Intraoperative Imaging in Hepatopancreatobiliary Surgery

Simple Summary There is evidence that oncological radicality with complete tumor removal may improve the survival of patients undergoing surgery for hepatopancreatobiliary malignancies. However, the complexity of vital vascular structures close to or embedded within the pancreas and the liver may increase both the surgical difficulty and the risk of achieving a non-radical resection. Preoperative staging after neoadjuvant therapy is usually challenged by the inability of correctly used imaging methods to distinguish vital tumors from fibrosis. Additionally, the inability to define the exact tumor borders often transfers to the operating room as well. Recently, more research has focused on the development of novel intraoperative imaging modalities and targeted contrast agents to improve preoperative and intraoperative diagnostics. We review current advances made in preclinical research and discuss clinical possibilities and future perspectives, including the characteristics of the ideal contrast agent. Abstract Hepatopancreatobiliary surgery belongs to one of the most complex fields of general surgery. An intricate and vital anatomy is accompanied by difficult distinctions of tumors from fibrosis and inflammation; the identification of precise tumor margins; or small, even disappearing, lesions on currently available imaging. The routine implementation of ultrasound use shifted the possibilities in the operating room, yet more precision is necessary to achieve negative resection margins. Modalities utilizing fluorescent-compatible dyes have proven their role in hepatopancreatobiliary surgery, although this is not yet a routine practice, as there are many limitations. Modalities, such as photoacoustic imaging or 3D holograms, are emerging but are mostly limited to preclinical settings. There is a need to identify and develop an ideal contrast agent capable of differentiating between malignant and benign tissue and to report on the prognostic benefits of implemented intraoperative imaging in order to navigate clinical translation. This review focuses on existing and developing imaging modalities for intraoperative use, tailored to the needs of hepatopancreatobiliary cancers. We will also cover the application of these imaging techniques to theranostics to achieve combined diagnostic and therapeutic potential.

Perhaps advances made in the use of intraoperative US best potential for intraoperative imaging. Studies proving th the intraoperative utilization of indocyanine green (ICG) w sources are also emerging [8]. Currently, even the most comp using mini-invasive approaches, where the concept of reliabl tive NIR fluorescent imaging is even more appealing given t tion [9].
Owing to the limitations of current imaging and the e molecular-image-guided surgery in neurosurgery [10] and search is aimed at the development of precise and accessible r in HPB surgery. This review considers the advances and p method, defines the characteristics of the ideal contrast agen ties in clinical translation.

Methods
This is a narrative review synthesizing information from the terms "hepatopancreatobiliary", "intraoperative imaging niques", as well as primary evidence from known examples. Perhaps advances made in the use of intraoperative US in HPB surgery depict the best potential for intraoperative imaging. Studies proving the added prognostic value of the intraoperative utilization of indocyanine green (ICG) with near-infrared (NIR) light sources are also emerging [8]. Currently, even the most complex surgeries are performed using mini-invasive approaches, where the concept of reliable and functional intraoperative NIR fluorescent imaging is even more appealing given the absence of tactile evaluation [9].
Owing to the limitations of current imaging and the established clinical benefit of molecular-image-guided surgery in neurosurgery [10] and gynecology [11], more research is aimed at the development of precise and accessible real-time imaging modalities in HPB surgery. This review considers the advances and perspectives of each imaging method, defines the characteristics of the ideal contrast agent, and describes the difficulties in clinical translation.

Ultrasound
US is used transabdominally and endoscopically (EUS); it is also the only routinely used intraoperative imaging modality in HPB surgery (IOUS) (Figure 2). Techniques such as contrast enhancement (CEUS), doppler mode, or elastography have also been implemented [12]. US became a standard of care in any surgical facility performing liver surgery in order to image complex and individually variable areas of liver anatomy and to improve tumor detection in real time; ablative liver therapies are not feasible without the use of IOUS [13] (Figure 3). The profound benefits and advances made with the use of IOUS in liver surgery are well documented [14][15][16]; the "radical but conservative" strategy of G. Torzilli et al. in performing more complex resections using IOUS led to a shift from major liver resections to more precise and complex parenchyma-sparing surgeries [17]. Even today, in the era of high-resolution MR and CT, IOUS, or more precisely, contrast-enhanced intraoperative ultrasound (CEIOUS), have superior qualities [16,18], and their accuracy has a fundamental effect on surgical strategies in liver surgery [19,20]. Nevertheless, the intra-operative use of US requires expertise, with a notable learning curve of 40 pancreatic and 50 liver cases [21]. Other potential problems include a lack of precise visibility in subcapsular areas and lesions less than 5 mm [22,23]. Image quality is limited in the setting of liver disease and in the area where ablative therapy was previously used. Also, there is currently an insufficient amount of data on the outcomes of US use in laparoscopic surgery [24].

