Clinical use of mass spectrometry (imaging) for hard tissue analysis in abnormal fracture healing

Risk factors for non-union development

Although there is still no way to predict the outcome of the fracture healing process, there are some risk factors that can be related to non-union development [25]. One of the most important factors is inadequate vascularization, as disturbance of the blood supply can increase the risk of non-union development [2], [18], [25], [38], [41], [42], [44]. This can be caused by high-energy trauma resulting in extensive soft tissue damage, excessive displacement, and operative treatment at the fracture site [22], [25], [38], [40], [41], [44]. In addition, the amount of vascularization can be related to the location of the fracture, which can influence its healing process, as delayed unions and non-unions occur mainly in diaphyseal parts of the bone [2], [18], [22], [25], [44]. Bones with lower blood supply (e.g. tibia and femoral neck) show higher non-union rates compared to, for example, humeral fractures, which receive higher blood supply [19], [40], [44]. High-energy trauma and open fractures can result in disruption of the biological environment and/or comminuted bone, which both are risk factors [1], [25], [44]. Another important factor is instability and/or movement between fracture fragments, as the natural fracture healing process is disturbed [2], [18], [22], [38], [40], [41], [42]. Other mechanical causes can be displacement and a fracture gap, in which the main fracture fragments are not in contact and the fracture gap is too large to bridge [2], [18], [21], [22], [25], [38], [41], [42], [44]. A contributing risk factor is the quality of the surgical treatment in terms of stabilization and reduction of the fracture gap [18], [20], [22], [41]. In addition, the type of fracture will also influence the fracture healing process, as unstable complex, comminuted, or segmental fractures have a higher risk of non-union development [8], [20], [25], [44]. A (deep) infection resulting from contamination of an open fracture or as a complication of operation is a risk factor for non-union development [2], [8], [18], [21], [22], [25], [40], [41], [44]. Furthermore, there are some factors that increase the risk of non-union development, but are not the main reason for impaired healing [4], [18], [39]. These contributing factors include old age, osteoporosis, malnutrition, anemia, alcohol consumption, smoking, genetics, systematic disorders (like diabetes mellitus), multiple trauma or fractures, and certain medications (such as corticosteroids, anticoagulants, chemotherapeutic agents, nonsteroidal anti-inflammatory drugs (NSAIDs), and some antibiotics) [2], [4], [8], [18], [20], [21], [22], 25], [40], [41], [44]. Because of the complexity and interaction between these risks factors, early detection and accurate prediction of the outcome of the healing process remain a challenge [8], [20], [25], [45]. Nevertheless, risk factors and prediction models could allow for early identification and treatment initiation of non-unions [25].

Key molecules involved in the bone fracture healing

Multiple classes of molecules have been reported to play a role in the normal bone fracture healing process, and can be divided into three main classes: pro-inflammatory cytokines, growth factors, and metalloproteinases and angiogenic factors [1], [3], [6]. Several cell types interact in a coordinated way with these molecules in the fracture healing process, including immune cells, bone and cartilage forming primary cells, and muscle mesenchymal stem cells (MSCs) [2], [3], [6], [7], [17]. Any alteration in local and systemic regulatory molecules can have an effect on the cellular response, which may result in impaired healing [1], [5], [7], [17]. Table 1 gives an overview of the molecules that have been shown to be involved in the bone fracture healing process.

Despite the numerous molecules that have been discovered to contribute to the bone fracture healing process, there is still no complete understanding of which of these molecules are essential for the process [2], [6], [8]. No biomarkers for impaired healing have been used yet in a clinical setting for outcome prediction, as the consequences of the imbalance in the expression of different molecules are largely unknown [2], [8], [22], [45]. Thus, these molecules fail to serve as predictors of non-union [8], [45].

Diagnosis and prediction of non-unions

For the diagnosis of non-unions and evaluation of fracture healing, clinicians use mainly clinical, radiographic, and laboratory examinations [2], [8], [40], [41], [45]. However, the diagnosis of non-unions is often difficult [1], [22], [45]. Non-unions are diagnosed when they have already established [18], [20], [45]. Therefore, prediction of outcome and/or detection at the beginning of the fracture healing process would allow for earlier diagnosis, but this is lacking.

The clinical assessment contains documentation of the patient’s history and the bone fracture history, as well as a physical examination [18], [21], [22], [40], [41]. The physiological assessment of a non-union can show, among others, pain and tenderness at the fracture site, persistent motion, the inability of complete weight-bearing, and reduced motion in adjacent joints [18], [21], [22], [40], [41], [43].

