Multispectral image-guided surgery in patients

An optical-imaging instrument that integrates multispectral imaging for the detection of fluorescence in the first and second near-infrared windows aids the surgical resection of liver tumours in patients.

Despite significant advances in systems for intraoperative imaging guidance, oncological surgery is still mostly performed without real-time image-based assistance owing to the lack of clinically oriented imaging systems and optical contrast agents that can help the surgeon localize lesions, ensure clear resection margins and find small metastases. The primary unmet clinical need for intraoperative imaging is the ability to optically differentiate the diseased tissue so that it can be resected more easily. Conventional non-invasive medical imaging modalities such as computed tomography, magnetic resonance imaging, single photon emission computed tomography and positron emission tomography are used for surgical planning but are impractical for real-time intraoperative use; also, they are too costly, even when taking into account the development of portable and compact equipment for intraoperative use^1,2. Ultrasound can be used for real-time imaging despite its limited resolution but requires direct contact with the tissue and therefore is incompatible with surgical conditions.

Owing to the favourable light transmission properties of near-infrared (NIR) fluorescence imaging across biological tissue (in comparison to visible light), this modality can provide real-time navigation during cancer surgery. The first NIR spectral window (NIR-I; wavelengths of 650–900 nm) has been extensively used for the visualization of healthy tissues, such as blood vessels, nerves, the ureter and endocrine glands^3,4 as well as cancerous tissue^5,6, and is under evaluation for image-guided surgery owing to its superb sensitivity and high temporal and spatial resolution. However, NIR-I imaging suffers from relatively high scattering, which limits its use to reflectance geometries in shallow tissue. Fluorescence from the second NIR spectral window (NIR-II; wavelengths of 1,000–1,700 nm) has been used for in vivo imaging applications because it allows for deep tissue imaging at high resolution^7,8 owing to reduced light scattering, minimal light absorption and extremely low autofluorescence levels. However, translating NIR-II imaging into the clinic has been limited by the lack of available detection systems and clinically approved targeted contrast agents. Over the past 15 years, the United States Food and Drug Administration (FDA) has approved several surgical grade image-guided systems² (in particular, the Novadaq SPY Elite in 2005, the Hamamatsu PDE in 2012, the Fluoptics Fluobeam 800 in 2014, the da Vinci Firefly in 2014, Quest’s Artemis in 2015, the Novadaq Pinpoint in 2016 and the SPY PHI in 2017). These instruments, based on reflectance geometries and operated in the NIR-I window, are designed for use during either open surgery or minimally invasive surgery and provide real-time image-guided identification of vital tissues at the limited penetration depth of 1 cm or less. None of these systems are capable of operating in the NIR-II spectral region. Reporting in Nature Biomedical Engineering, Jie Tian, Zhen Cheng, Sam Gambhir and colleagues describe the first-in-human evaluation of an intraoperative NIR-II camera that integrates the detection of signals from visible, NIR-I and NIR-II wavelengths for clinical use¹. The researchers demonstrate the advantages of NIR-II imaging over those of the NIR-I and visible spectral windows for the intraoperative detection of hepatocellular carcinomas and for their image-guided resection.

Tian and co-authors’ instrument detects a broad NIR spectrum of wavelengths — from 650 nm to 1,700 nm — and the integrated system enables multispectral and multichannel imaging of distinct target tissues simultaneously (Fig. 1a). For example, a spectral region (such as the NIR-IIb region) can be filtered to distinguish malignant tumour tissue for resection, while another channel (for instance, from the NIR-IIa region) is used to characterize lymph node metastatic infiltration. Additional NIR-I channels can be tailored to visualize vital tissues that should be preserved, for instance by using 800-nm NIR for visualizing blood vessels and 700-nm NIR for visualizing nerves. Also, because the authors’ NIR-II imaging approach is less influenced by ambient lighting, it can be used in the operating room with the lights switched on; this is not the case for NIR-I imaging, which requires low-light settings (despite the fact that white light noise can be filtered out using prisms and dichroic filters; Fig. 1b). The authors’ multispectral system therefore improves the performance of image-guided surgery, making it a quick and safe option for the reduction of the likelihood of follow-up surgeries and for improving overall survival.