Optical Imaging with Fluorescent Agents
A well-studied method in HPB surgery uses optical imaging with near-infrared fluorescence (NIR) agents [25][26][27]. Infrared light has a better tissue penetration, of up to 1 cm, compared with visible light, and the minimal autofluorescence of tissues in the NIR spectrum improves the target-to-background ratio (TBR) [28]. As multispectral cameras with a fusion of RGB and NIR imaging are widely available, imaging is fast and includes no damaging radiation. NIR imaging was initially used to determine cardiac output and

Ultrasound
US is used transabdominally and endoscopically (EUS); it is also the only routinely used intraoperative imaging modality in HPB surgery (IOUS) (Figure 2). Techniques such as contrast enhancement (CEUS), doppler mode, or elastography have also been implemented [12]. US became a standard of care in any surgical facility performing liver surgery in order to image complex and individually variable areas of liver anatomy and to improve tumor detection in real time; ablative liver therapies are not feasible without the use of IOUS [13] (Figure 3). The profound benefits and advances made with the use of IOUS in liver surgery are well documented [14][15][16]; the "radical but conservative" strategy of G. Torzilli et al. in performing more complex resections using IOUS led to a shift from major liver resections to more precise and complex parenchyma-sparing surgeries [17]. Even today, in the era of high-resolution MR and CT, IOUS, or more precisely, contrast-enhanced intraoperative ultrasound (CEIOUS), have superior qualities [16,18], and their accuracy has a fundamental effect on surgical strategies in liver surgery [19,20]. Nevertheless, the intra-operative use of US requires expertise, with a notable learning curve of 40 pancreatic and 50 liver cases [21]. Other potential problems include a lack of precise visibility in subcapsular areas and lesions less than 5 mm [22,23]. Image quality is limited in the setting of liver disease and in the area where ablative therapy was previously used. Also, there is currently an insufficient amount of data on the outcomes of US use in laparoscopic surgery [24].

Optical Imaging with Fluorescent Agents
A well-studied method in HPB surgery uses optical imaging with near-infrared fluorescence (NIR) agents [25][26][27]. Infrared light has a better tissue penetration, of up to 1 cm, compared with visible light, and the minimal autofluorescence of tissues in the NIR spectrum improves the target-to-background ratio (TBR) [28]. As multispectral cameras with a fusion of RGB and NIR imaging are widely available, imaging is fast and includes no damaging radiation. NIR imaging was initially used to determine cardiac output and