The radiographic appearance of the fracture is still the gold standard used to confirm the non-union, to determine its classification, and to check the integrity of implants [18], [40], [41], [45]. General radiological signs of non-unions are the absence of bone bridging, persistent fracture lines, the presence of loose or broken implants, lack of progressive healing on serial radiographs, and hypertrophic or absent callus [20], [21], [22], [43]. Furthermore, multiple additional imaging modalities can be helpful, for example, computed tomography (CT), magnetic resonance imaging (MRI), and bone scans [18], [21], [22], [40], [43], [45].

Laboratory evaluations are used to define whether an infection is present as well as to discover metabolic or endocrine abnormalities that can explain the non-union [18], [22], [41], [43]. The laboratory evaluation should include a complete blood count (CBC), erythrocyte sedimentation rate (ESR), and C-reactive protein for the detection of infections as well as tests for the nutritional status when necessary [18], [21], [22], [40], [43]. Nevertheless, these tests have no direct linkage with the bone healing process. A few studies have been performed to analyze biological markers of fracture healing as predictors of impaired healing, for example, transforming growth factor-β (TGF-β) and bone morphogenetic proteins (BMPs), but they have not been applied in general clinical practice and sufficient evidence is lacking [1], [17], [45].

Despite the diagnostic options available, non-unions can only be diagnosed when they have already been established. Currently, it is not possible to predict the outcome of a bone fracture in the early phase, e.g. during the initial surgery. Early prediction would allow for adaption of treatment, if necessary, preventing non-union development and, thereby, avoiding additional surgeries and long burden for the patient.

Risk factors for PTOA development

Cartilage health depends on a combination of biological, mechanical, and structural factors [14], [15], [16]. Any disturbance in one of these factors, which cannot be compensated for by the other factors, will potentially result in the development of OA [15].

PTOA cannot be predicted, but some risk factors are known. Certain injuries often result in development of PTOA, including joint impact loading; anterior cruciate ligament (ACL); meniscal, ligament, or joint capsule tears and injuries; joint dislocations; and recurrent joint instability [10], [11], [12], [15], [26], [46]. One of the major risk factors for the development of PTOA is damage to the chondral surface during injury, which is related to the injury severity [11], [13], [23], [26]. In intra-articular fractures, the underlying subchondral bone is also affected, which increases the risk of PTOA development [11], [13], [15], [56]. The severity of the injury depends on the magnitude, orientation, and rate of the applied force and determines the absorbed energy and mechanical damage to cartilage and the surrounding structures [9], [12], [16], [24], [47]. The injury results in chondrocyte death and ECM damage, which contributes to cartilage loss [13], [16], [26], [47]. Nevertheless, Zhang et al. suggested that chondrocyte catabolism instead of death results in cartilage degeneration, although the chondrocytes eventually will die [48]. Although mechanical trauma is a major risk factor, it is insufficient to explain all the changes in the cartilage in the unaffected areas [15], [16]. Therefore, the severity of the injury is not the only factor that determines whether damaged cartilage will repair or undergo progressive degeneration [12]. After injury and/or surgery, the joint kinematics change, which results in abnormal mechanical loading and contact pressures at the articular cartilage [12], [23], [24], [56]. Thus, chronic joint instability, incongruity, and malalignment contribute to the development of PTOA via abnormal distribution of joint loadings, which is worsened by surface gaps and step-offs [10], [11], [12], [13], [15], [16], [24], [26], [56]. It has been shown that these increased contact stresses can be related to cartilage thinning [24]. In addition, the biological response to the joint injury, for example, bleeding and inflammation, can lead to the development of PTOA [13], [46], [47]. Other risk factors of PTOA development are the involved joint, certain bone fracture types and related dislocation, time since injury, genetics, obesity and other co-morbidities, age, gender, muscle weakening, and joint disuse or overuse [9], [11], [13], [15], [16], [26], [46], [47], [57]. The relative contribution of each factor is poorly understood as PTOA is considered a multifactorial condition, which makes optimal treatment planning challenging [13], [24], [46], [55].

