Introduction

The quest for advanced intraoperative diagnostic and therapeutic tools has spanned centuries, but only with the recent surge in Big Data and Artificial Intelligence has a truly effective solution become feasible1.

Recent decades have seen significant advances in optical technologies for surgery2. Initially, efforts focused on bridging the gap between human vision and imaging systems, progressing from Standard Definition (SD) to High Definition (HD), 4 K, and 3D imaging. As technology evolved, the focus shifted toward surpassing human vision— achieving microscopic visualization, and revelation of the previously invisible. In other words, the surgical augmented eye3.

This evolution has given rise to surgical optomics4, an emerging interdisciplinary field that pairs optical imaging technology with advanced computational methods to provide relevant, real-time, quantitative, and functional information during surgical procedures. “Omics” refers to the comprehensive characterization and quantification of biological systems5. Surgical optomics applies this approach to enable surgeons to quantitatively assess tissue composition and pathophysiological properties (e.g., perfusion, hydration), at both the micro- and macroscopic level, with unprecedented precision. In providing real-time access to previously unattainable data, Surgical Optomics has the potential to enhance intraoperative decision-making, yet the scope, applications and limitations of surgical optomics remain underdefined. We aim to delineate this emerging field by exploring the most prevalent and widely available surgical optical imaging techniques, as well as those with significant potential for future advancements (Fig. 1).

Fig. 1
figure 1

Schematic representation of key technologies in Surgical Optomics.

Full size image

The electromagnetic spectrum – how different wavelengths are leveraged for surgical use

The electromagnetic spectrum refers to the full range of electromagnetic radiation characterized by wavelength or frequency. The wavelengths visible to humans (400–780 nm) represent only a small portion of this continuum, which encompasses every type of energy wave6. From long, slow, and low-energy radio waves to ultra-short, fast, high-energy gamma rays, the spectrum contains seven broad classes based on how these waves behave when interacting with matter7 (Fig. 2).

Fig. 2: Diagram of the electromagnetic spectrum.
figure 2

Orange stripe display wavelengths, which increase in length towards the left. Blue stripe display frequency, which is inversely proportional to wavelength (decreasing towards the left). Surgical Optomics primarily focuses on wavelengths within the range between infrared and ultraviolet, including the visible spectrum.

Full size image

Although every class in the electromagnetic spectrum has been applied in the medical field, Surgical Optomics focuses on wavelengths within the ultraviolet, visible, and infrared light ranges (approximately from 300 to 3000 nm), offering a cost-effective, high-resolution imaging without the use of ionizing radiation8.

By studying the interaction between light and matter, it is possible to understand how different electromagnetic wavelengths behave when propagating through various biological tissues — how each one is absorbed, reflected, and refracted in distinct settings (Fig. 3). This enables the characterization of tissue molecules and facilitates objective clinical conclusions9.

Fig. 3: Schematic of light tissue interactions.
figure 3

Absorption occurs when light is absorbed by tissue molecules and converted into other forms of energy (e.g., heat). Reflection refers to light bouncing specularly (at a single angle) off the surface. Scattering involves deflection of light in various directions by tissue structures (e.g., organelles and fibers) which may also alter the photon’s wavelength. Fluorescence is the re-emission of absorbed light at a different wavelength. Transmission describes the process of photons passing through tissue without being dissipated. These phenomena occur simultaneously and at varying degrees depending on tissue properties and light characteristics.

Full size image

This study examines the most prevalent and promising surgical optical imaging techniques, emphasizing their current applications and potential for future advancements. The analysis is categorized into two primary approaches: exogenous fluorescence, represented by fluorescence-guided surgery, which utilizes externally administered contrast agents for enhanced visualization10, and endogenous fluorescence, including hyperspectral imaging and spatial frequency domain imaging (SSOP), which exploit intrinsic tissue properties to provide functional and structural insights9. This framework aims to elucidate the advantages, limitations, and future prospects of these modalities in advancing surgical precision and outcomes.

Exogenous fluorescence

Principle

Fluorophores are chemicals that can absorb light at a specific wavelength and re-emit it at a typically higher (lower energy) wavelength. Fluorescent imaging exploits this physical phenomenon by utilizing spectrally resolved light sources that excite the injected exogenous fluorescent dye, which in turn emits a signal. This signal is detected by cameras equipped with light-collection filters and processed through specialized hardware and software, ultimately displaying the fluorescent data, potentially superimposed on a white-light illuminated image.