Optical Imaging with Fluorescent Agents
A well-studied method in HPB surgery uses optical imaging with near-infrared fluorescence (NIR) agents [25][26][27]. Infrared light has a better tissue penetration, of up to 1 cm, compared with visible light, and the minimal autofluorescence of tissues in the NIR spectrum improves the target-to-background ratio (TBR) [28]. As multispectral cameras with a fusion of RGB and NIR imaging are widely available, imaging is fast and includes no damaging radiation. NIR imaging was initially used to determine cardiac output and hepatic function [29,30]. At present, its spectrum of use includes the assessment of tissue perfusion, ophthalmic angiography, sentinel lymph node evaluation, ureter visualization, and tumor mapping.
The only FDA-and European Medicines Agency-approved fluorescent dyes are indocyanine green [31] and methylene blue (MB) [32]. Both dyes have been implemented in HPB surgery as perfusion agents. MB is often used as a visible stain as it is rapidly recognizable to the naked eye. However, with an excitation peak of around 700 nm, background tissue shows more autofluorescence in fluorescent images. The potential use of MB in HPB surgery is based on the ability to mark anatomic liver parenchymal resection margins [33] and detect bile leaks after liver resections [34].
The use of ICG is more widespread because it has an excitation peak of approximately 800 nm. This peak allows for superior visualization compared with MB because of the elimination of background autofluorescence, although it is difficult to detect with the eye alone, in contrast to MB [35]. Consensus guidelines for the use of fluorescence imaging in hepatobiliary surgery were published in 2021 by a group of Asia-Pacific experts [36] focusing on ICG use in liver and biliary surgery. ICG can be injected directly into the biliary system or intravenously, which could be considered a safer route. After intravenous application and binding to plasma proteins, ICG is taken up by hepatocytes and then fully eliminated via bile. These kinetics make it effective for imaging the extrahepatic biliary anatomy [37][38][39], a particularly beneficial trait in cases with anomalous or intricate biliary anatomies. Despite the notable benefits, ICG is unable to visualize the intrahepatic biliary tree because of limited tissue penetration or precisely identify common bile duct stones, probably because of the high concentrations of dye in bile.
In tumor imaging, ICG has a sensitivity of up to 99% in identifying hepatocellular carcinoma (HCC) lesions [40], which has led to a more widespread use primarily in eastern countries, where the incidence of HCC is high. Yet, the average rate of false positivity is 10.5% [41]. Colorectal liver metastases (CRLM) tend to display a rim pattern of fluorescence; this may be due to the extensive central necrosis that is common in CRLM, or possibly, it may be due to distorted biliary extraction in immature hepatocytes surrounding the tumor tissue that, conversely, do not uptake the dye [42]. A newly published study on imaging superficial CRLM using ICG identified the ability to detect "disappearing lesions" after downstaging chemotherapy in 15 patients [43]. While the depth of the visualization is the main limit of ICG imaging, combining fluorescence and IOUS has proven to be superior to preoperative CT or IOUS alone in the detection of CRLMs ≤ 3 mm [23]. "Positive" and "negative" staining techniques have been described to help guide anatomic resection margins [44]. Apart from better intraoperative visualization and proof of concept data ( Figure 4), studies of patient outcomes have emerged, showing the superiority of ICGassisted liver surgeries for distinct indications [8,45].
The use of nonspecific dyes in pancreatic surgery has many limitations, and only a few reports exist. Healthy pancreases show a similar ICG uptake to tumor tissue, which creates ineffective tumor-to-background ratios (TBR) in carcinomas [46]. The way to potentially improve TBR with nonspecific dyes is using a second window technique (high-dose ICG injected intravenously 24 h prior to surgery). Given its enhanced permeability and retention effect (EPR), a TBR of 4.42 was achieved in 20 patients with malignant lesions enrolled in an open-label clinical trial [45,47]; three out of eight benign lesions were fluorescent as well.
Shirata et al. used intravenous ICG injected intraoperatively to image pancreatic ductal adenocarcinoma (PDAC), pancreatic NETs, and cystic neoplasms in 23 consecutive patients, proving the proposed hypothesis of visualization based on the hypovascularization (cystic lesions) or hypervascularization of lesions [48]. The reported TBR of the NETs was 1.99, with all the lesions successfully visualized intraoperatively (100% sensitivity). Cystic neoplasms showed lower fluorescent signals and a TBR of 0.54, with fair-to-poor visualization. The TBR of PDAC was of no statistical significance. These data are in line with other presented case reports; the COLPAN study reported 100% sensitivity in the laparoscopic ICG imaging of pancreatic NETs with a mean TBR of 7.7, peaking at 20 min after intravenous application [49]. Hutteman et al. reported no clear visualization of pancreatic carcinoma using intraoperative ICG, with a mean TBR of 1.22 ± 0.39 [46]. A recently published meta-analysis of ICG use in pancreatic surgery found six papers with a total of 64 lesions reported [50]; the overall sensitivity for all pancreatic lesions was 75% and a mean TBR of 1.22. The authors correctly pointed out the unknown effect of neoadjuvant therapy on visualization and the unknown effects of ICG use on recurrence-free and overall survival in pancreatic surgery.
