To the Editor:

Acanthamoeba keratitis is often challenging to diagnose with traditional microbiological methods [1]. Confocal microscopy is a quick and noninvasive test that is a useful adjunct for diagnosis since it can image intrastromal acanthamoeba cysts that might not be captured on a corneal scraping, making it perhaps even more sensitive than culture or smear [2]. The Nidek Confoscan4 (NIDEK technologies, Fremont, CA) and Heidelberg HRT3 Rostock Cornea Module (RCM) (Heidelberg Engineering, Heidelberg, Germany) confocal microscopy systems are two systems currently used in the diagnosis of acanthamoeba keratitis [3]. The purpose of this study was to compare the diagnostic yield of the Nidek and Heidelberg confocal microscopes in the diagnosis of acanthamoeba keratitis.

The study population was a convenience sample of 6 eyes from 5 patients who were seen at the Proctor Foundation between 2013 and 2015 with suspected acanthamoeba keratitis. Participants had confocal microscopy performed by a single operator with both the Nidek Confoscan4 and Heidelberg HRT3 Rostock machines on the same day, before any corneal scrapings were performed. Two experienced ophthalmologists independently reviewed the 12 sets of scans for the presence of acanthamoeba cysts based on typical features (e.g., bright spots, double-walled structures, signet rings), masked to participant identity, clinical information, and microbiological test results [4]. All captured images were presented to the graders (HRT3 RCM range: 401–681 images per eye; Confoscan4 range: 701–800 images per eye). Discrepancies were resolved by discussion until consensus was reached. Ethical approval was obtained from the University of California, San Francisco; written informed consent was obtained from all participants.

Of the six eyes with suspected acanthamoeba keratitis, five subsequently had a culture positive for acanthamoeba. The remaining eye had a negative culture and smear and was ultimately considered not to have acanthamoeba keratitis. When assessing the five culture-positive eyes, the ophthalmologists graded all five of the Heidelberg scans as positive for acanthamoeba (two with high confidence and three with low confidence), and three of the Nidek scans as positive (three with high confidence and one with low confidence; the two Nidek scans incorrectly classified as negative were both of low confidence). Figure 1 depicts representative images from each device. For the culture-negative eye, the ophthalmologists graded the scans of both devices as negative for acanthamoeba, with low confidence for each. On average the Heidelberg device was 40% more sensitive than the Nidek device, though the small sample size provided very low certainty of this estimate (bootstrapped 95% CI of difference: 0–100%, 9999 replications). The difference between the two devices did not approach statistical significance (P = 0.48, McNemar’s test).

Fig. 1: Representative confocal microscopy images of three eyes with acanthamoeba keratitis.
figure 1

Each row is a distinct eye, with an image from the Heidelberg HRT3 Rostock Cornea Module (CRM) on the left and an image from the Nidek Confoscan4 on the right. The image thought most likely to represent acanthamoeba cysts by the masked ophthalmologist graders was chosen in each case. Images from the top two eyes (a–d) were judged to be positive for acanthamoeba for both microscopes, whereas images from the bottom eye (e, f) were judged positive for the HRT3 CRM but not the Confoscan4.

Full size image

The Heidelberg HRT3 RCM appeared to be a more sensitive diagnostic test than the Nidek Confoscan4 in this study, though the small sample size precluded rigorous statistical testing. Two of five culture-positive eyes had discrepant results, with both given a diagnosis of acanthamoeba-positive by the Heidelberg but not the Nidek. A review of confocal microscopy reported an axial resolution of 4 µm for the HRT3 RCM versus 8–25 µm for the Confoscan4, so it is plausible that the higher-resolution device would be more sensitive [5].

Graders generally had low confidence in their diagnosis for both devices. In particular, graders had low confidence for all scans of the two eyes with the discrepant grades—both for the positive grades (i.e., the HRT3 RCM scans) and the negative grades (i.e., the Confoscan4 scans). Confocal grading is inherently subjective and in our experience extremely dependent on the experience of the grader. The ophthalmologist graders in this study had the most experience with confocal diagnosis of acanthamoeba keratitis at our institution, and were confident in their diagnostic abilities. Nonetheless, it is possible that other graders might have reached a different conclusion for one or both machines.

Several additional limitations should be noted. Confocal microscopy images a small area of the cornea in a given scan, so it is possible scans were obtained in an area remote from the pathology. Stromal inflammation can result in false negatives if the inflammatory cells and edema mask acanthamoeba cysts, or false positives if macrophages are misclassified as acanthamoeba cysts. As mentioned above, the small sample size reduced the statistical power of the study, and the presence of only a single culture-negative case made it difficult to make conclusions about test specificity. Finally, the generalizability to other clinics with a different patient population, different confocal operator, or different graders is unclear.

Despite its limitations, this small study showed that both the Heidelberg and Nidek confocal microscopes provided the correct diagnosis for more than half of culture-positive acanthamoeba keratitis cases, and suggested that the Heidelberg machine may be slightly more sensitive.