Key molecules relevant for cartilage health

The molecular pathways involved in the development of PTOA are still mostly unknown, which is mainly due to the complexity of cartilage interaction with surrounding joint tissues and response to external factors [12], [15], [16]. Inflammatory cytokines and chemokines are two major molecular classes involved in the development of PTOA and interact with chondrocytes and other joint tissue cell types [13], [15], [46], [49]. PTOA development is affected by cytokines and chemokines in both positive and negative ways [46], [49]. Cartilage loss is caused by progressive cartilage damage when the catabolic pathway is more active than the anabolic pathway and, therefore, the chondrocytes are unable to restore the damage [12], [14], [15], [16], 56]. Table 1 gives an overview of the molecules that have been shown to contribute to cartilage loss and to be involved in intra-articular fracture repair.

A number of metabolic changes after a joint injury have been shown, but relating these changes to healthy or pathological joint adaptions is challenging and using them as predictors of clinical outcome is extremely difficult [11], [14], [15], [46], [47], [49]. Therefore, at the moment, no predictive biomarkers for PTOA are available [15], [16], [49].

Diagnosis and prediction of PTOA

Early detection of OA-related changes to the cartilage would allow for early intervention, which might prevent the development of end-stage OA [15]. The detection of PTOA (or OA) is very difficult and limited by the lack of symptoms in the early phase [12], [14], [15]. The variability in the time between injury and the development of PTOA symptoms depending on the severity of the injury and differences between patients is adding to this challenge [15], [16], [26]. In addition, joint injuries that can cause PTOA are difficult to diagnose and often overlooked in cases of other joint tissue injuries [9]. Therefore, PTOA is often diagnosed after the onset of symptoms based on clinical and radiographic findings, when irreversible damage to the articular cartilage has already occurred [15], [55]. This clearly illustrates the need for early detection and prediction before irreversible changes occur.

Radiography and CT scans are often used for the assessment of articular fracture reduction and diagnosis of OA [13], [58]. With radiography, OA is diagnosed based on the bone (sclerosis and osteophytes) and cartilage changes as well as reduction of the joint space [15], [49], [51], [52]. CT scans have a higher sensitivity than radiography for articular cartilage assessment, with improved detection of surface gaps, step-offs, and other surface disruptions allowing for estimation of the absorbed energy during injury [13], [16], [24], [58]. Double-contrast CT can be used for the measurement of cartilage thickness to determine areas of cartilage thinning [24]. MRI can be used for the measurement of, among others, cartilage thickness and the detection of cartilage and bone marrow lesions [11], [24], [26]. Nevertheless, there is currently no method to measure the extent of cell and matrix damage [9], [12]. MRI has been proposed for studying the amount of damage for injuries without visible mechanical disruption [9].

Biomarkers that can be measured in, for example, blood, urine, or synovial fluid could contribute to early detection of OA, but they are currently not used for diagnosis because of lack of specificity [15], [49], [55]. Different molecules have been investigated, for example, ECM molecules and inflammation markers, but they have not been applied in general clinical practice, as there is only limited consensus about their predictive value [15], [16], [49], [55].

PTOA can only be diagnosed when irreversible changes to the articular cartilage have occurred, despite the current diagnostic options. Currently, it is not possible to predict the outcome of injury-related cartilage damage in the early phase, e.g. during the initial surgery. Early prediction would allow for adaption of treatment, if necessary, delaying or even preventing PTOA development and, thereby, avoiding additional surgeries and long burden for the patient.

MSI of bone

Most of the research performed with MSI on bone has been performed with time-of-flight (ToF) SIMS, but more recent literature reports the application of MALDI-MSI. The limited number of studies applying MS proteomics methods on bone did not use MSI, as they were either applied to cell cultures and ECM, or blood products [68]. One of these studies was focused on bone fracture healing, in which Grgurevic et al. showed that the detected proteins were involved in cell growth and proliferation, transport, and coagulation using liquid chromatography-MS [69].

ToF-SIMS imaging of bone

One of the application fields of ToF-SIMS is to study bone mineralization in relation to, among others, fracture healing, bone-implant interaction, and disturbances of minerals (see Figure 3A) [71]. This is mainly due to the high sensitivity of SIMS for elements and elemental and molecular distributions of, for example, Ca, P, PO₂, and PO₃. In addition, different fragments of hydroxyapatite and collagen have been reported [61], [62], [65], [71], [72].

Figure 3:

Examples of application of mass spectrometry for the direct analysis of hard tissues, namely ToF-SIMS on bone, MALDI-MSI on cartilage, and CO₂ laser coupled to REIMS on bone and cartilage.