By administering these fluorophores in Fluorescence Guided Surgery (FIGS), specific structures that would otherwise be indiscernible to the naked eye can be set apart from their surroundings, giving the surgeon critical support for adequate real-time decision-making11.

Each fluorophore used in FIGS has specific optical and pharmacokinetic characteristics that allow for its use in different situations. Among the clinically available fluorophores, those that emit light in the Near Infrared (NIR) wavelength spectrum (700–1700 nm) are of special interest for translation into surgical practice because the signal at this wavelength can penetrate deeper into tissues with less photon absorption, scattering, or auto-fluorescence, resulting in high-contrast, three-dimensional data12,13,14.

Clinical application

Since its first use nearly a century ago11, fluorescence in surgery has proven to be a real-time, dynamic, easy-to-use, non-invasive, low-toxicity, and low-cost way to obtain otherwise invisible data in both open and minimally invasive surgery10. Clinically approved FIGS uses non-targeted fluorescent contrast agents such as indocyanine green (ICG) and methylene blue (MB)14,15, which allow for multiple applications, including surgical navigation, structure identification, tumor delineation, and metabolic activity evaluation3,16.

Fluorescence can be applied dynamically—imaging an area over time to draw clinical conclusions—or statically, using a single image to evaluate the fluorescence intensity of a specific area relative to the background or other areas. The amphiphilic nature of ICG allows it to bind to plasma proteins, achieving uniform blood distribution when injected intravenously. This makes it ideal for real-time, dynamic perfusion assessment17,18, as demonstrated by several clinical trials on fluorescence guided angiography (FA), including in gastrointestinal surgery (e.g., anastomosis evaluation) and transplant surgery (e.g., vascular patency and real-time liver perfusion monitoring15,19,20,21,22,23,24,25).

ICG has also advanced oncological surgery for lymph node mapping and tumor detection. In 2005 Kitai et al. demonstrated its feasibility for lymph node visualization in breast cancer26. Subsequent studies showed it to be as effective as existing alternatives while being less disruptive to surgical workflow and free of ionizing radiation27,28. Fluorescent dyes are also valuable for lymph node mapping in gastrointestinal procedures, such as esophageal, gastric, and colonic surgeries29,30,31. Nonetheless, dose, volume, and injection methods remain topics of discussion32. Tumor detection applications often rely on passive fluorophore diffusion into tumors via aberrant vasculature, pharmacodynamic properties of the fluorophore, or the binding of fluorescent probes to targeted ligands33,34. Off label, ICG has been used to visualize various cancers, including liver, breast, colorectal, head and neck tumors, and melanoma28,35,36,37,38.

Liver applications of ICG have been extensively studied due to its hepatic metabolism35,39. Hepatocellular carcinomas emit a strong fluorescent signal due to impaired biliary excretion40, while liver metastases exhibit a fluorescent ring from healthy immature hepatocytes surrounding the tumor, attributed to downregulation of anion transporters41. ICG hepatic metabolism also enables real-time identification of extrahepatic biliary structures without radiation42, potentially reducing operative time and laparoscopic-to-open conversion rates, though its impact on postoperative complications remains unproven43,44. Finally, ICG is used to demarcate anatomical boundaries in segmental hepatic resections. It guides parenchymal delineation through both positive and negative staining techniques, enabling precise surgery while preserving healthy liver tissue45.

