JIMT

Recently, optical imaging has become widely used in medical diagnostics¹. Biomedical optical imaging enables precise diagnosis of various legions, often providing superior resolution compared to traditional radiological or ultrasound methods. The portability of optical imaging devices, along with their use of non-ionizing radiation, has further contributed to the growing adoption of this technology in clinical settings. Among optical imaging techniques, fluorescence imaging is particularly valuable for biochemical and molecular diagnostics due to its ability to provide optical molecular contrast. Consequently, specific fluorescent molecular markers have been introduced into medical diagnostics to generate contrast between different biochemical components^2,3. However, the clinical application of exogenous fluorophores is limited by their potential toxicity, posing challenges for their translation into clinical practice. Alternatively, certain endogenous biomolecules exhibit autofluorescence when excited by specific light sources in the range from ultraviolet to near-infrared. This intrinsic autofluorescence enables label-free, imaging-based diagnostics, eliminating the risk of dye toxicity. Despite this advantage, autofluorescence signals are often difficult to interpret because their sources are often unknown or mixed, and signal intensity can be influenced by various environmental factors.

On the other hand, fluorescence lifetime, defined as the decay rate of fluorescent signals, can provide robust intrinsic biochemical information about biological samples. Because fluorescence lifetime is independent of signal intensity, it allows for quantitative analysis of biochemical properties. The most common method for capturing the decay rate is time-correlated single photon counting (TCSPC)⁴. This technique measures the arrival times of individual photons to generate a histogram, which is then analyzed using non-linear exponential fitting to determine the decay rate. Although TCSPC provides accurate measurements of fluorescence lifetime, it can be time-consuming, as it requires counting over 10,000 individual photons to produce a reliable histogram. Recently, high-speed fluorescence lifetime measurement methods have been developed, making it feasible to use fluorescence lifetime as a real-time diagnostic tool^5-7.

This review explores the application of fluorescence lifetime imaging (FLIm) in medical diagnostics, with a particular focus on cancer and vascular disease. In oncology, FLIm has demonstrated potential in distinguishing between healthy and malignant tissues by leveraging the distinct metabolic activities of tumors. In vascular disease, FLIm has been used to characterize various arterial plaque components, such as inflammation, collagen, muscle cells, and lipids, by identifying their unique fluorescence lifetime signatures. This review synthesizes current research on FLIm, highlighting its advantages and potential for enhancing diagnostic accuracy and improving patient outcomes in both fields.

Endogenous fluorophores such as collagen, elastin, nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD), and porphyrins can provide valuable insights into functional and structural changes within the tumor environment. Especially, numerous studies on the autofluorescence of metabolic coenzymes, NADH and FAD, have demonstrated that these endogenous fluorophores serve as indicators of metabolic alterations within biological tissues. The optical redox ratio, defined as the fluorescence intensity ratio of FAD to NADH+FAD⁸, has been utilized to assess shifts in metabolic pathways between glycolysis and oxidative phosphorylation (OXPHOS). In addition, the metabolic alterations can be monitored by the temporal decay profile of NADH and FAD^9,10, as their fluorescence signals decay at different rates depending on the ratios of their free and protein-bound states. These characteristics can be exploited to provide multifaceted information, helping to address complications arising from the spectral overlap of various endogenous fluorophore in tissue.

Alterations in cellular metabolism in the tumor environment tend to rely more on glycolysis rather than OXPHOS, known as Warburg effect¹¹. This phenomenon presents an opportunity to interrogate tumors by measuring the fluorescence signals of NADH and FAD. Extensive research into the potential use of endogenous FLIm for cancer diagnosis has been conducted across various organs, including oral cavity^12,13, skin¹⁴, breast¹⁵, lung^16,17, and brain^18,19. Especially, FLIm of endogenous fluorophores has shown promise in the fields of surgical resection. Precise delineation of tumor margins during surgical resection is critical for the complete removal of cancerous tissue while preserving the functionality of the affected area, which significantly impacts patient outcomes. Traditional clinical assessment of tumor margins is typically performed visually by surgeons during operations using white light imaging, followed by a histopathological assessment of the excised tissue. However, traditional methods have limitations, such as sampling errors during surgery and prolonged feedback times from histopathological evaluations. In contrast, label-free FLIm offers immediate, non-invasive feedback to surgeons, potentially improving the accuracy and efficiency of tumor resection. As a result, FLIm instruments have evolved to meet the clinical demands, including enhanced accessibility to the measurement sites, seamless integration with other instruments, real-time image acquisition, a large field of view, and precise differentiation of pathological regions. Sun et al.²⁰ demonstrated a flexible fiber-optic fluorescence lifetime endoscope for in vivo applications, which is composed of a pulsed nitrogen laser with a wavelength of 337 nm, a gradient index lens, and a gated intensified charge-coupled device (CCD) camera with time-gated widefield FLIm acquisition. In this study, imaging of a hamster model of oral carcinogenesis revealed that tumor tissues exhibited shorter lifetimes compared to normal tissues. Clinical trials using the same instrument on head and neck cancers as well as brain tumors demonstrated weaker fluorescence intensity at the 460 nm emission wavelength. Additionally, differences in lifetime between tumor and normal tissues were observed, highlighting the potential of FLIm as a tool for image-guided surgery^19,21.

