The methods for acquiring the diagnostically relevant information can roughly be grouped in two types, they are either based on direct observation or imaging of the patient (in vivo), or they are based on analysis of extracted material from the patient (ex vivo). The former both suffer and benefit from being closely tied to the patient. Direct imaging of the oral cavity of the patient requires that the necessary imaging devices are available at the point of care, thereby ruling out expensive devices or such which require highly skilled personnel. On the other hand, the immediate availability of the patient allows increased flexibility to adjust imaging to the situation which simplifies the work and provides fast feedback. The methods which rely on extracted samples may, if resources are available, be analysed at the point of care, but such approaches also provide options where the sample is transported to a nearby laboratory for analysis. This may enable use of the same resources to cover a larger region, thereby allowing cost-efficient use of more expensive equipment. The sample-based approach may also allow self-sampling, which could be a suitable option for saliva and brush samples.

Sample-Based Methods

Methods based on imaging of samples (ex vivo) avoid the need for local access to instrumentation and staff to handle it, since samples can be transported to where technical resources and expertise are available. This is an important advantage for being able to offer equivalent health care at a reasonable cost also in less accessible rural areas.

Histology Using Tissue Biopsies

Histological analysis of tissue biopsy is the current ‘gold standard’ for clinical examination, it is based on tissue biopsy, slide preparation, histological staining and microscopic analysis (manual or computer assisted). Tissue specimens can be obtained by scalpel or a punch, depending on the appearance of the tissue or preference of the treating dentist/surgeon. Where technically possible, complete excision of diseased tissue is preferable. Otherwise, a ‘mapping procedure’ may be used to mark where incisional biopsies (Fig. 1) are obtained in areas with different reaction patterns. It is of importance that representative biopsies are obtained: that they are sufficiently large to include normal and suspicious tissue and the full depth of the mucosa, in order for the pathologist to give a diagnosis without requiring further samples. Tissue specimens should be submitted, in buffered formalin to ensure proper fixation, to a pathology laboratory for slicing, staining, and histopathological diagnosis. Based on the report, it is decided when the patient should be reviewed at the specialist dental care, considering the risk of tumour development. This may require repeated biopsies and may be warranted if a change in the clinical picture occurs. The value of histologic assessment is unclear because the description and grading of dysplasia are subjective [10]. However, the general opinion is that assessment of PMOD by clinical examination alone is challenging and in order to confirm a clinical diagnosis, tissue biopsy is required, especially in the event of lesions in multiple locations.

Fig. 1Cytology Using Brush Biopsies

Cytological analysis relies on microscopic examination of collected cells that have been spread out on a glass slide and stained. Cell samples are taken by a brush from PMOD or changes that are suspected to be cancer (Fig. 2). Samples are either smeared directly onto the glass or deposited into a small bottle of preservative liquid which is later spread onto the glass. Staining is most often performed according to Papanicolaou, which is a polychromatic substance involving multiple dyes that stain different components of the cell with different colours and intensities [17]. Currently, the consensus is that liquid-based cytology (LBC) provides an improvement on specimen adequacy, visualisation of cell morphology and diagnostic reproducibility [18]. LBC also simplifies cell collection due to easier handling and fewer transfer errors. A cytotechnologist, aided by a microscope, looks for any cells showing signs of malignant changes among approximately 100,000 cells in a sample, marking cells of interest. This process takes 10–15 min. As a following procedure, a cytopathologist analyses the samples and the indicated areas of interest to set the diagnosis.

Fig. 2

Liquid based cytology process cycle: a sampling with a brush, b deposition into a preservative liquid, c whole slide image cytology, d close-up of the cells, e extraction and annotation of the cells for AI analysis, and f mosaic of the extracted cells

The Cochrane meta study, summarising results from 20 oral cytology studies, reports sensitivity of 90% and specificity of 94% for the detection of oral cancer and PMOD [19]. Deurling et al. analysed and compared 1352 brush biopsies taken for cytological diagnosis with the same number of pathological-anatomical analyses of tissue biopsies of oral lesions. The results showed that LBC based on brush biopsies has high sensitivity (95%) and specificity (85%) and the authors concluded that brush samples are fully reliable for the diagnosis of neoplasia in the oral cavity [20]. Thus, with current technology, brush biopsies appear to be a reliable alternative to tissue biopsies for diagnosing cell changes [21].