While recommendations on ICG dosage and the timing of administration in liver and biliary surgery exist [36], there is a large inter-institutional discrepancy between dosages, ranging from a bolus of 2.5 mg to 5 mg/kg, and the timing of ICG administration in pancreatic tumor imaging [45,46]. Perhaps one of the main challenges is the fact that, while liver metastasis and the primary tumor site may benefit from the second window technique, peritoneal metastases are likely to be missed at the time of exploration and could require intraoperative ICG applications [51]. In contrast, the administration of ICG intraoperatively leads to unavoidable background fluorescence in the liver and biliary tract. As the current papers suggest, different pancreatic neoplasms will probably require different ICG administration timing. The timing of administration is currently being studied in a Japanese trial: jRCT1051180076.
Pancreatic surgery is burdened with high rates of R1 resections in up to 80% of cases [52,53] and difficulty in staging patients after neoadjuvant therapy, especially in evaluating the degree of treatment response within the tumor. The high rate of R1 resections is partially influenced by vascular invasion and a previous lack of widespread adoption of standardized pathologist protocols [54]. R0 resection is, however, an independent prognostic factor of survival after pancreatic resection [52], therefore, the ability to precisely image pancreatic tumors in the operating room would be an important milestone in pancreatic surgery. The era of minimally invasive surgery, where tactile feedback is omitted, calls for even more precise intraoperative imaging. New methods emerge with the aim of improving fluorescent imaging, e.g., the second near-infrared wavelength window NIR-II (1000-1700 nm), which suggests a higher sensitivity and detection of up to 8mm deep [55]. More studies evaluating the impact of fluorescence on intraoperative decision making are needed. Reporting was suggested by Lauwerends et al. [56].  [46]. A recently published meta-analysis of ICG use in pancreatic surgery found six papers with a total of 64 lesions reported [50]; the overall sensitivity for all pancreatic lesions was 75% and a mean TBR of 1. 22. The authors correctly pointed out the unknown effect of neoadjuvant therapy on visualization and the unknown effects of ICG use on recurrence-free and overall survival in pancreatic surgery. While recommendations on ICG dosage and the timing of administration in liver and biliary surgery exist [36], there is a large inter-institutional discrepancy between dosages, ranging from a bolus of 2.5 mg to 5 mg/kg, and the timing of ICG administration in pancreatic tumor imaging [45,46]. Perhaps one of the main challenges is the fact that, while liver metastasis and the primary tumor site may benefit from the second window technique, peritoneal metastases are likely to be missed at the time of exploration and could require intraoperative ICG applications [51]. In contrast, the administration of ICG intraoperatively leads to unavoidable background fluorescence in the liver and biliary tract. As the current papers suggest, different pancreatic neoplasms will probably require different ICG administration timing. The timing of administration is currently being studied in a Japanese trial: jRCT1051180076.
Pancreatic surgery is burdened with high rates of R1 resections in up to 80% of cases [52,53] and difficulty in staging patients after neoadjuvant therapy, especially in evaluating the degree of treatment response within the tumor. The high rate of R1 resections is partially influenced by vascular invasion and a previous lack of widespread adoption of standardized pathologist protocols [54]. R0 resection is, however, an independent prognostic factor of survival after pancreatic resection [52], therefore, the ability to precisely image pancreatic tumors in the operating room would be an important milestone in pancreatic surgery. The era of minimally invasive surgery, where tactile feedback is omitted, calls for even more precise intraoperative imaging. New methods emerge with the aim of improving fluorescent imaging, e.g., the second near-infrared wavelength window NIR-II (1000-1700 nm), which suggests a higher sensitivity and detection of up to 8mm deep [55]. More studies evaluating the impact of fluorescence on intraoperative decision making are needed. Reporting was suggested by Lauwerends et al. [56].

