Direct potassium hydroxide (KOH) testing is a simple, quick and inexpensive technique integral to dermatological practice for identifying fungal organisms (Figure 1C) (49, 50). It involves retrieving the specimen from the nail bed and underneath the nail plate then dissolving it in KOH (51). KOH dissolves the keratin, allowing microscopic visualization of the fungal septate hyphae (51). Specimens can be further treated with stains such as Calcofluor White, Evans Blue, Gram, Giemsa, and India ink (52, 53).

KOH testing has 61% sensitivity and 95% specificity (51). It is cost-effective and can determine the presence of fungal organisms within an hour. However, it cannot specify the exact type of pathogenic organism (19). Retrieving adequate amount of specimen is critical in ensuring the success of KOH testing. To optimize accuracy, specimens should not be interpreted immediately after applying KOH, as it takes at least 15–30 min for the KOH to adequately dissolve the keratin (19). Overall, KOH testing is a quick and inexpensive diagnostic tool for confirming presence of fungi in nails, enabling clinicians to commence treatment for onychomycosis (Table 1).

Fungal culture can identify the specific pathogen subtype (53, 64). After removing debris from the nail plate, the subungual specimen is cultured in Sabouraoud dextrose agar (SDA) at 26–30°C for up to a month (18, 52). The media is treated with chloramphenicol and gentamicin to prevent bacteria from interfering with fungal growth (52). Laboratories often culture the subungual specimen in SDA with and without cycloheximide which prevents growth of NDMs. Repeat cultures are required to diagnose NDM onychomycosis, as NDMs are common contaminants of skin surfaces (18).

Fungal culture has 99% specificity, its pooled sensitivity is 56% (range 29–82%) according to a recent meta-analysis (51, 53). Sensitivity of fungal culture is largely dependent on the expertise of the testing center (65). Although specialized mycology centers may report fewer false negatives, most tertiary centers and general clinics observe low sensitivity rates. Clinical utility of fungal culture is further limited by delays in retrieving results (several weeks to months). Therefore, fungal culture is recommended when identifying the fungal organism is necessary (53).

Histopathological assessment involves examining the microscopic features of nail clipping specimens embedded in paraffin blocks (18, 52). To acquire an adequate nail sample for examination, at least 4 mm of the free edge of the nail plate should be retrieved using a dual-action or heavy-duty nail nipper (66). Samples can be transported to the laboratory in a dry container or in formaldehyde (67). Softening nail samples before routine processing with solutions such as chitin-softening agent, 4% phenol or 10% Tween 40, facilitates sectioning and thus optimizes the quality of sections (66, 67). After paraffin embedding and sectioning, stains highlight presence of fungi (Figure 1D). Using the periodic acid-Schiff (PAS) staining method, periodic acid oxidizes hydroxyl groups of the cell wall in spores, hyphae, pseudohyphae and yeasts into aldehydes (52, 53). These then react with Schiff to produce a red color (52, 53). Using the Grocott-Gomori methanamine silver (GMS) staining method, the chromic acid oxidizes the cell wall then reduces the methanamine silver nitrate into metallic silver to produce a dark brown color (53).

Histopathology of nail clippings is highly sensitive (84%) and specific (89%) (51). It can also be stored and retrospectively reviewed in refractory cases, as the paraffin embedding allows for long-term storage. Histopathology is however rather costly than KOH testing and unable to specify the exact subtype or viability of the causative organism (18, 19, 53, 68).

Nail dermoscopy (onychoscopy) is a non-invasive bedside tool that allows clinicians to visualize microscopic features of abnormal nails. Key dermoscopic features of distal and lateral subungual onychomycosis include a jagged proximal edge of the onycholytic area with spikes and longitudinal striae (Figures 1E,F) (30, 54). These features and a ruins aspect are associated with total dystrophic onychomycosis (54). Homogenous opacity is found in superficial onychomycosis (54).

Onychomycosis can also present with longitudinal melanonychia (fungal melanonychia). In such cases, white or yellow streaks, non-longitudinal homogenous pattern, yellow coloration, reverse triangular pattern, subungual hyperkeratosis, multicolor pattern and nail scaling are positive predictors of fungal melanonychia (Figures 1G,H) compared to nail matrix naevi or subungual melanomas (69). As nail dermoscopy is quick, non-invasive and inexpensive, it has the potential to help physicians identify onychomycosis by the bedside and decide whether to proceed to mycological assessment (70).

Reflectance confocal microscopy (RCM) is a real-time imaging tool that allows clinicians to observe features of abnormal nails at near-histologic resolution by the bedside. It uses a 830 nm laser in reflectance mode which divides the nail unit into thin horizontal sections for examination (56). Reflectance confocal microscopy of onychomycosis reveals networks of bright filamentous septate hyphae (56, 71, 72). It has 52.9–91.67% sensitivity and 57.58–90.2% specificity for detecting onychomycosis (55–57). Reflectance confocal microscopy is expensive and not subsidized in many countries, and further studies are needed to support its utility in the clinical setting (72). Therefore, it is yet to be integrated into clinical practice in many countries. In addition, it is difficult to assess thick nails with RCM as its depth of imaging is limited to ~200 μm.