In recent years, several large-scale clinical trials and systematic reviews have underscored the clinical value of ICG‑based perfusion assessment across a range of surgical applications. The EssentiAL trial, a phase III randomized study with 839 patients undergoing minimally invasive rectal surgery, demonstrated that intraoperative ICG fluorescence angiography reduced anastomotic leak rates by 4.2% (RR 0.645, 95% CI 0.422–0.987; p = 0.041)46. Multiple recent meta-analyses of ICG use in colorectal surgery have consistently reported a significant reduction in anastomotic leak risk (pooled relative risks ≈0.5 and NNT 22–23)47,48,49,50. Similarly, for oncologic liver resection, different meta‑analyses (with sample sizes of 417 to 1260) have consistently concluded that ICG guidance significantly decreases operative time (by 16–21 min), intraoperative blood loss ( ~ 100 mL), hospital stay ( ~ 1–1.6 days), and postoperative complications, with a significantly higher one-year disease‑free survival (OR 2.87) relative to non-ICG controls51,52. Beyond these oncological applications, ICG has been increasingly leveraged in the field of Liver Transplant for graft evaluation, donor safety, and anatomical guidance, as shown in a review by Lau et al. which demonstrated that ICG clearance during normothermic machine perfusion can reliably predict early allograft dysfunction and correlates with postoperative outcomes, while intraoperative ICG angiography provides real-time assessment of arterial, venous, and biliary patency20. Recent clinical studies have applied ICG cholangiography during donor hepatectomy to delineate biliary anatomy, demonstrating that the use of ICG reduces biliary complications relative to historical rates, improves precision in bile duct transection53,54, leads to faster biliary anastomosis (mean time 28 vs. 37 min, p = 0.02) and fewer postoperative biliary leaks (2.5% vs. 8.3%, p = 0.04)55.

Open challenges

FIGS has demonstrated its value as a tool that provides surgeons with supplementary, previously invisible information for clinical decision-making. Despite the growing adoption of fluorescent contrast agents in clinical practice and their injection generally being regarded as safe, rare but severe complications, such as anaphylactic shock, have been reported56,57. Furthermore, prolonged washout times of fluorescent contrast agents, such as 30 min for ICG58) pose challenges for the repeated testing required in applications like perfusion monitoring.

Current clinically approved fluorescence systems cannot provide quantitative results because they fail to measure the absolute concentration of fluorescent agents in tissues and lack control over basic variables that affect signal readings59. Fluorescence intensity directly varies with the square distance between the tissue and the imaging system. Without rigorous control or normalization, the observed signal cannot reliably indicate actual perfusion60,61,62,63. Additionally, each commercial system uses proprietary algorithms with unique cutoff values, rendering results from different systems mostly incomparable64,65. Fluorescence signals may appear blurry or lack resolution due to light scattering within the tissue. The attenuation of both excitation light and fluorescence emission hinders accurate localization and quantification of fluorescence, which is essential for precise and quantitative diagnostic applications. As a result, surgeons rely only on qualitative images, leading to subjective, inconsistent, and biased interpretations66,67. Current evidence remains low and insufficient to recommend FIGS broadly for reducing surgical complications36,68,69.

Efforts to achieve objective quantification have introduced methods such as signal-to-background-ratio (SBR), which compares the mean signal from target tissues to background fluorescence. SBR can provide pseudo-quantification and more standardized results70. Calibration devices like the Green Balance™ ICG Reference Card (Diagnostic Green; Aschheim–Dornach, Germany) and the CalibrationDISK™ (SurgVision BV, Harde, the Netherlands) can improve reproducibility by controlling variables such as camera distance and illumination, generating space-resolved pseudo-quantitative results in static images71,72.

Time-resolved quantification has also been explored as an alternative to absolute fluorescence intensity parameters73. For instance, fluorescence-based enhanced reality (FLER) (Fig. 4) analyzes temporal fluorescence characteristics rather than the intensity of individual pixels, generating quantitative perfusion cartography that overlaid onto real-time laparoscopic videos. These color-coded maps, which highlight varying levels of tissue perfusion, offer more clinically valuable insights compared to traditional qualitative ICG systems61,62,74. Although FLER has shown experimental60,61,62,63,75,76,77 and clinical78 accuracy, its implementation remains time-consuming and adds complexity to surgical workflows79. Repeated analyses encounter the same issues observed with standard ICG assessment. Other approaches that rely on time-resolved metrics (e.g., time to peak, maximum fluorescent value over time, slope of the time to peak), face challenges due to the lack of standardized variables, leading to heterogeneous results preventing standard implementation in clinical practice80.

Fig. 4: FLER system used in a left hemicolectomy.
figure 4

a After ICG is injected and recorded over time with an infrared camera (b) Fluorescence Time to Peak is computed (c) and a virtual perfusion cartography is built (Red = High time to peak ∴ low perfusion). d The virtual cartography is superimposed onto real-time images to obtain a mixed reality effect and display the quantified ICG signal directly on the bowel.

