The ability to automatically perform tissue differentiation is the most crucial factor for the progress of tissue-specific laser surgery. First approaches in that area showed valuable results [30–33]. Moreover, optical methods for the discrimination of tissues seem to meet the needs of a remote feedback control system, as it does not require direct contact with the tissues. Prior findings from our workgroup showed the general ability to differentiate several soft and hard tissue types ex vivo by diffuse reflectance spectroscopy [17, 18]. However, performing laser ablation of biological tissues is known to cause multiple alterations, including a change of optical properties [20, 34]. A successful implementation of feedback-controlled laser surgery requires the differentiation of tissues under conditions of laser ablation. Hence, it was the aim of this study to investigate the general viability of optical tissue differentiation on physiological soft tissue types by diffuse reflectance spectroscopy under conditions of laser surgical intervention.
When regarding the results before laser ablation, a high discrimination performance was found for the soft tissue pairs investigated in this study, with a mean AUC of 0.97. The average classification performance for the identification of each single tissue type in comparison to all other tissue types investigated in this study (confusion matrix) turned out to be 86.5%. These results confirm the tissue differentiation performance on non-ablated ex vivo soft tissue types, which was found in a previous study of our work group . However, some differences were encountered: the AUC and sensitivity values for the tissue pair mucosa/nerve increased from 0.93 (AUC) and 0.92 (sens.) in the prior study to 1.00 (AUC&sens.) in the present study. The results of the tissue pair nerve/fat showed reduced results with an AUC of 75%, a specificity of 75%, a sensitivity of 65% and an classification error of 0.34. These parameters were found to be lower than in the prior study and are assumed to be due to the bio-morphological similarity of the two tissue types, as discussed further below . However, the differences in the tissue discrimination performance between the two studies may be due to the fact that we used a different spectrometer and another data pool with different statistical parameters in this study. We currently used a spectrometer with lower resolution, performing 385 measurements in a range from 350 nm to 650 nm with an inter-measurement point distance of 0.8 nm. Another set-up was used in our prior study, consisting of a spectrometer that operated 1150 measurements in the same range with a distance of 0.26 nm, e.g., resulting in a higher resolution . Moreover, the extend of the obtained data and the statistical analysis can further cause an aberration of the results. The statistical analysis in the current study was carried out based on 14 tissue samples per type of tissue, whereas in the prior study 12 tissue samples were used. Furthermore the statistical analysis, used in the current study, is based on a total of 21.000 measurements whereas in the prior study half of the data points where used (10.200 spectra). The further analysis is currently based on 10 Principal Components (prior study: 6 PCs).
Different types of lasers have been used for the purpose of tissue ablation. The excimer laser proved to allow only a low degree of tissue ablation per pulse [35, 36], whereas Nd:YAG lasers, Ho:YAG lasers as well as continuous wave and long-pulsed CO2 lasers allow for a sufficiently high ablation rate performing laser surgery. However, these lasers are meant to cause a heavy thermal impact with large carbonization zones . Short-pulsed (< 1 μs) CO2 lasers, ultra-short pulse lasers (Ti-Sapphire) and free running Er:YAG lasers provide sufficiently high ablation rates per pulse for rapidly processing bio-tissue as well [3, 30]. Due to the fact that the tissue response is highly dependent on the wavelength of the incident light , the Er:YAG laser is known to be specifically suitable for fast processing of both soft and hard tissue [38, 39]. Its wavelength (2.94 μm) is very close to the absorption maximum of water, the main chromophore of biological tissue, at 3 μm. The laser energy is absorbed in a very small volume of tissue, with precise removal of the irradiated tissue. It was demonstrated that Er:YAG-laser ablation comes with accurately limited lesion edges, low thermal damage, and corresponding undisturbed wound healing [40–42]. Due to these aspects the Er:YAG-laser was chosen for this study.
However, when exposing tissue to laser light, an alteration of bio-morphological properties occurs depending on wavelength, energy and irradiation time. The ablated tissue area is known to develop a carbonization zone, which scatters and absorbs incident light, followed by a zone of tissue denaturation . In ER:YAG Laser systems, these undesirable thermal effects are rather small but still had a detectable depth that was reported to be ≤ 5 up to 30 μm [40–42]. Even this small area of carbonization and denaturation may cause an alteration of the optical properties, followed by a modification of the resulting diffuse reflectance spectra . It is assumed that the mentioned effects of laser energy are tissue-specific. The laser impact on optical properties will vary according to the water content and the histological partition of each specific tissue type [21, 44]. Additionally, the specific vascularization of each tissue type may influence the impact of laser light on tissue. Hemoglobin is known to be one of the major absorbers in biological tissue. Lukionova et al. reported an irreversible alteration of erythrocytes after exposing them to laser energy .
However, the influence of Er:YAG-laser ablation did not heavily alter the optical differentiation performance between the tissue pairs in this study. In general, the results after laser ablation yielded a high differentiation quality with a mean AUC of 0.97. This average value was found to be similar to the average AUC value for all tissue pairs before laser ablation. The total classification error—calculated for all tissue types of this study—was 16.8%, which yield a slightly reduced classification performance after laser ablation of 3% compared to the performance before laser ablation. More specifically, promising results were observed for the differentiation of the tissue pair skin/fat, skin/muscle, fat/muscle, muscle/nerve, mucosa/fat, mucosa/skin and mucosa/nerve, with constant differentiation qualities of 1.0 (AUC). The differentiation of these tissue pairs is meant to be of importance concerning a guided laser surgical system that will follow the anatomical tissues layer by layer. Similar findings were observed concerning the specificity of tissue differentiation. However, a slight decline of sensitivity after laser treatment was detectable for the majority of tissue pairs—but all values still ranged above 88%.