Molecular assays including polymerase chain reactions (PCR), flow cytometry and mass spectrometry are advanced diagnostic tools that involve analysis of the fungal DNA causing onychomycosis.

Polymerase chain reactions involves amplifying the fungal DNA then detecting this with specialized fluorescent primers (73). Therefore, it can detect small amounts of pathogenic organisms within nails (74). Real-time PCR is the most frequently used form of PCR as it is relatively simple to conduct, can detect multiple organisms and has a low risk of contamination (73, 74). Various assays have been developed to facilitate commercial use of PCR technology when detecting dermatophytes, but these are not widely available (74). They report sensitivity of 85–100% and specificity of 94–100% (58–60).

Hafirassou et al. investigated usefulness of panfungal and pandermatophyte assays for real-time PCR compared to fungal culture in detecting onychomycosis (61). The pandermatophyte assay was 90% sensitive relative to culture. The panfungal assay showed a low sensitivity of 47% relative to culture due to multiple fungal species residing within diseased and healthy nails, as demonstrated by the candida and aspergillus assays. Further studies are warranted to examine PCR use in a real-world setting and reduce the risk of false positives.

Other molecular assay techniques include flow cytometry and mass spectrometry. Flow cytometry separates cells according to size, granulosity and presence of DNA and protein markers (18). Mass spectrometry involves charging chemical species and separating ions according to their mass-to-charge ratio (18, 53, 75). These techniques are experimental, requiring further research and development prior to integration into clinical practice.

Artificial intelligence (AI) has caused a recent paradigm shift in clinical medicine (76, 77). Its diagnostic performance has been shown to be comparable to that of specialist clinicians in identifying diabetic retinopathy and skin cancer (78, 79). Many dermatologists recognize that AI has great potential to improve dermatologic care (80, 81). Currently, it is mainly explored in the setting of skin cancer, ulcers, psoriasis and other inflammatory skin diseases, predicting skin-sensitizing substances, histopathological assessment, and gene expression profiling (82).

Developing a large database of photographs capturing a wide range of disease presentations is critical in AI training (82). Onychomycosis is an ideal candidate for AI as it is a common condition with minimal racial differences (63). Clinics worldwide can contribute to the database, and the AI technology would not be limited to specific populations.

In 2018, Han et al. developed AI for diagnosing onychomycosis (63). Two datasets, A1 (n = 49,567) and A2 (n = 3,741), were generated. A2 consisted of images of clinically diagnosed onychomycosis (63). For A1, standardized clinical images were generated by a hand and foot image selecting convolutional neural network (CNN), followed by a nail part extracting regional CNN (R-CNN) then a fine image CNN. Two CNN algorithms, ResNet-152 and VGG-19, were then trained to classify nails as onychomycosis or one their differential diagnoses using the two datasets. These algorithms were then validated against datasets (B1, B2, C, and D) of mycologically confirmed onychomycosis cases.

Algorithms trained with A1 were more accurate in diagnosing onychomycosis than those trained with A2. Moreover, the AI (two-layered feedforward neural networks computing the combined output of ResNet-152 and VGG-19) achieved test sensitivity/specificity/area under the curve values of (96.0/94.7/0.98), (82.7/96.7/0.95), (92.3/79.3/0.93), and (87.7/69.3/0.82) for the B1, B2, C, and D datasets, respectively. AI performed better than most of the 42 dermatologists. The Youden index (sensitivity + specificity−1) of AI which reflects its diagnostic accuracy was significantly higher than that of the dermatologists (p = 0.01) when evaluating the B1 and C datasets.

More recently, the group reported that the deep neural network at operating point achieved 70.2% sensitivity, 72.7% specificity and AUC of 0.75 in diagnosing onychomycosis in a prospective cohort of 90 patients (62). This was comparable to the performance of dermoscopy (sensitivity 72.7%, specificity 72.9%, AUC 0.755; p = 0.952) and experienced dermatologists (mean Youden index 0.230 ± 0.176; p = 0.667).

Although there is limited literature on this topic, these results showed that AI has the potential to assist clinicians decide whether they should test nails for onychomycosis. AI can also help improve telemedicine, as it can act as an accessible resource for patients to evaluate their own nails and, if needed, promptly initiate a formal clinician review. To further develop AI, the database should be optimized to include rigorously confirmed onychomycosis and various non-onychomycosis onychopathies that are evaluated with relevant mycological examinations. Finding a suitable means of distributing this technology to the public is also necessary. Finally, whilst integrating AI into clinical practice is important, dermatologists should use their clinical judgment to prevent overdiagnosis and excessive testing which would increase the burden on health care costs.