Full size image

Machine learning algorithms have shown promise in fluorescence analysis, reducing subjectivity by identifying dynamic intensity patterns over time. These algorithms can distinguish between cancerous and normal tissue with minimal disruption to surgical workflow81. The ongoing CLASSICA project uses machine learning to analyze ICG perfusion patterns, aiming at differentiating benign and malignant rectal polyps82. However, these analyses are conducted in the postoperative period and are not available in real time during surgery.

Targeted fluorescent agents represent another frontier in the advancement of FIGS, aiming to overcome current limitations of oncologic applications and enable real-time analysis. These fluorescent probes typically use a biomarker conjugated to a fluorophore, thereby providing a signal at the target site and enabling real-time visualization of primary tumors, neoplastic lymph nodes, and distant metastasis12,59. Among the myriad fluorescent probes developed, antibodies and peptide-based recognition elements are of particular interest. Fluorophore-antibody probes, inspired by targeted cancer therapies, have demonstrated exceptional precision, with high sensitivity and specificity in detecting multiple tumors in both experimental settings83,84 and clinical trials85. Similarly, fluorescent peptide probes rely on a labeled peptide sequence that can selectively bind to or become modified by the molecular target expressed at the tumor site, offering numerous advantages, including rapid distribution and reduced or absent immunogenicity86. The drawback is that each targeted probe is specific to a tumor and must undergo extensive development and navigate regulatory hurdles before clinical use. Additionally, as with non-targeted fluorophores, quantifying targeted fluorescence to correlate intensity with tumor characteristics remains a challenge. Targeted fluorescent agents have also been evaluated in endoscopic applications. Phase II and phase III trials87,88 have demonstrated their feasibility for tumor identification, highlighting their potential as an alternative for endoscopic follow-up of esophageal and rectal lesions89,90,91.

The limitations of non-targeted and targeted FIGS have spurred interest in alternative imaging technologies. While FIGS offers considerable benefits, its transition to a standard surgical tool requires addressing the lack of quantification, improving standardization, and navigating regulatory challenges for targeted agents.

Dye-free methods have been developed as alternative approaches for tissue characterization that do not require exogenous agents. Autofluorescence, which operates on the same optical principles as exogenous fluorescence but without the need for a contrast agent, has been extensively studied within FIGS. However, its application is limited to tissues with naturally occurring intrinsic fluorophores, such as the parathyroid92,93. Although effective for specific applications, autofluorescence is static and has a significantly lower signal-to-background ratio, which has hindered its development as a general alternative to exogenous fluorescence. In contrast, emerging dye-free techniques that exploit light-tissue interactions—such as reflectance and scattering—are offering new opportunities for tissue characterization. These methods provide safer, more reliable imaging solutions without the need for exogenous agents.

Spectral imaging

Principle

Spectral imaging (SI) refers to imaging techniques that capture multiple bands across the electromagnetic spectrum to determine tissue composition noninvasively. Unlike standard cameras, which use only three visible spectral bands (red, green, and blue — RGB), SI covers a broader range, often including spectral bands corresponding to both visible and invisible light94 (Fig. 5). As different tissue types (e.g., different organs) have unique optical properties (e.g., absorption, reflection and scattering) at different wavelengths, SI can potentially recover their detailed molecular composition, even beyond the visible surface95,96,97. SI is classified based on the number of spectral bands captured and their spectral bandwidth. It can be broadly divided into multispectral imaging (MSI) and hyperspectral imaging (HSI). MSI typically captures up to tens of relatively broad, non-contiguous spectral bands using filters and illumination methods. HSI, in contrast, captures up to hundreds of narrow spectral bands per pixel, providing high spectral and spatial resolution98.

Fig. 5: Schematic representation of a spectral imaging hypercube.
figure 5

SI acquires 3-dimensional data by measuring the tissue reflectance spectrum at each pixel.

Full size image

Clinical application

Numerous image processing and machine learning algorithms have been developed to efficiently handle the large datasets of multidimensional images generated by hyperspectral cameras99. In surgical settings, both regression and classification algorithms have been exploited to extract clinically relevant parameters, ultimately providing valuable insights that can support surgical decision-making100,101.