Optoacoustic-Based Imaging
Optoacoustic (OA), or photoacoustic imaging, is another emerging method with significant potential. The principle of "light-in, sound-out" is utilized [58]; the absorption of near-infrared light generates acoustic waves via the thermoelastic expansion of tissues

Optoacoustic-Based Imaging
Optoacoustic (OA), or photoacoustic imaging, is another emerging method with significant potential. The principle of "light-in, sound-out" is utilized [58]; the absorption of near-infrared light generates acoustic waves via the thermoelastic expansion of tissues that are detected using a computerized transducer. In biological tissues, sound scatters Cancers 2023, 15, 3694 6 of 17 1000 times slower than light, circumventing the resolution/imaging depth tradeoff that hinders the application of optical imaging [59]. In addition, as biological tissue is largely transparent to near-infrared light, imaging depths of approximately 5 cm are achievable with no sacrifices in resolution [60,61]. While optical imaging agents, i.e., ICG, methylene blue, and IR-800-CW dyes, are detectable using both optical imaging and optoacoustic imaging, these agents are optimized to generate the largest fluorescence signal and represent suboptimal optoacoustic agents. Optical imaging with agents such as ICG requires a path length double of that of optoacoustic imaging while also utilizing higher-energy light, which is more susceptible to photon absorption and scatter in biological tissue. This inherent disadvantage of fluorescence restricts imaging to an imaging depth of <6 mm to maintain suitable resolution [62]. However, optoacoustic signals have been detected as several centimeters. Optoacoustic imaging is often conducted in individual perpendicular planes, which may be combined following acquisition to establish a 3D tomographic image. This provides an advantage in HPB surgical imaging, as it allows for the visualization of tumors within the liver and pancreatic parenchyma, as they are often deeper than 6 mm from the anterior surface, and molecular imaging agents have been detected within the liver, i.e., high hemoglobin content, in preclinical models [63].
Both endogenous and exogenous contrast agents can be detected with optoacoustic imaging. While endogenous contrast agents eliminate the problems of utilizing dyes in tissues, they are usually weak reporters with non-unique spectra [64]. However, there is a current lack of contrast agents that are developed and optimized for optoacoustic imaging [62][63][64][65][66]. The unique optical properties, e.g., optical absorption as a function of wavelength, of different contrast agents lead to the ability of "unmixing" images to determine the location and concentration of each unique agent simultaneously. This unmixing capability in the context of optoacoustic imaging is called multispectral optoacoustic tomography (MSOT). Specifically, multiwavelength illumination is used to identify the absorption and emission spectra for every contrast agent in an area to differentiate between background and individual contrast agent signals in a murine model ( Figure 5). This capability could be utilized to provide information about a tumor's molecular features and/or structure's metabolism within tissues [67][68][69][70][71][72][73]. For example, a response to oxyhemoglobin may outline an artery in the vicinity of angiogenesis as a symptom of tumor progression. The development of OA agents can significantly improve the capabilities of in vivo imaging, such as identifying deep tumors following the administration of an OA agent in a murine model ( Figure 6). This is a major advantage when aiming to distinguish tumors from peri-tumoral fibrosis, necrosis, and inflammation. For example, benign and malignant gallbladder polyps were shown to have different OA signal intensities [74]. that are detected using a computerized transducer. In biological tissues, sound scatters 1000 times slower than light, circumventing the resolution/imaging depth tradeoff that hinders the application of optical imaging [59]. In addition, as biological tissue is largely transparent to near-infrared light, imaging depths of approximately 5 cm are achievable with no sacrifices in resolution [60,61]. While optical imaging agents, i.e., ICG, methylene blue, and IR-800-CW dyes, are detectable using both optical imaging and optoacoustic imaging, these agents are optimized to generate the largest fluorescence signal and represent suboptimal optoacoustic agents. Optical imaging with agents such as ICG requires a path length double of that of optoacoustic imaging while also utilizing higher-energy light, which is more susceptible to photon absorption and scatter in biological tissue. This inherent disadvantage of fluorescence restricts imaging to an imaging depth of <6 mm to maintain suitable resolution [62]. However, optoacoustic signals have been detected as several centimeters. Optoacoustic imaging is often conducted in individual perpendicular planes, which may be combined following acquisition to establish a 3D tomographic image. This provides an advantage in HPB surgical imaging, as it allows for the visualization of tumors within the liver and pancreatic parenchyma, as they are often deeper than 6 mm from the anterior surface, and molecular imaging agents have been detected within the liver, i.e., high hemoglobin content, in preclinical models [63]. Both endogenous and exogenous contrast agents can be detected with optoacoustic imaging. While endogenous contrast agents eliminate the problems of utilizing dyes in tissues, they are usually weak reporters with non-unique spectra [64]. However, there is a current lack of contrast agents that are developed and optimized for optoacoustic imaging [62][63][64][65][66]. The unique optical properties, e.g., optical absorption as a function of wavelength, of different contrast agents lead to the ability of "unmixing" images to determine the location and concentration of each unique agent simultaneously. This unmixing capability in the context of optoacoustic imaging is called multispectral optoacoustic tomography (MSOT). Specifically, multiwavelength illumination is used to identify the absorption and emission spectra for every contrast agent in an area to differentiate between background and individual contrast agent signals in a murine model ( Figure 5). This capability could be utilized to provide information about a tumor's molecular features and/or structure's metabolism within tissues [67][68][69][70][71][72][73]. For example, a response to oxyhemoglobin may outline an artery in the vicinity of angiogenesis as a symptom of tumor progression. The development of OA agents can significantly improve the capabilities of in vivo imaging, such as identifying deep tumors following the administration of an OA agent in a murine model ( Figure 6). This is a major advantage when aiming to distinguish tumors from peritumoral fibrosis, necrosis, and inflammation. For example, benign and malignant gallbladder polyps were shown to have different OA signal intensities [74].   The visualization of molecules within 200 uM blood vessels can be achieved wit the artifacts associated with vascularly dense tissues or the vessels themselves with [75][76][77]. In contrast, ICG loses fluorescence intensity after binding to proteins in b vessels when visualized with NIR [78], therefore limiting use in highly vascular tis which is often the case in HPB surgery. Furthermore, the ICG signal may be hind further by blood pooling within the surgical field secondary to blood loss, which i uncommon in hepatobiliary surgery [79]. Of importance, hemoglobin is one of the strong endogenous contrast agents that allow for the identification of microvas changes and tissue oxygenation using MSOT [58,80], and it has been successfully us monitor tumor responses to antiangiogenic agents in mouse models [81]. Exogenous trast agents used for OA imaging in preclinical mouse studies include organic dyes, as ICG, and nanoparticles, such as gold, silver, tungsten, iron oxide, and carbon nanot [64,70,71]. OAI showed a 3.7 TBR in the resected specimens of patients with PDAC w targeted with cetuximab-IRDye800, a NIR fluorescent agent that binds to the epide growth factor receptor [82]. This study, as well as several others [83], tried to com fluorescence with OAI in a multimodal approach; fluorescence had a sensitivity of 96 a positive margin was identified, and the separate differentiation of chronic pancre from PDAC was possible. However, the restricted depth of imaging and relatively level of false positive lesions remains the main limitation of ICG with OAI.
Besides frequent subcutaneously grafted tumor models, MSOT has been success used in orthotopic pancreatic cancer mouse models and created 3D images [61,75 which has driven further interest in this modality. However, while ICG is readily avai and a contrast agent known to most surgeons, more contrast agents that are specifi designed for optoacoustic imaging are needed for further clinical translation.