No local anaesthesia or suturing are needed for the brush biopsy and acquiring the sample is relatively easy, in contrast to a tissue biopsy. Brush biopsies are less expensive, require less resources and are considerably less invasive for the patient. In ongoing studies, dental hygienists and dentists are being evaluated as performers of brush tests in general dental practice (GDP). Preliminary results show that 98% of acquired brush biopsies can be used for diagnostics [22, 23]. Accordingly, sampling could routinely be done in GDP and possibly also as self-sampling, similar to screening for cancer of the cervix, which has been successful [24].

Liquid BiopsiesSaliva-Based Biomarkers

Saliva contains various biomarkers that signal disease, and it has great potential for early cancer diagnosis with cost-effective and easy collection, storage, transport and processing. A systematic literature review [25] has evaluated studies regarding possible saliva markers for PMOD and oral cancer in both unstimulated and stimulated morning saliva. These studies showed that most of the possible markers are proteins. The authors’ conclusion was that a combination of biomarkers in saliva could be used as a screening tool to improve early detection and diagnostic safety of PMOD and cancer. However, methods of saliva collection, processing, storage methods and analysis must be standardised, prior to clinical implementation. Likewise, limit values for different salivary biomarkers must be defined for healthy individuals, and for individuals with PMOD or oral cancer. Additionally, metabolites are differentially expressed in saliva of subjects with oral cancer as compared to normal subjects [26]. The best possibility for the development of diagnosis using saliva-based biomarkers, should be based on a combination of biomarker panels after standardisation of the procedure, which could then be used as an effective screening tool to improve early detection and diagnostics [26]. The use of saliva-based biomarkers for cancer detection is still in early development, and no study has yet reported on the diagnostic accuracy of salivary sample analysis according to the Cochrane review [19].

Blood Samples

Circulating tumour DNA (ctDNA) are small pieces of DNA found in the bloodstream, originating from cancerous cells that have died. ctDNA can be differentiated from normal cell DNA by the presence of several alterations. Results so far suggest that ctDNA can detect cancer in close to 70% of the head and neck cancer cases, while the performance varies significantly, correlating with tumour type among all cancers and stages (between 8 and 100%). The larger the size and spread of the tumour, the higher the sensitivity of the method. A proof of concept in a prospective randomised screening trial shows that a better prognosis for patients diagnosed using ctDNA is still to be presented [27, 28].

Assisted Screening of the Oral Cavity

As an alternative or complement to sample-based diagnosis, the assisted screening of the oral cavity of the patient can potentially provide a quick, painless, and affordable examination that can be integrated into any routine examination. Among the various techniques, optical methods are easy to handle, safe, and already implemented within dentistry. In this section, fluorescence, Raman techniques and optical coherence tomography (OCT) are considered as methods with the highest potential for the purpose of this review.

Direct Inspection Under White Light

The initial and conventional oral examination is the visual inspection and palpation of the oral cavity by a specialist, based on WHO’s oral cancer diagnosis protocol [29]. This method allows direct feedback to the patient. Unfortunately, the reliability of such analysis is low, subjective, and requires skilled personnel. An alternative, which makes the optical analysis more objective, is to take digital white light photographs of the oral cavity. This allows remote human analysis, as well as computer-supported analysis. To further improve the contrast, different types of staining may be used, e.g. toluidine blue [30]. Light-based methods, including white-light and autofluorescence (AF), have shown a sensitivity of 87% and specificity of 50% for the detection of PMOD and oral cancer in meta-analysis of 23 studies [19].

Fluorescence Techniques

Tissue is composed of naturally fluorescing molecules (fluorophores) that, when excited with light in their specific absorption spectrum, emit a spectrum of light with longer wavelengths referred to as AF. Ultraviolet-violet light is commonly used as excitation light with fluorescence in the visible range that can be directly observed by the human eye. Pathological changes in the tissue lead to alterations in the tissue fluorophores and thereby AF, where cancerous tissue often emits AF with a lower intensity. The main contributors to autofluorescence in normal oral tissue are known to be epithelial fluorescence emitted from metabolic molecules originating from the cell cytoplasm, and stromal fluorescence originating from collagen [31]. Changes seen in AF are reported to be associated with loss of collagen in the stroma and the increased metabolic activity of the cells. In cancer, changes in the cells in the stroma precede changes in epithelial cells and invading tumours [31, 32].