The FLIm systems have become more robust, compact, and capable of rapid data acquisition with multispectral detection, making them suitable for clinical environments²². One example is real-time point-scanning multispectral time-resolved fluorescence spectroscopy, which augments FLIm maps on top of wide-field white-light images by superimposing fluorescence lifetime measurement points in real time^12,23 (Fig. 1A). Additionally, research has also been conducted to differentiate breast cancer tissue composition using FLIm-derived data and machine learning-based image segmentation²⁴. For clinical translation, a compact, handheld FLIm endoscope system has been introduced²⁵, utilizing a 355 nm wavelength as the excitation source and collecting fluorescence data across three spectral bands optimized for key endogenous fluorophores: collagen, NADH, and FAD. A pilot clinical study with this system explored its feasibility in distinguishing between benign, dysplastic, and early-stage cancerous oral lesions. This study demonstrated that the use of fluorescence lifetime-derived features, combined with machine learning analysis, achieved high diagnostic performance¹³ (Fig. 1B). Fernandes et al.¹⁷ introduced a fiber-based autofluorescence FLIm micro-endoscopy system and demonstrated promising results in differentiating between cancerous and non-cancerous lung tissues in ex vivo samples, with a sensitivity of 81.0% and specificity of 71.4% (Fig. 1C). On the other hand, a macro-FLIm system capable of imaging large areas of brain tissue, up to 18 mm in diameter, has been employed to assess entire excised tissues²⁶. This system was used to record fluorescence lifetime images from freshly excised rat brain tissues with glioma models and from human glioblastoma samples obtained during surgery. In the human samples, glioblastoma tissues exhibited longer fluorescence lifetimes compared to non-infiltrated white matter, indicating the potential of FLIm to assist in the rapid identification of tumor margins during surgical procedures (Fig. 1D). Therefore, FLIm-based diagnostics in the field of oncology are advancing with the miniaturization of imaging probes and the development of advanced image processing techniques, which are expected to enable accurate, real-time diagnosis and provide essential guidance during surgical operations.

Figure 1. Fluorescence lifetime imaging (FLIm) in clinical environments. (A) Fiber-based FLIm augmenting the surgical field of view for precise tumor margin delineation in head and neck cancer¹². (B) Endoscopic imaging of fluorescence lifetime imaging microscopy (FLIM), assisted by machine learning segmentation, for detecting oral cancer¹³. (C) Changes in fluorescence lifetime observed in lung cancer tissue ex vivo, imaged using FLIM integrated with a fiber-optic imaging bundle¹⁷. (D) Macroscale FLIm images of excised glioblastoma tissues²⁶. Images reproduced from the references with original copyright holder’s permission.

Coronary artery disease is the leading cause of death worldwide²⁷. It is primarily caused by thrombotic occlusion following the rupture of a coronary plaque, which can result in acute myocardial infarction or sudden cardiac death. High-risk coronary plaques exhibit several distinct histopathological features, such as a large lipid-rich core, a thin fibrous cap, and abundant macrophage infiltration²⁸. Plaque formation involves the recruitment of smooth muscle cells (SMCs) in response to intra-intimal lipid accumulation and related immune activation. These SMCs produce collagens, which play a pivotal role in stabilizing plaques²⁹. Therefore, assessing these key biochemical components can provide valuable biological insights into the risk of plaque rupture.