(A) ToF-SIMS image of healthy sheep vertebra showing Ca⁺ (inorganic compound), C₄H₈N⁺ (organic compound), and overlay of the selected m/z values. ToF-SIMS measurement was performed with a Bi₃⁺-primary ion beam and a lateral resolution of 10 μm. Reprinted with permission from Müller et al. [65]. (B) MALDI-MSI image of osteoarthritic cartilage showing an overlay image of SM (d18:1_16:0) in the damaged superficial part of the tissue, PC (18:0_18:1) specific for chondrocytes, and DMPE (34:1) in the matrix of the tissue. The arrows point at chondrocyte pellets. MALDI-MSI image was taken at a spatial resolution of 15 μm. Reprinted with permission from Barré et al. [70]. (C) Mass spectra for the mass range m/z 500–1200 obtained from a human femoral head with a CO₂ laser (continuous wave with a laser power of 20 W with a pulse duration of 0.20 s) coupled to REIMS from cartilage (top) and cortical bone (bottom).

Different studies have used ToF-SIMS on bone tissue to study its elemental and molecular composition. In osteoporotic animal models, ToF-SIMS was used to show decreasing trabecular number and thickness, and declining cortical bone based on calcium and collagen distribution [61], [65]. For quantification of the calcium content, calcium hydroxyapatite scaffolds have been used in a rat model with encouraging results, although quantification was affected by the surface properties of the standards [62]. Furthermore, the bone-titanium implant interaction has been studied with ToF-SIMS, showing higher mineralization around the implant after a longer time period and specific molecules for bone tissue, the implant, and the interface [60], [72]. Multiple studies have researched the effect of a strontium-enriched calcium phosphate cement on fracture healing on healthy and osteoporotic bone [64], [73], [74], [75]. It was shown that this implant improved bone formation and strontium could be found in the newly formed bone as well as pre-existing tissue as strontium diffused up to 6 mm from the implant [64], [73], [74], [75].

MSI of cartilage

Several MSI studies have been performed on cartilage, but most studies use electrospray ionization or liquid chromatogram MS/MS. These methods have been used to study mainly proteins, peptides, and lipids in cartilage, other joint tissues, serum, and synovial fluid. The MSI studies mainly employed ToF-SIMS and MALDI for the analysis of cartilage and other joint tissues [31].

ToF-SIMS imaging of cartilage

With ToF-SIMS, osteoarthritic and healthy human cartilage were studied and showed that these two types of cartilage could be distinguished by specific peaks [76]. Furthermore, colocalization of cholesterol and other lipids in lipid droplets was seen in the superficial area of the cartilage, with a similar localization of fatty acids in osteoarthritic cartilage [76]. Calcium and phosphate ions were observed in the areas surrounding chondrocytes in osteoarthritic cartilage [76].

MALDI-MSI of cartilage

In a study on peptides and proteins in human healthy and osteoarthritic cartilage using MALDI-MSI, osteoarthritic-specific peptides and proteins were found with a higher intensity in the deep cartilage [30]. Distinctive peptides for OA included fibronectin, cartilage oligomeric matrix protein, and fibromodulin [30]. The peptides of young, old, and osteoarthritic equine cartilage showed different profiles between the groups using MALDI-MSI [77]. Increased intensities of fibronectin peptides and decreased intensities of fibromodulin peptides were seen in osteoarthritic cartilage, while collectin-43 and cartilage oligomeric matrix protein were age-specific [77]. Recently, in a MALDI-MSI study about the lipid distribution in human osteoarthritic cartilage, the superficial layer and the cartilage layer could be distinguished by specific molecules, but more interestingly a phosphatidylcholine was found to be specific for chondrocyte cluster cells (see Figure 3B) [70]. Furthermore, N-glycans have been studied in cartilage and bone marrow with MALDI-MSI, as the subchondral bone is affected by OA by bone marrow lesions [78]. Different N-glycans showed different distributions over the cartilage and bone marrow, with a high mannose N-glycan being prominent in osteoarthritic cartilage and a disialylated biantennary complex glycan was prominent in bone marrow of the patient with bone marrow lesion stage 1 [78]. The distribution of triamcinolone acetonide (TAA) in cartilage was shown for sample submerged in TAA with MALDI-MSI using an essential derivatization step [79].

MALDI-MSI has already been applied to cartilage to study a wide range of molecules, including lipids and peptides, although most research is applied on late osteoarthritic tissue. Applying this technique to impaired healing research on cartilage could contribute to a better understanding of, among others, the early phase of PTOA.