Photodynamic Imaging
The concept of photodynamic diagnosis, or photodynamic imaging (PDD), ut the application of photosensitizers that accumulate in targeted tissues that can the imaged upon excitation using specific infrared (IR) wavelengths. The photodynam fects of ICG and similar molecules are expected upon excitation with an infrared laser nm) during the surgical procedure. The visualization of molecules within 200 uM blood vessels can be achieved without the artifacts associated with vascularly dense tissues or the vessels themselves with NIR [75][76][77]. In contrast, ICG loses fluorescence intensity after binding to proteins in blood vessels when visualized with NIR [78], therefore limiting use in highly vascular tissues, which is often the case in HPB surgery. Furthermore, the ICG signal may be hindered further by blood pooling within the surgical field secondary to blood loss, which is not uncommon in hepatobiliary surgery [79]. Of importance, hemoglobin is one of the few strong endogenous contrast agents that allow for the identification of microvascular changes and tissue oxygenation using MSOT [58,80], and it has been successfully used to monitor tumor responses to antiangiogenic agents in mouse models [81]. Exogenous contrast agents used for OA imaging in preclinical mouse studies include organic dyes, such as ICG, and nanoparticles, such as gold, silver, tungsten, iron oxide, and carbon nanotubes [64,70,71]. OAI showed a 3.7 TBR in the resected specimens of patients with PDAC when targeted with cetuximab-IRDye800, a NIR fluorescent agent that binds to the epidermal growth factor receptor [82]. This study, as well as several others [83], tried to combine fluorescence with OAI in a multimodal approach; fluorescence had a sensitivity of 96.1%, a positive margin was identified, and the separate differentiation of chronic pancreatitis from PDAC was possible. However, the restricted depth of imaging and relatively high level of false positive lesions remains the main limitation of ICG with OAI.
Besides frequent subcutaneously grafted tumor models, MSOT has been successfully used in orthotopic pancreatic cancer mouse models and created 3D images [61,75,83], which has driven further interest in this modality. However, while ICG is readily available and a contrast agent known to most surgeons, more contrast agents that are specifically designed for optoacoustic imaging are needed for further clinical translation.