Autofluorescence lifetime imaging has been introduced as a label-free method to capture the various biochemical components of plaques. In 2001, Marcu et al.³⁰ successfully discriminated lipid-rich lesions in human coronary arteries using the time-resolved laser induced fluorescence spectroscopy (TR-LIFS) by analyzing the fluorescence spectra of arterial fluorescent compounds³¹. Additionally, macrophages, key indicators of plaque formation, were detected in vivo in an atherosclerotic mouse model³². TR-LIFS effectively identified rupture-prone atherosclerotic plaques, distinguishing necrotic cores and fibrous plaques from regions of merely thickened intima³³. These studies demonstrated the feasibility of fluorescence lifetime-based plaque characterization, prompting further research to characterize multiple biochemical components using TR-LIFS. It was observed that the biochemical components involved in atherosclerosis, such as collagen, SMCs, and low-density lipoprotein (LDL), exhibit different emission spectra and fluorescent lifetimes. This led to the proposal of multispectral FLIm for cardiovascular in vivo imaging³⁴. Multispectral FLIm enabled the classification of LDL-rich atherosclerotic plaque by analyzing fluorescence lifetime across multiple spectral bands³⁵. More recently, multispectral fluorescence lifetime imaging microscopy (FLIM) has been developed, allowing for the label-free characterization of multiple plaque components³⁶. Fig. 2A presents multispectral FLIM images alongside their co-registered histological validations of high-risk plaques. The images reveal that regions rich in lipids and macrophages exhibit shorter lifetimes compared to regions abundant in collagen and SMCs, demonstrating the potential of multispectral FLIM as a diagnostic tool for assessing high-risk atherosclerotic plaques.

Figure 2. Fluorescence lifetime imaging (FLIm) on atherosclerotic plaques. (A) FLIm microscopy images with four different stainings³⁶. (B) IVUS-FLIm imaging for predicting newly formed plaque and foam cell regions⁴³. (C) Machine learning-assisted characterization of biochemical composition using in vivo optical coherence tomography (OCT)-FLIm images⁵³. (D) OCT-FLIm images highlighting normal and necrotic core regions⁵⁰. Images reproduced from the references with original copyright holder’s permission.

In the diagnosis of atherosclerosis, intravascular ultrasound (IVUS) and intravascular optical coherence tomography (IV-OCT) are significant tools, as they provide cross-sectional images of arteries³⁷. However, since FLIm offers only biochemical information about tissue compositions, FLIm-based diagnostic devices have been developed in combination with structural imaging modalities such as ultrasound imaging, photoacoustic imaging (PAI), and OCT^38-40. In a benchtop setup, FLIm was combined with ultrasound backscatter microscopy and PAI, allowing for the clear distinction between fibro-lipidic plaques and fibrotic plaques³⁸. To achieve luminal imaging, FLIm was combined with IVUS in a catheter-based system, allowing for in vivo co-registered IVUS-FLIm imaging of swine vessels^41,42. Additionally, human coronary artery segments were analyzed using the IVUS-FLIm system, revealing that FLIm can identify plaque progression in atherosclerosis, such as peroxidized-lipid-rich foam cell accumulation and recent plaque formation⁴³ (Fig. 2B).

While IVUS-FLIm provides complementary diagnostic information in atherosclerosis, OCT offers another option for visualizing structural details with much higher spatial resolution. Park et al.⁴⁴ developed a benchtop dual-modal imaging system that combined OCT and FLIm to analyze atherosclerotic plaques of excised tissues. Plaques such as pathological intimal thickening, fibroatheroma, thin-cap fibroatheroma, and fibrocalcific plaque were manually characterized based on OCT, while collagen-rich, lipid-rich, and collagen/lipid-poor areas were identified based on FLIm⁴⁵. These tissue compositions were successfully identified using linear discriminant analysis. However, the limited penetration depth of OCT, due to its 830nm central wavelength light source, restricted the detection of certain morphological features.

