: Light for Sight; 2018.
Sideroudi H, Flockerzi E, Daas L, et al. Baseline characteristics of 1976 patients with ectatic corneal disorders at a single center from 2010 to 2021: a cross-sectional study of the Homburg Keratoconus Center. Eye Contact Lens. 2023; 49(9): 392–398. [PubMed]
Shneor E, Millodot M, Blumberg S, Ortenberg I, Behrman S, Gordon-Shaag A. Characteristics of 244 patients with keratoconus seen in an optometric contact lens practice. Clin Exp Optom. 2013; 96(2): 219–224. [CrossRef] [PubMed]
McGhee CN, Kim BZ, Wilson PJ. Contemporary treatment paradigms in keratoconus. Cornea. 2015; 34(suppl 10): S16–S23. [PubMed]
Rabinowitz YS . The genetics of keratoconus. Ophthalmol Clin North Am. 2003; 16(4): 607–620, vii. [CrossRef] [PubMed]
Al Shammari Z, Al Shammari R, Al Orf S, et al. Prevalence, clinical features and associated factors of keratoconus patients attending ophthalmology department, King Khalid Hospital, Hail City, Saudi Arabia. EC Ophthalmol. 2016; 3(5): 388–400.
Mazharian A, Flamant R, Elahi S, Panthier C, Rampat R, Gatinel D. Medium to long term follow up study of the efficacy of cessation of eye-rubbing to halt progression of keratoconus. Front Med (Lausanne). 2023; 10: 1152266. [CrossRef] [PubMed]
McMonnies CW . Mechanisms of rubbing-related corneal trauma in keratoconus. Cornea. 2009; 28(6): 607–615. [CrossRef] [PubMed]
Hawkes E, Nanavaty MA. Eye rubbing and keratoconus: a literature review. Int J Kerat Ect Cor Dis. 2014; 3(3): 118–122.
Abad JC, Ocampo-Patiño A, Vanegas-Ramírez C. Short-term bilateral keratoconus progression after deep anterior lamellar keratoplasty in one eye and intracorneal ring segments and corneal crosslinking in the other due to eye rubbing. J Cataract Refract Surg. 2020; 46(12): e44–e47. [CrossRef] [PubMed]
Akkaya S, Ulusoy DM. Serum vitamin D levels in patients with keratoconus. Ocul Immunol Inflamm. 2020; 28(3): 348–353. [CrossRef] [PubMed]
Lasagni Vitar RM, Fonteyne P, Knutsson KA, et al. Vitamin D supplementation impacts systemic biomarkers of collagen degradation and copper metabolism in patients with keratoconus. Transl Vis Sci Technol. 2022; 11(12): 16. [CrossRef] [PubMed]
Lu X, Chen Z, Lu J, Watsky M. Effects of topical 1,25 and 24,25 vitamin D on diabetic, vitamin D deficient and vitamin D receptor knockout mouse corneal wound healing. Biomolecules. 2023; 13(7): 1065. [CrossRef] [PubMed]
Gorimanipalli B, Shetty R, Sethu S, Khamar P. Vitamin D and eye: current evidence and practice guidelines. Indian J Ophthalmol. 2023; 71(4): 1127–1134. [CrossRef] [PubMed]
Shivakumar S, Shetty R, Ghosh A, Jeyabalan N. Vitamin D enhances the autophagic lysosomal clearance in oxidatively stressed human corneal epithelial cells: a therapeutic intervention for keratoconus. Invest Ophthalmol Vis Sci. 2019; 60: 2819.
Kemer Atik B, Emul M, Kirgiz A, Cubuk MO, Ahmet S, Kandemir Besek N. Comparison of deep anterior lamellar keratoplasty and penetrating keratoplasty in keratoconic eyes with previous hydrops-related corneal scar. Cornea. 2023; 42(11): 1446–1450. [CrossRef] [PubMed]
Yamaguchi T, Ohnuma K, Tomida D, et al. The contribution of the posterior surface to the corneal aberrations in eyes after keratoplasty. Invest Ophthalmol Vis Sci. 2011; 52(9): 6222–6229. [CrossRef] [PubMed]
Figure 1.
 
Workflow of the AI model using the Orange software.
Figure 1.
 
Workflow of the AI model using the Orange software.
Figure 2.
 
Receiver operating characteristics (ROC) curve for the AI model.
Figure 2.
 
Receiver operating characteristics (ROC) curve for the AI model.
Table 1.
 
Demographic and Preoperative Features of Eyes Included in the AI Model
Table 1.
 
Demographic and Preoperative Features of Eyes Included in the AI Model
Table 2.
 
Preoperative Differences in Clinical Risk Factors Between the Groups Clinically Classified as Unfavorable and Favorable
Table 2.
 
Preoperative Differences in Clinical Risk Factors Between the Groups Clinically Classified as Unfavorable and Favorable
Table 3.
 
Preoperative Differences in Tomographic Parameters Between the Groups Clinically Classified as Unfavorable and Favorable
Table 3.
 
Preoperative Differences in Tomographic Parameters Between the Groups Clinically Classified as Unfavorable and Favorable
Table 4.
 
Clinical Risk Parameters
Table 4.
 
Clinical Risk Parameters
Table 5.
 
Tomographic Parameters
Table 5.
 
Tomographic Parameters
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×