Regression techniques in SI

Regression techniques are employed to recover and quantify relevant functional tissue properties. This can be achieved using SI, as endogenous light-absorbing molecules in tissue - referred to as chromophores - absorb and scatter light at specific and known wavelengths in unique patterns102,103,104,105,106,107. Hemoglobin in blood is a key endogenous chromophore. SI leverages the distinct spectral signatures of oxygenated and deoxygenated hemoglobin to enable real-time imaging of tissue oxygenation and perfusion108,109. The quantification of functional tissue parameters has been applied across various clinical contexts, especially in colorectal surgery. This includes site selection for colon anastomosis110,111, assessment of colon tissue vascularization112, and the identification of optimal bowel transection points113. Recently, the authors developed the first real-time SI system that was applied in prospective patient studies106.

Beyond tissue perfusion and oxygenation, regression-based SI can also monitor water distribution114, serving as a sensitive and non-invasive tool for continuous, real-time monitoring of tissue hydration across diverse perioperative settings115,116.

Classification techniques in SI

The unique composition of different tissue generates distinct spectral signatures, enabling precise intraoperative differentiation and characterization of specific organs and anatomical structures based on their “spectral fingerprints” (Fig. 6).

Fig. 6: Automated surgical scene segmentation from hyperspectral imaging data.
figure 6

Human organs exhibit characteristic spectral profiles, which can be leveraged for highly accurate surgical scene segmentation using deep learning116. The plots show the median spectra (solid line) across gastrointestinal organ annotations for the sample image, with the shaded areas indicating the standard deviation.

Full size image

This ability of SI to distinguish between tissues in surgical settings was first demonstrated by Zuzak et al. during a laparoscopic cholecystectomy, where a near-infrared MSI system showcased the in vivo visualization of distinct anatomical structures117. Since then, SI has proven effective for anatomical navigation in a plethora of procedures, including thyroid surgery118, neurosurgery119,120, and liver surgery121. When integrated with advanced deep learning algorithms, SI enables automatic tissue recognition122. By classifying each pixel of a HS image, these models can automatically identify specific organs123,124,125 and critical anatomical structures (e.g., biliary ducts, vessels, and nerves)4,123,124,125,126,127,128,129,130. This capability paves the way for real-time augmented reality applications and robotic automation in surgery, offering the potential to enhance precision in surgical interventions4.

Cancer hallmarks, such as angiogenesis and increased metabolism, can alter hemoglobin and oxygen concentration, cellular structures, and vascular networks within tumor sites131. These tumor-associated changes can be leveraged by SI for various clinical purposes, including cancer tissue classification122,132, metastasis detection133,134,135, tumor site delineation108,136, and sentinel lymph node identification across multiple tumor types38. Several studies have investigated hyperspectral imaging (HSI) for the non-invasive detection of colorectal cancer in freshly resected samples135,136. However, these studies are limited by small datasets or rely on classical machine learning models, such as support vector machines (SVMs) or multi-layer perceptrons (MLPs), which only consider spectral features smoothed with low-pass spatial filters, without employing deep learning techniques. Prior research on tissue classification135,136 has shown that these traditional approaches without spatial smoothing, underperform compared to convolutional neural networks (CNNs)122. Ongoing studies are further exploring the applicability of MSI systems in the endoscopic setting, particularly, it is being evaluated by the authors for its ability to distinguish between normal and cancerous tissue in the colon and rectum based on their distinct optical signatures, without requiring intravenous contrast agents or tissue biopsy. Preliminary pre-clinical results have revealed clear spectral distinctions between healthy and pathological tissue in colorectal cancer specimens, further confirming the feasibility and safety of MSI in this context (Fig. 7).

Fig. 7: Hyperspectral data of a specimen of human colon with a tumor.
figure 7

A Representative spectral curve output generated by a commercially available hyperspectral camera. B Human colon specimen with regions of interest (ROIs) corresponding to colorectal cancer (#21, #22) and healthy tissue (ROIs #1–20). Absorbance and reflectance data are presented for wavelengths ranging from 500 to 1000 nm. The spectral curves are not normalized, which poses challenges for interpretation by clinical personnel. Close collaboration with data scientists is essential to derive clinically relevant insights for routine surgical practice.

Full size image

In addition to the functional and structural mapping capabilities discussed above, HSI combined with machine learning can now be applied as a true theranostic tool. By guiding and assessing thermal damage during laser ablation therapies137, HIS can reliably predicting both thermal thresholds and necrosis extent, with a strong correlation to histopathology scores138 thus achieving high accuracy in delineating ablation margins during liver procedures in animal models139.