To enable real-time, simultaneous acquisition of OCT and FLIm images, a dual-modal system was developed using a 1,310 nm swept-source laser for OCT and a high-speed lifetime acquisition algorithm⁴⁰. Following benchtop studies, intravascular imaging was subsequently performed using a single imaging probe that combined OCT and FLIm. Prior to this combination, other imaging modalities, such as near infrared spectroscopy⁴⁶, near infrared fluorescence⁴⁷, and near infrared autofluorescence⁴⁸ were integrated with OCT in a catheter-based system. The basic concept of integrating OCT with additional biochemical imaging for vascular imaging was first reported in 2011⁴⁷. In this approach, the OCT light was transmitted through the core of a double-clad fiber (DCF), while additional light for other modalities was transmitted through the first cladding of the DCF. Similarly, FLIm and OCT were integrated into a DCF based catheter^49-51. In 2018, Lee et al.⁴⁹ developed an integrated FLIm and OCT system with a custom-built optical rotary junction that covers the broadband range of both OCT and FLIm light sources. Rapid combined FLIm and OCT imaging at a rate of 100 frames per second was achieved by utilizing swept-source OCT and real-time FLIm acquisition using an analog mean delay method⁵². To mitigate autofluorescence noise generated by the 355 nm light source, the imaging probe was designed with a fused silica-based ball lens, and FEP tubing was used for the catheter imaging window. This system successfully obtained multimodal images of lipid-rich plaques and a normal aorta in a rabbit model in vivo, revealing distinct FLIm characteristics. This imaging study was subsequently applied to coronary artery diseases, where atherosclerotic plaques in swine coronary arteries were imaged. In the study, the different biochemical components of atherosclerotic plaques were successfully characterized using a machine learning algorithm⁵³ (Fig. 2C). This automatic characterization demonstrated the feasibility of detecting multiple biochemical compositions in a label-free manner, showing promise for revolutionizing high-risk plaque imaging. A first-in-man clinical trial is currently underway (NCT04835467)⁵⁴.

On the other hand, Chen et al.⁵⁰ also developed a combined intravascular OCT-FLIm system using a different lifetime acquisition technique based on frequency analysis. They obtained OCT-FLIm images of normal and necrotic core regions from human coronary arteries, showing elongated fluorescence lifetimes in channel 3 (~540 nm) at the necrotic core regions, where lipids and macrophages were co-localized (Fig. 2D). The research group led by Prof. Marcu, known for pioneering autofluorescence FLIm for the label-free diagnosis of atherosclerosis and cancer, also developed a dual-modal intravascular OCT-FLIm system. Their imaging studies on excised human coronary arteries revealed distinct fluorescence lifetime characteristics between collagen-rich regions and macrophage-infiltrated regions⁵¹. To address chromatic shift between the two modalities, they developed a new type of imaging probe suitable for both IVUS-FLIm and OCT-FLIm⁵⁵. Imaging of human coronary specimens was successfully performed, with the results validated through histopathological staining⁵¹. More recently, they reported the development of an integrated imaging system combining FLIm with polarization-sensitive OCT, enabling the capture of more detailed information about the coronary artery⁵⁶. These advancements suggest that catheter-based intravascular FLIm, in combination with morphological imaging modalities such as IVUS and OCT, is nearing readiness for clinical application.

In this review, we explored the applications and technical advances of autofluorescence-based FLIm for diagnosing lesions such as tumors and atherosclerosis. With the development of high-speed FLIm acquisition method, FLIm has evolved into a system capable of real-time diagnosis. Particularly in cancerous tumor surgery, FLIm systems have been developed primarily in endoscopic and fiber-based forms to enhance integration with other equipment and improve accessibility during the surgical procedures. This design allows for flexible sampling points over a wide area, thereby improving the system’s adaptability and effectiveness in various surgical settings. For diagnosing atherosclerosis, FLIm systems have been developed in catheter-based forms and integrated with morphological imaging modalities, such as IVUS and IV-OCT, to provide accurate diagnosis and visualize biochemical components of different plaque types.

Label-free FLIm, enhanced by advanced image processing techniques such as machine learning algorithms, is emerging as a promising diagnostic tool, capable of distinguishing various biochemical compositions with greater precision. Ongoing clinical trials and research efforts aimed at refining these tools highlight their potential to revolutionize medical diagnostics, offering more precise, non-invasive, and real-time insights into complex disease processes. As FLIm-based systems continue to be developed and validated, they are likely to become integral components of personalized medicine, providing critical information that guides treatment decisions and improves patient care outcomes.

None.

Hongki Yoo owns stocks in Dotter, a company developing an intracoronary imaging system and catheter. Hongki Yoo is currently serving as an associate editor for Journal of Innovative Medical Technology; however, he was not involved in the peer reviewer selection, evaluation, or decision process of this article. The others have no potential conflict of interest relevant to this article.

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (Project Number RS-2023-00208888 & RS-2024-00401786) a Korea Medical Device Development Fund grant funded by the Korean government (Ministry of Science and Information and Communication Technologies, Ministry of Trade, Industry and Energy, Ministry of Health & Welfare, Ministry of Food and Drug Safety) (Project Number, RS-2023-00254566).