Remarkably, the differentiation parameters increased for the tissue pair nerve/fat after laser ablation: The differentiation performance rose up to 85%, the sensitivity up to 88%, compared to the results before laser ablation (75%/65%). However, the identification of nerve tissue is a crucial step concerning laser surgery guidance—heavy damage to nerve tissue was demonstrated by several studies using high energy lasers. It was reported that nerve injury by lasers may lead to major sensory and/or motor impairment, affecting the patient’s function and aesthetics [6, 7, 9, 45, 46]. As assumed in a prior work, the biological similarity of nerve and fat is followed by a reduced potential of optical differentiation . Fat tissue is known to comprise large amounts of lipids like triglycerides, cholesterol and fatty acids . Referring to bio-morphological criteria of nervous tissue, every nerve fiber of a peripheral nerve that was used in the current work is surrounded by a thin layer of myelin called the epineurium. In turn, each nerve fiber bundle is surrounded by another myelin sheath called the perineurium. Both of these structures consist of up to 75% lipids, e.g. 25% cholesterol, 20% galactocerebroside, 5% galactosulfatide, 50% phospholipids . Hence, the tissue pair nerve/fat provides a high biological similarity, at least at the superficial layers of the samples. However, we used a constant set-up of 30 laser pulses, causing a histological ablation depth of 350 to 400 μm for all soft tissue types investigated in this study. For that reason, the surrounding myelin sheath may have been partly ablated by the laser, uncovering the bare nerve fibers. The axonal structure of nerve tissue is known to have a different biological structure compared to fat tissue, with a very low content of intracellular lipids. Hence, we assume that the laser-modified nerve structures without the surrounding myelin sheath provide a higher potential for optical differentiation, which is due to their biological diversity. On the other hand, it has to be taken into account that harming the myelin sheath already may alter nerve function and is therefore not desirable from a clinical point of view. This fact has to be considered when the results will be transferred to feedback-controlled nerve preservation during laser ablation.
An impairment of the differentiation performance was found for the tissue pair skin/nerve after laser ablation, with an AUC, a specifity and sensivity of 0.88. The underlying structure the epidermis is dominated by the connective tissue of the dermis. The epidermis and dermis of pigs provide a thickness of about 400–500 μm, followed by the subcutaneous tissue—similar to human skin [47, 48]. As mentioned above, the ablation depth was found to be between 350–400 μm in this study. Hence, it is assumed that the ablation of skin removed the epidermis and exposed the underlying dermal tissue components, i.e., collagen and elastic fibres (Figure 3). Laser ablation of nerve tissue removes parts of the myelin sheath but may additionally expose the cytoskeleton of peripheral nerve tissue which is composed of protein rich neurofilaments similar to connective tissue . Taking the results of the confusion matrix into account which shows that the classification error of ablated skin mainly occurred when comparing with nerve after laser ablation, it can be concluded that the optical properties of the connective tissue of the sub-epidermal tissue and the scaffold tissue of nerve show similar diffuse reflectance spectra, followed by a reduction of the differentiation performance due to their biological similarity. Considering normal body anatomy the differentiation of the tissue pair skin/nerve is not meant to be of major importance concerning a feedback system for laser guidance as major nerve branches do not run next to the skin. However, after trauma or cancer resections the anatomy will be heavily altered and the differentiation of skin and nerve may become a major issue for tissue specific laser ablation.
Compression of the tissue—when applying measurement techniques in direct contact with the tissue—is known to have an impact on optical properties in both in vivo and ex vivo studies [50–53]. Hence, we used remote techniques for applying the illumination light, acquiring the reflectance spectra and for laser tissue ablation, to avoid any bias by mechanical pressure on the tissue samples.
Complete darkness, which would avoid any bias from light sources others than the illumination light of the set-up, is not meant to meet the requirements of a surgical procedure on real patients. For our experiments, we have chosen a set-up with surrounding stray light to simulate an applicable environment for surgical procedures. To eliminate the influence of stray light, the diffuse reflectance spectra were adjusted by a mathematical algorithm .
In the current preliminary investigation we performed a total of 300 measurements per tissue type for each of the 14 tissue samples in order to show the general feasibility of tissue identification and differentiation by this method and further gain a data pool for each of the 5 tissue types. For the clinical in vivo implementation in a feedback system for laser surgery, it is necessary to establish a greater data pool as a base for tissue identification. Then, a minimal number of spectra can be recorded after each laser pulse to identify the tissue type using the trained LDA after the transformation dictated by the PCA.
The promising results of this study have to be considered with care concerning some limitations: First, the study was conducted on pigs’ tissue. Interspecies differences, e.g. human/pig, may show varying results when transferring this method to other animal models or humans. Second, ex vivo tissue is similar but not identical to in vivo tissue due to its decreasing moisture and blood content, the missing blood circulation and its progressing de-oxygenation of hemoglobin [55, 56]. Thus, further research is necessary to transfer the technique to in vivo tissue, taking into account the influence of circulation and oxygenation. Third, the ablation was performed with an Er:YAG-laser, which is known to cause minimal alterations to the surrounding tissue. As any laser interaction with biological tissue depends considerably on the wavelength, the results of this study may not be transferable to other laser types. Even though, this study demonstrated the general viability of tissue differentiation under the influence of laser ablation by diffuse reflectance spectroscopy.