There is growing evidence of the correlation between HIS and histopathological assessment. In the preclinical models of hepatic outflow obstruction, declines in HSI-derived oxygen saturation and rises in deoxyhemoglobin were spatially co-localized with histologically confirmed venous congestion and necrosis zones on H&E staining121,140. Additionally, in a clinical series of major hepatectomies, intraoperative HSI measurements were significantly correlated with post-operative serum markers of liver injury, which aligned with pathologically confirmed areas of ischemic damage141.

Open challenges

While SI holds promise as a valuable intraoperative imaging tool, several challenges hinder its widespread adoption in clinical settings142. Limitations often emerge from the inherent challenges of balancing advances across all three critical dimensions of SI (spatial, spectral, and temporal resolution). Optimizing one dimension frequently necessitates trade-offs in the others, complicating efforts to achieve simultaneous and comprehensive improvements.

Quantitative imaging with SI faces significant challenges, including reproducibility and the robust quantification of functional tissue parameters. Current devices report non-quantitative chromophore indices (e.g., hemoglobin index) rather than clinically relevant tissue properties. These indices are generated using methods such as the Modified Beer-Lambert Law, time-domain diffuse spectroscopy, and second-derivative techniques, which have inherent limitations143,144,145,146, including the reliance on unrealistic assumptions that do not hold in a clinical scenario or restricted applicability of the results to point measurements147. Machine learning approaches show potential148 but require extensive annotated datasets and reliable reference gold standards — also defined as “ground truth” (e.g., blood gas analysis, or pathology), which remain scarce. Physics-based diffuse reflectance simulations for training machine learning algorithms149,150,151 have equally shown potential but struggle with domain gaps and the lack of complex spatial context (i.e., limits in generating realistic spatially resolved maps of optical properties and simulating complex surfaces that include curvature or texture).

The scarcity of large, diverse medical HSI datasets further limits the development of robust deep learning algorithms. Most existing studies rely on single-center data, with only one known multicenter HSI study4, and none to date exploring real-time applications. Although the number and size of medical HSI datasets are growing— some now encompassing tens to hundreds of patients4,125,152 — few surgical HSI datasets are publicly available120,153,154.

Device variability further aggravates these challenges. Parameters such as imaging methods (e.g., pushbroom, snapshot), number of spectral bands, spectral range, spectral and spatial resolution, sensitivity, and noise levels differ substantially across HSI devices9,155. Standardization is also lacking across file formats, accompanying metadata (e.g., exposure time, gain, environmental factors, clinical data), measurement protocols, and annotations. Such variability complicates cross-study validation and comparison.

Confounding factors remain mostly underexplored in medical HSI, yet they are critical to developing reliable deep learning models, preventing shortcut learning and algorithmic failure156. Seidlitz et al. have begun addressing this challenge by examining subject-, therapy-, and imaging-related confounders in automated sepsis diagnosis using HSI skin images152. For semantic surgical scene segmentation, recent studies have investigated model generalization across shifts in geometric domains124,125, perfusion states157 and illumination conditions158.

Slow acquisition times also impede HSI integration into surgical workflows. Recent advances in high-speed spectral imaging devices for laparoscopic surgery show promise109,159. In open surgeries, dynamically changing lighting conditions necessitate continuous light source calibration, disrupting the surgical workflow and potentially compromising patient safety. Emerging techniques, such as those using specular highlights160 or deep learning158, aim to address inaccuracies in functional tissue parameter measurements and segmentation errors caused by uncalibrated illumination.

To advance SI adoption, multidisciplinary, well-designed, large-scale research initiatives are essential. These efforts should aim to identify the spectral characteristics required for specific diagnostic inquiries and establish which tissue types can be reliably differentiated. Future studies should prioritize capturing predefined spectral target structures, to streamline data processing and image acquisition. Additionally, tailored data extraction algorithms will be crucial to minimize delays between image acquisition and data availability, thereby enhancing the clinical utility of SI in surgical applications142.

Single snapshot imaging of optical properties

Principle

While spectral imaging requires complex data processing to recover optical parameters from raw measurements Spatial Frequency Domain Imaging (SFDI) uses spatially modulated light patterns projected onto a tissue surface to directly probe its optical properties. By analyzing the reflected light at different spatial frequencies, SFDI separates the effects of absorption and scattering, allowing for the extraction of quantitative optical properties (e.g., absorption coefficient, scattering coefficient) across the tissue.