The main advantages of AF-based examination methods are that no substance needs to be administered to the patient, no fixation or staining of samples is required, and the measurement techniques are relatively simple and inexpensive. Tiwari et al. and Lima et al. [33, 34] have reviewed the efficacy of clinically evaluated fluorescence imaging methods. The studies report different sensitivity and specificity values for AF imaging; however, they all show a higher sensitivity by combining AF imaging with conventional oral examination. Different commercial or research-grade devices have been implemented for visualisation of AF, where VELscope®, a handheld instrument for direct visualisation of AF using a blue excitation light, was used more extensively than other devices. Through the VELscope®, malignant tissue shows darker (no or low AF emission) compared to the surrounding healthy tissue that emits green AF (Fig. 3). However, it is reported that the excitation light and the pathology and anatomic site of the lesions affect the total AF [31] which is not considered in most of the clinical studies and is a probable reason for the variation in reported efficacies. Direct visualisation of AF is reported not to have a high diagnostic performance for distinguishing dysplasia or early cancer, which questions its reliability, especially in GDP, where experience with the system is usually lacking. Still, the WHO considers VELscope® to be an effective tool for the prevention of oral cancer [35].

Fig. 3

a Assisted screening using VELscope® used with safety goggles, b white light image of epithelial cancer in the right lip, and c fluorescence image of the same site. The person in (a) is not a real patient. b and c are re-published with permission from Tandläkartidningen [1]

With the addition of photosensitizing agents, i.e. exogenous fluorescent markers, the contrast of the target cancer tissue is enhanced, giving more reliable signals or images compared to AF. Among the considered agents, 5-aminolevulinic acid (5-ALA) is the most investigated substance for oral cancer, with the possibility of being topically applied to the lesion [34, 36]. 5-ALA is yet not approved for oral cancer detection; however, initial studies show a high sensitivity but a lower specificity (90–100% vs. 60–90%) [34] and dependence of the results on the anatomic site [37]. The use of either spectroscopy methods together with spectral analysis or advanced quantitative imaging could be beneficial for visualising both the AF and the exogenous marker’s fluorescence.

Raman Techniques

Information on the chemical composition of tissue can be extracted by measuring the vibrational modes of the molecules in the Raman scattered light spectrum. The method is label-free and specific, but the signals are weak, and interpretation of the optical spectra requires extensive knowledge and experience of the method, making the extraction of information highly dependent on the analysis and statistical methods. However, the interest for implementing the technology clinically and for surgical applications is rising, leading to adaptation of the technology for such circumstances [38, 39]. Systems used in the clinics are mostly spectroscopic, endoscopic [40] or in the form of a handheld imager [41]. Studies show promising results for oral cancer detection; however, the results are reported differently based on the measurement systems used and molecular content detected. A high sensitivity has been achieved for discrimination of oral cancer from normal tissue distinguished mainly at the protein, amino acid level, and with beta-carotene as the main molecular markers [42]. A clear distinction of OSCC and normal tissue is seen as an increase in protein and DNA, and a decrease in lipid content [43]. Difference in the water content of tissue has also shown to be useful for distinguishing tumour margins during oral cancer surgery [44]. Moreover, Raman imaging microscopy has shown positive results for grading of chondrogenic tumours based on the tissue histology slides [45]. However, Raman systems are currently relatively expensive for implementation in small clinics.

Optical Coherence Tomography

OCT is a method based on low-coherence interferometry that scans 2D images from which 3D images can be reconstructed. OCT is suitable for imaging of the tissue microstructure with micrometre resolution and allows an imaging depth of about 1–3 mm in the tissue. Application of OCT is well established within routine ophthalmology and cardiovascular examination. The potentials are numerous for other clinical imaging and screening applications as well, including oral cancer screening. There are, however, several obstacles for expansion of the OCT application to a routine screening: the system costs, the need for careful preclinical studies, non-trivial interpretation of the images and relatively bulky systems. As the technology develops, handier systems at affordable costs will become available, making OCT a suitable option for routine screening of oral lesions.

A sufficient number of studies have reported on the capability of OCT to distinguish between normal, dysplasia and malignant oral tissue [46, 47], suggesting that the method has potential for this application. Smith et al. [48] could with sensitivity and specificity > 93% distinguish cancer vs non-cancer (normal and dysplasia). Polarisation-sensitive OCT imaging (PS-OCT) has shown a higher image quality and resolution [49, 50] and very high accuracy in distinguishing benign from malignant (dysplasia or early cancer) oral lesions in mice [49]. Disregarding the image quality and details apparent in the image, interpretation of the images is often not trivial and therefore appropriate image analysis and classification methods are beneficial for facilitating the implementation of OCT in the clinic. Examples of OCT images from the oral mucosa obtained using a non-PS spectral domain system are shown in Fig. 4.

Fig. 4

OCT image of a ventral tongue mucosa and b vestibular mucosa in the lower lip. The images illustrate approximately 0.8 mm deep in the tissue with layers of epithelium (EP), basement membrane (BM) and lamina propia (LP). The layer thicknesses are clearly different between a and b