Photodynamic Imaging
The concept of photodynamic diagnosis, or photodynamic imaging (PDD), utilizes the application of photosensitizers that accumulate in targeted tissues that can then be imaged upon excitation using specific infrared (IR) wavelengths. The photodynamic effects of ICG and similar molecules are expected upon excitation with an infrared laser (805 nm) during the surgical procedure. Most of the data on photodynamic imaging are in preclinical proof-of-concept studies in small case series in various proposed areas of HPB surgery, notably in the successful imaging of HCC using ALA [84,85] and the determination of tumor margins for cholangiocarcinoma in a mouse model [86]. Another study tested laparoscopic photodynamic imaging to detect carcinomatosis in a staging laparoscopy for pancreatic cancer. Staging laparoscopies are a common practice for high-risk patients with HPB malignancy because of a lack of ability in preoperative imaging to precisely determine the early stages of peritoneal dissemination; however, this is currently solely dependent on the surgeon's experience. Indicating an area that needs more improvement, the study reported an increased rate of laparoscopic detection of peritoneal dissemination with fluorescence and even higher rates with spectrophotometry compared with white light in mouse models. Perhaps the greatest limitation of PDD in laparoscopy, especially for ALA, which has emission peaks in the spectrum of red-to-near-infrared light, is the low concentrations of fluorophores not limited to superficial layers. The study used spectrophotometry to overcome this limitation. The rate of ex vivo detection in human specimens using spectrophotometry was 63% [87]. New photosensitizing agents and nanocarriers are being tested [88,89]. There is also an interest in finding adjuvant agents to accelerate protoporphyrin accumulation in tumors, e.g., the currently active clinical trial NCT03467789, which is evaluating the effect of vitamin D.
PDD possesses potential characteristics for real-time imaging, and the concept of combining PDD and photodynamic therapy (PDT) to monitor tumor responses to PDT or treat resection margins is perhaps appealing. While studies on PDT alone are ongoing in HPB-e.g., the current clinical trial NCT03033225 for the treatment of unresectable pancreatic cancer using PDT [90]-the concept of combining PDD and PDT is not documented to the best of our knowledge. More studies are needed to address the real potential of photodynamic imaging and/or therapy in clinical settings.

Intraoperative 3D Imaging
Creating a three-dimensional image is a natural next step in today's digital era [91]. Three-dimensional modeling and its derived volumetric calculations are not uncommon in liver and biliary surgery. Data proving more precise preoperative planning in liver resection for HCC in 3D vs. 2D were recently published [92]. The idea of supporting intraoperative orientation led to the creation of 3D holograms, or 3D modeling, which faced the challenge of changing organ shapes during intraoperative manipulation. Today, "lastminute simulation" is mostly reported rather than intraoperative navigation, necessitating more research [93,94].