Absorption is primarily dictated by the tissue’s chemical composition, as photons can be absorbed by endogenous molecules, such as hemoglobin, lipids and water. In contrast, scattering arises from local variations in the refractive index and is influenced by cellular structural properties. These interactions are quantified as optical properties: the absorption coefficient and the scattering coefficient, respectively161. As a result, the absorption coefficient reflects the extent and nature of molecular interactions and is typically associated with functional metrics such as oxygenation, hydration or metabolism. In contrast, the scattering coefficient is related to sub-cellular structures, including nuclei, mitochondria, and collagen.

Single Snapshot of Optical Properties (SSOP) is an advanced imaging technique derived from SFDI that aims to simplify and speed up the process of acquiring optical property maps of tissues. Unlike traditional SFDI, which requires multiple spatially modulated patterns at different phase shifts and spatial frequencies, SSOP uses only a single spatially modulated illumination pattern to extract the tissue’s optical properties. Advanced algorithms and pre-calibrations are used to compute tissue absorption and reduced scattering coefficients from just one image162,163.

Clinical application

The measurement of optical properties, namely absorption and scattering coefficients, has been shown in the literature to highlight the presence of diseases or tissue alterations intraoperatively164,165,166,167,168,169,170,171.

Technologies for measuring optical properties typically employ modulation techniques to resolve the temporal and spatial behavior of light propagation161. Most of these technologies require either a temporal sequence of images or a spatial scanning to measure and to generate optical properties maps, limiting their applicability for real-time intraoperative guidance. To address this limitation, SSOP was developed to provide quantitative measures of individual optical properties for each pixel in a wide field image within a single snapshot, thereby enabling real-time intraoperative imaging172 (Fig. 8).

Fig. 8: SSOP Preclinical results.
figure 8

A Oxygen saturation images of a swine bowel during baseline (t = −1 min), occlusion (t = 10 min), and after release (t = 14 min); (B) Corresponding color images; (C) Time course of occlusion measure at the center of the occlusion section; and (D) Pearson’s correlation analysis between normalized lactates and SSOP-StO2 in correspondence to all ROIs. Of note capillary lactates were normalized according to systemic lactates to reduce variability. Lactates in ischemic ROIs (blue dots) were significantly higher compared to vascularized ROIs (red dots procedures) adapted from172.

Full size image

The SSOP technology was demonstrated preclinically in pigs to image oxygenation in real time using the Trident imaging system173. This technology was also validated in a variety of preclinical procedures in pigs, such as esophagus, liver, colorectal, stomach or kidney surgical procedures using metabolic biomarkers such as lactate for correlation170,174,175. Recently it was successfully translated to a clinical trial, where it was used to image the oxygenation of the externalized colon of 10 patients undergoing left anterior resection of the colon78.

Open challenges

Current SSOP implementations, while enabling real-time optical property imaging with acceptable visual quality, face challenges such as edge artifacts and lower resolution compared to SFDI and spectral imaging. The integration of deep learning with Graphics Processing Units (GPU) computation and CNNs have enabled significant advances in biomedical optics, including achieving real-time processing despite network complexity. For instance, a study has demonstrated the feasibility of CNN-based, GPU-accelerated SSOP technology, which integrates 3D profile correction and high-quality image reconstruction from a single SSOP input.172,176,177,178.

A preclinical in vivo trial highlights the capability of SSOP to generate real-time oxygenation rate maps, aligning with results from ICG fluorescence imaging. However, the simultaneous acquisition of oxygenation and fluorescence data has not yet been explored, and integrating or displaying both datasets concurrently could present technical challenges173.

Currently, most studies have been conducted in preclinical settings using equipment that is neither commercially available nor widely accessible. The approval and adoption of the described technologies for clinical use will be essential to facilitate future clinical studies aimed at evaluating their impact on intraoperative surgical decision-making.

Future of surgical optomics

Surgical Optomics represents the convergence of optics, surgery, and computational analytics to create a new paradigm for precision surgery. By turning the operating room into a data-rich environment, this field has the potential to revolutionize surgical care, enabling better outcomes, reduced complications, and faster recovery for patients.

Emerging techniques in surgical optomics, currently in early preclinical development, hold significant promise for advancing intraoperative diagnostics.