Electrocautery

Electrocautery combined with rapid evaporative ionization MS (REIMS) for rapid, in situ profiling has been mainly used for cancer research and tumor margin assessment (see Supplementary Table 1) [32], [35], [80], [83]. The coupling of electrocautery with REIMS is often referred to as the intelligent Knife (iKnife) [32], [35], [80], [83]. With this approach, electrocautery is used for cauterizing and cutting of the tissue resulting in aerosol-rich smoke containing aerosolized molecules, which is transported to the mass spectrometer via a tube using a Venturi pump [28], [32], [83], [84], [85], [86]. The basis for this technique is rapid thermal evaporation of biological tissue by, for example, electrocautery resulting in the formation of gaseous ions [29], [32], [80], [82], [83], [84], [85]. Several research groups have shown that different tissues show different and characteristic mass spectra of specific lipid patterns allowing for differentiation between healthy and cancer tissue, different tumor grades, and tumor necrosis [32], [82], [83], [84], [85], [86]. In general, these studies show high accuracy (range 74.7–97%), sensitivity (range 78.6–97.7%), and specificity (range 89.7–100%) (see Supplementary Table 1) [32], [83], [84], [85], [86], [87]. Although most of the research is currently performed ex vivo, the usage of the electrocautery-REIMS has also been tested in vivo [32], [84], [85], [86].

In a similar manner, Balog et al. used electrosurgical dissection via an endoscopic polypectomy snare connected to REIMS for the classification of gastrointestinal tract tissue [87]. This technique was able to differentiate between healthy intestinal wall tissue, colorectal cancer, and adenomatous polyps [87].

Laser-based methods

Laser ablation of tissues has shown to result in characteristic ion patterns and combined with MS could be used for the near real-time in vivo identification of tissues [81], [88]. For this, both ultraviolet and infrared lasers have been used, whereby the CO₂ laser is the most common employed infrared laser in surgical procedures [35], [80], [81], [88].

Schäfer et al. used laser desorption ionization mass spectrometry (LDI-MS) for the ex vivo analyses of biological tissues comparing different ultraviolet and infrared lasers [88]. The reproducibility of the spectra obtained with the CO₂ laser was higher than of those obtained with the ultraviolet (Nd:YAG) laser, resulting in high accuracies for tissue identification with the CO₂ laser (see Supplementary Table 1) [88]. The mass spectra generated with the CO₂ laser differed considerably between the two modes of the laser (continuous wave and pulsed mode) and was associated with the different laser powers [88]. This indicates that laser-tissue interaction differs per mode and, therefore, should be optimized per tissue type and clinical application. The obtained spectra showed high similarity with spectra obtained with electrocautery, suggesting similar ion formation mechanisms, which was confirmed by Balog et al. [84], [88]. Genangeli et al. showed that the quality and reproducibility of the mass spectra obtained with the CO₂ laser coupled to REIMS were higher compared to electrocautery [89]. In addition, the CO₂ laser enabled to generate a lipid-rich signal from hard tissues, i.e. bone, bone marrow, and cartilage [89]. An example of the mass spectra obtained from hard tissues is provided in Figure 3C. The use of the CO₂ laser has been tested in vivo [88].

Fatou et al. developed an in vivo near real-time analysis technique, named the SpiderMass, which uses resonant infrared laser ablation (RIR-LA) for the production of mainly lipid ions [28], [29], [90]. Different signals could be obtained for normal, cancer, and necrotic tissue and high correct classification rates have been shown (see Supplementary Table 1) [28], [90]. The minimal invasiveness of this technique was shown in vivo on human skin and canine tissue, although the technique can result in local dehydration of the tissue [28], [90].

Woolman et al. developed a handheld sampling probe using picosecond infrared laser (PIRL) ablation MS for the sampling of tissue with minimal thermal damage and tissue removal [91]. In different studies, m/z values in the signal were identified, which belonged to fatty acids, ceramides, and (fragments of) phospholipids, including phosphatidic acids, phosphatidylethanolamines, and phosphatidylcholines [91], [92], [93]. Ex vivo studies using PIRL showed good separation of tissue types and medulloblastoma subgroups with high classification rates (see Supplementary Table 1) [91], [92], [93]. PIRL combined to MS has not been used in vivo yet.