Fig. 1: Intraoperative image-guided surgery via a multispectral imaging system.

a, The integrated visible colour–NIR-I–NIR-II imaging system¹. b, The visible colour–NIR-I–NIR-II imaging system operates via three independent detectors. CCD, charge-coupled device; ROI, region of interest. c, White light (colour), NIR-I and NIR-II imaging of residual cancer following the resection of hepatic tumour tissue from a patient who had been injected with the fluorophore ICG before the surgery. Residual cancer is only detectable with NIR-II imaging. Panels a,c reproduced with permission from ref. ¹, Springer Nature Ltd; panel b reproduced with permission from ref. ¹⁴, Taylor & Francis, Ltd.

Besides assembling the integrated system, Tian and co-authors used indocyanine green (ICG; the only FDA-approved NIR fluorophore that shows fluorescence emission tailing up to 1,600 nm⁹) for both NIR-I and NIR-II wavelengths. The authors show that ICG-based NIR-I/II fluorescence imaging (Fig. 1c) enables the intraoperative detection of tumour lesions that were not detectable via preoperative diagnostic modalities, enhancing tumour detection rates by 46% and 56% through NIR-I and NIR-II imaging, respectively. Also, the use of NIR-I/II imaging significantly increased the accuracy of the staging and prognosis of hepatic tumours in 23 liver cancer patients, for which the data generated maximum tumour-to-background ratios of 1.45 and 5.33 for intraoperative NIR-I and NIR-II imaging, respectively. Moreover, the NIR-II spectral window provided an even higher tumour detection sensitivity than the NIR-I spectral window (100%, with a 95% confidence interval (CI) of 89–100% versus 90.63% with a 95% CI of 75–98%). Notably, NIR-I/II imaging could also help to identify residual lesions during intraoperative procedures that are notoriously difficult for the surgeon to recognize visually or by intraoperative ultrasonography. Lesions were also detected in resected specimens ex vivo by using both NIR modalities.

ICG-based NIR-II imaging could be used to visualize several other types of tumours, for diagnosis or for treatment evaluation, and represents a good starting point for the clinical evaluation of the NIR-II window. As a non-targeted fluorophore, however, ICG is a suboptimal cancer tracer, as it is unable to distinguish malignant tumour from benign lesions; in fact, it typically leads to a high false positive detection rate and suffers from an extremely short half-life in blood (less than 5 min), which limits its use for surgical procedures (ICG is mostly used to visualize the vasculature). Also, ICG is unstable in aqueous environments (it binds to serum proteins) and is taken up predominantly by the liver, making it useful for the diagnosis of liver fibrosis and cirrhosis (but not for the optical imaging of other tissues). Even in patients with pulmonary neoplasms, it is hard to detect any lesions located 5 mm beneath the surface owing to the extremely low uptake of ICG in tumour tissue, even at a high dosage of 5 mg kg^–1 (ref. ¹⁰). In fact, by using ICG, Tian and colleagues could only identify partial lesions (or their rim) in patients with poorly differentiated hepatocellular carcinomas. Hence, there is a pressing need for the development of clinically translatable NIR imaging tracers targeted to tumours or to vital tissues.

The translation of targeted NIR agents for image-guided surgery has two fundamental challenges: their non-specific uptake by normal tissues and the incomplete elimination of unbound agent from the body^6,11. The design of targeted contrast agents with fast clearance from background tissues (and then from the body) thus remains a key aim. Small molecule dyes, polymers and organic and inorganic nanoparticles in particular have been explored as targeted agents for NIR-I/II fluorescence imaging^9,12. And, in the development of fluorophores with inherent chemical structures for targeting specific tissues, tissue-specific components and fluorophore domains are combined into a single molecule for the detection of tissue features in real time⁴; the resulting compact structural design enables the efficient systemic clearance of the unbound contrast agent, reducing background signal and improving the contrast of the signal emitted in the targeted tissue. The physiological filtration and clearance of these compounds are consistent with known criteria¹³: hydrophobic compounds larger than 5.5 nm in hydrodynamic diameter are processed by the liver, and charge-balanced agents with smaller hydrodynamic diameters are cleared by renal filtration^3,13.

Tian and colleagues’ first-in-human demonstration of NIR-II fluorescence-guided tumour resection opens the door to future testing of multispectral NIR-I/II imaging technology in other clinical settings that may similarly benefit from the technology’s high sensitivity (in particular, in dermatology, ophthalmology, neurology and gynaecology). After about a decade of preclinical evaluation^7,8,9,10,11, NIR-II imaging is ripe for clinical testing.