Optical elastography is an advanced imaging modality that quantifies the biomechanical properties of biological tissues using light-based measurements. By leveraging the altered mechanical characteristics of pathological tissues compared to healthy ones, optical elastography has the potential to serve as a noninvasive intraoperative diagnostic tool179.

Optical Coherence Tomography (OCT) is already utilized in some medical fields as a non-invasive imaging technique that produces high-resolution, cross-sectional images of biological tissues by measuring the backscattered light. Employing low-coherence interferometry with near-infrared light, OCT achieves varying tissue penetration depths based on optical properties. This technique provides micrometer-scale structural images, enabling detailed visualization and analysis of tissue morphology180.

Full-Field Optical Coherence Tomography (FF-OCT), a variation of OCT, offers high-resolution, en face (surface-parallel) images rather than cross-sectional views. FF-OCT captures the entire field of view simultaneously, allowing for faster imaging over larger areas compared to conventional OCT. This technique is particularly suited for intraoperative pathology, where detailed surface imaging is essential for accurate tissue assessments. FF-OCT allows for real-time, label-free imaging, showing promise for identifying cellular and subcellular structures in tissue samples181.

Intraoperative Raman Spectroscopy (RS) provides real-time molecular information of tissues during surgery by exploiting Raman scattering— a phenomenon where light interacts with molecular vibrations, resulting in wavelength shifts that reveal the molecular composition of tissues. This technique excels in distinguishing healthy tissues from diseased ones, such as differentiating cancerous from non-cancerous tissues, with exceptional precision182.

All the cutting-edge imaging techniques discussed in this paper represent important progress in surgical optics; however, they remain experimental and require further development before clinical integration. Recent studies have begun to quantify clinical benefits, suggesting real potential for improving outcomes. For instance, ICG-guided lymphadenectomy in oncologic liver surgery improved 3-year disease-free survival from 68.2% to 81.4% (HR 0.53, p = 0.012)183, and meta-analyses in colorectal surgery report significant reductions in intraoperative blood loss (WMD –4.38 mL, p = 0.001) and postoperative complications (WMD –0.04, p = 0.027)184. Moreover, multispectral and HSI tools have advanced beyond pilot feasibility and recent studies have demonstrated that real-time intraoperative oxygenation mapping is now practical, though integration into standard workflow remains an active area of development101.

While technical feasibility has been demonstrated, routine clinical adoption will require overcoming implementation barriers. Future research will focus on assessing their clinical feasibility, cost-effectiveness, and potential to enhance diagnostic precision and support real-time surgical decision-making.

As detailed above, further important steps towards the advancement and implementation of surgical optomics in clinical practice are:

  • Standardization and Validation: Establishing clinical standards for using these techniques to ensure reliability and reproducibility across diverse surgical contexts.

  • Integration into Surgical Workflows: Developing portable, easy-to-use systems that seamlessly integrate into operating rooms.

  • Broad Accessibility: Reducing costs and system complexity to make Surgical Optomics accessible for a wider range of healthcare settings.

Conclusion

Surgical optomics represents a groundbreaking advancement in precision surgery. By harnessing the properties of light and bespoke advanced computational methods, including machine learning, it can enable real-time visualization, characterization, and quantification of tissue and organ properties at both microscopic and macroscopic scales. Although still in the exploratory and developmental stage, this innovation has the potential to enhance surgical outcomes by facilitating precise resection of diseased tissue, minimizing collateral damage to healthy structures, and improving patient safety.

While Surgical Optomics is still in the exploratory and developmental stage, its future envisions operating rooms transformed into data-driven environments where real-time optical imaging technologies will enable the rapid interpretation of complex optical data, empowering surgeons with unprecedented insights.

As the technology advances, these tools will become more compact, faster, and seamlessly integrated into surgical workflows. However, its successful integration depends on surgeons’ comprehensive understanding of these technologies and their inherent limitations. promise to enable unprecedented precision, ushering in an era of personalized, minimally invasive, and highly effective surgical interventions. Continued research and development in this field hold the potential to redefine the standard of care, enhancing patient safety and improving outcomes on a global scale. Crucially, the future application of these technologies will require seamless collaboration among surgeons, engineers, and data scientists, emphasizing the importance of interdisciplinary integration in achieving this transformative vision.