Characteristics of Ideal Tracer for Molecular Imaging
In nuclear medicine, the clinical value of labeled tracers used to image and monitor the metabolic activity of targeted tissues has been well documented [95,96]. Similar tracer characteristics are required for intraoperative imaging, e.g., pharmacokinetics, costs, and safety. While the fast elimination of contrast agents in nuclear medicine is desired, intraoperative imaging may favor extending the imaging window throughout the resection phase of surgery or the possibility of the repeated use of contrast agents. Creating a detectable signal with low amounts of contrast agents and a high TBR, including differentiation between tumors and surrounding fibrosis and inflammation, should be the main goals when developing probes for intraoperative imaging. Of note, MSOT can detect collagen as a measure of fibrosis.
While the passive targeting of nonspecific dyes (e.g., fluorescent imaging HCC based on EPR) is possible, it requires higher concentrations of contrast agents and generally has a lower labeling efficiency [97], as discussed previously. Therefore, active targeting seems to be the ideal design, perhaps even reaching microdosing levels. Most targeting contrast agents consist of a targeting component and a signaling component [98], although a dye that would be tissue-specific is an appealing concept. The strong affinity of contrast agents for target tissues and their fast elimination from blood and off-targeted tissues ultimately create a high TBR. Antibodies are one of the most widely tested targeting agents because of their natural targeting ability and a range of molecules that are already FDA-approved [99]. Clinical translation has been, however, hindered by several challenges (Table 1). Antibodies have a fairly large molecular size (150 kDa), which limits penetration within solid tumors that typically possess dense stroma, e.g., stroma in pancreatic cancer form 60-90% of the tissue [100]. Solid tumors also have various regions of hypoxia and acidity, which may affect the binding capacity of antibodies [101]. Furthermore, a notable intertumoral and interpatient variability in the overexpression of targeted receptors exists [102]. The high costs and administration several days before surgery must also be considered. Both anti-VEGFR (bevacizumab) [103] and anti-carcinoembryonic antigen (CEA) [57,104], conjugated to a fluorescent dye, showed suboptimal signal-to-background ratios, ranging from 1.4 to 2.1 when tested on patients with pancreatic and colorectal cancer. This prompted the early termination of the NCT 02743975 trial using bevacizumab. A study using anti-CEA antibodies for the detection of colorectal and pancreatic cancer liver metastasis in human patients showed a mean TBR of 1.7, with two false positives out of nineteen lesions [105]. Interestingly, the false-positive lesions had no CEA expression, indicating parallel mechanisms of antibody retention or fluorescence signal.
Given that an antibody's size may be a limiting factor, antibody fragments with similarly low immunogenic potential have emerged in preclinical testing. The variable kinetics of molecules of different sizes, binding affinities, and half-life influence further diffusivity, EPR effects, and the clearance and homogeneity of tracer distributions within tumors [106]. Antibodies generally compensate for slow extravasation with prolonged circulation due to minimal renal clearance (thus requiring administration several days ahead of surgery). While decreasing the size of a particle leads to increased vascular permeability (allowing for even same-day imaging [107]), this also leads to faster renal clearance, rapidly lowering the intravascular concentration needed for diffusion into tumorous tissue. A recent in vivo study on mouse models, including pancreatic ductal adenocarcinoma models, showed that, while imaging is faster using nanobodies, peak fluorescence is lower, contrary to the same antibody [108]. Nanobodies have high specificity, low background retention, and more homogenous tissue penetration [109]; furthermore, they are stable, making them ideal for further modifications and labeling, which leads to an unlimited range of possible particles to be developed and tested. A spectrum of nanoparticles, peptides, and coated particles was tested in vitro and in animal models [98,110] (Tables 2 and 3); the importance of the timing of administration and dosage was also demonstrated in preclinical settings [107], but a clinical translation is yet to be started (an overview of the advantages and disadvantages of ligands based on their sizes is in Table 4).
Another important consideration is whether a tracer has an extracellular or intracellular target; catalyzes an enzymatic reaction; or mediates a chemical transformation [98]. Solid tumors have significantly lower pH because of specific cancer cell metabolisms; thus, targeting acidic pH is another appealing option [111]. The mechanism is interesting since the effectivity of many other tracers might be oppositely or negatively influenced by acidity.

Conclusions and Future Directions
Significant progress in the treatment of HPB malignancies has been made over the last few decades, including advanced techniques for liver resections, improved systemic therapy, and perioperative care [138][139][140]. Complex liver resections, with the possibility of resecting up to 70% of the parenchyma, allow more patients to be considered for curative resections, and use a combination of resections and ablative techniques. Neither would be possible without the implementation and improvement of intraoperative imaging. Conversely, given that more patients on systemic therapy become eligible for surgical resection, preoperative staging and intraoperative orientation have become more grueling. Additionally, the pressure to perform more mini-invasive surgeries, including HPB surgeries, is palpable in Western countries. While in open surgeries, the surgeon's tactile sensation is one of the main guiding instruments; its absence in laparoscopic surgery, and even more so in robotic surgery, can hinder the orientation. Lastly, preoperative preparation is crucial for the guidance of the surgery, but delays between staging and surgery are sometimes unavoidable, necessitating further reevaluation in the operating room.
All of the above requires further research and development in intraoperative imaging. Preclinical data show endless amounts of possibilities, but the superiority of one contrast agent over the others is yet to be demonstrated. Combinations of several modalities have been tested. From a surgeon's perspective, intraoperative imaging needs to be precise and easily implemented and evaluated. The concept of the same-day administration of a contrast agent is appealing in the era of ERAS and fast-track surgeries. Importantly, the costs will have an important effect on the broad implementation of any new techniques. Multipurpose contrast agents aimed at preoperative, intraoperative, and possibly therapeutic use are not solely focused on qualities sought in the operating room, but the possibility of the broad clinical implementation of a drug is undeniably more appealing for development. The strenuous process of clinical translation should not discourage the further drive to improve intraoperative imaging given its potential and possible broad implementation. Connecting preclinical research with clinicians is critical in order to maximize the potential of research and resources. Vice versa, the surgical community needs to be open to helping facilitate the clinical translation of preclinical research.