Other techniques

Schäfer et al. combined a cavitron ultrasonic surgical aspirator (CUSA), which uses ultrasound for the disintegration of tissue, to sonic-spray ionization MS [81], [94]. A clear differentiation could be made between different tissues and tumor types (see Supplementary Table 1) [94]. With this technique, besides lipid-related ions, the spectra also contained ions of metabolic constituents, carbohydrates, and peptides [94]. The combination of CUSA with MS has not been used in vivo yet, but is expected to give similar results [94].

The MasSpec Pen extracts molecules (metabolites, lipids, and proteins) from the tissue via a water droplet, which is analyzed in a mass spectrometer [95], [96]. This device allows for the rapid and nondestructive analysis of tissue, independently of the tissue shape and rigidity [95]. The MasSpec Pen has been shown to be able to differentiate between different tissue types and cancer and normal tissue with high accuracy, sensitivity, and specificity (see Supplementary Table 1) [95], [96]. No visible tissue damage was observed in the in vivo testing of the MasSpec Pen [95].

Implementation and limitations

Although each potential in vivo MS technique has its own application field, only a CO₂ laser coupled to REIMS has recently been used on bone and cartilage (see Table 2) [89]. The implementation and limitations of the different techniques will be discussed to explore which technique could be used for the prediction of non-unions and PTOA.

The main advantages of these ambient techniques are the possibility to analyze intact samples in vivo in their native environment with the near real-time feedback [37], [80], [84], [91], [97]. Furthermore, multiple techniques, including electrocautery and the CO₂ laser, use surgically approved handpieces for the ablation and ionization of tissue, which allows for easier implementation in the operating room [80], [81], [84], [89]. In contrast, other techniques, like the SpiderMass and the MasSpec Pen, are not medically approved, which will extend the implementation process in the operating room in the near future.

The amount of unwanted tissue damage generated by these in vivo techniques varies depending on the approach used. The main disadvantage of the electrocautery is the thermal and mechanical damage to healthy surrounding tissue, which is unavoidable during surgery but additional damage is an unwanted side effect [28], [35], [37], [90], [95]. Consequently, the spatial resolution of this method is low and defined by the size of the blade used [37], [98]. Furthermore, this tissue damage complicates the use on healthy tissues and the histopathological validation of tissues sampled with electrocautery [32], [37], [83], [92]. Other methods prevent extensive tissue damage, with limited damage for different ablation-based methods (infrared and ultraviolet lasers, CUSA, and PIRL) and almost no damage for other techniques (SpiderMass and MasSpec Pen) [28], [37], [89], [90], [91], [92], [93], [94], [95], [96], [97]. For CO₂ laser and PIRL-based techniques, another advantage compared to electrocautery is the minimal thermal spread as well as the fact that these probes interact mostly with the top tissue surface [80], [81], [91], [92], [93]. Compared to the electrocautery, CO₂ laser-based techniques allow for the more precise sampling of the tissue with less tissue damage [88], [89]. The higher reproducibility of the CO₂ laser compared to other lasers allows for more frequent sampling during surgical interventions to improve the efficiency of analysis and decrease the invasiveness of intentional sampling [88].

In addition, different techniques have their own disadvantages. As already mentioned, electrocautery results in extensive tissue damage [28], [35], [37], [90], [95]. The possible contamination with tissue debris limits the usage of the SpiderMass, which is the result of the close contact of the device and the tissue [90]. In the PIRL-based technique, a separate plume collection tube is used for the transfer of the molecules to the mass spectrometer [91], [92], [93]. The need for this additional tube might limit the translation to the operating room because of practical reasons. For the MasSpec Pen, the disposable tip can be cleaned by automated flushing of the system or needs to be replaced [95]. Due to the contact of the tip with the tissue, contamination can occur when the tip is only flushed while replacing the tip every time is clinically unpractical [37].

Until now, implementation of in vivo methods in the clinic has been limited, which can partly be explained by the lack of multicenter and validation studies performed in the operating room [81]. In addition, the implementation is constrained by the need for the development of large data sets (which can take years) to improve the accuracy of the built tissue classification models as well as the need for validated and quick data processing workflows [36], [80], [81]. These databases are generated based on ex vivo measurements of patient material and associated clinical information. The potential for intra-operative MS techniques in the improvement of patient care is foreseen as revolutionary, especially for those techniques using existing surgical handpieces. For the prediction of non-unions and PTOA, we propose the usage of a CO₂ laser coupled to REIMS, as it can be used on hard tissues, employs an existing surgical handpiece, and has low invasiveness. This in vivo MS technique can be used intra-operatively in patients for the prediction of outcome of fracture healing by sampling the bone and/or cartilage without tissue removal.