Microscopic assessment of lymph node status in gynecological malignancies using full-field optical coherence tomography

Abstract

Accurate assessment of lymph node (LN) status is critical in cancer management, particularly in gynecological malignancies. However, preoperative identification of pathological LNs remains a significant challenge with current imaging modalities. Full-field optical coherence tomography (FF-OCT) is a non-invasive microscopic technique based on tissue reflectivity and light interference, providing real-time, high-resolution images in < 10 min, with no tissue preparation or alteration required. Our aim was to evaluate the diagnostic accuracy of FF-OCT in identifying LN metastatic foci measuring ≥ 0.2 mm in gynecological cancers, in an intraoperative setting. Comparative analysis of 80 fresh ex vivo LNs with FF-OCT versus gold standard pathology showed high accuracy (97.6%), sensitivity (92.3%), and specificity (98.2%) of FF-OCT. These results support the suitability of FF-OCT integration into clinical practice for real-time assessment of LN status, thereby improving intraoperative decision making while enabling subsequent routine histological analysis.

Introduction

In gynecological malignancies, rapid and accurate assessment of lymph node (LN) status is critical because it directly influences surgical decisions, adjuvant treatment, and prognosis^1,2. Detecting LN micro- or macrometastases with preoperative imaging modalities such as contrast-enhanced computed tomography (CT), magnetic resonance imaging, and positron emission tomography/CT remains a significant technical challenge due to limitations in scan time and spatial resolution³. While histology is the gold standard for assessing LN status, results are often delayed by the complex and unavoidable technical processes involved⁴.

Definitive histological evaluation can only be performed after surgical LN removal. Additionally, the frozen section technique for intraoperative LN analysis has several limitations, including morphological artifacts that reduce sensitivity for detecting micrometastases (< 2 mm)⁵ and limited availability for LN ultrastaging, which is time-consuming and requires expert pathologists for interpretation⁶. Furthermore, depending on the surgical learning curve or the body mass index of patients, the rate of so-called ‘empty packets’ (i.e., the absence of LNs, with only adipose tissue retrieved) can reach up to 7–20%^7,8. These challenges highlight the need for an innovative technique for intraoperative LN status assessment.

Recent studies in ovarian and breast cancer^9,10 have assessed the use of full-field optical coherence tomography (FF-OCT) for rapid, histology-like examination of LN tissue and demonstrated its efficacy in distinguishing between healthy tissue and tumor cells¹¹. Unlike other real-time optical imaging methods (e.g., fluorescence, confocal, or laser Doppler imaging), FF-OCT provides high-resolution (1 μm axial) cross-sectional images of fresh tissue samples in real time. More importantly, it is suitable for an intraoperative use as it allows rapid scanning (< 15 min) without the need for tissue preparation, thereby preserving samples for subsequent histological analysis^12–14.

These attributes make FF-OCT a valuable tool for advanced intraoperative LN imaging, offering a temporal advantage over traditional histology, particularly suitable for assessment in cervical cancer where guidelines recommend intraoperative sentinel LN analysis¹. Furthermore, advances in deep learning can be applied to all numerical images, including FF-OCT, which may help to automatically assess LN metastases during surgery in the near future¹⁵.

Our aim was to assess the diagnostic accuracy of FF-OCT in identifying metastatic LN foci measuring ≥ 0.2 mm in a cohort of freshly removed ex vivo LNs from patients with gynecological cancers, compared to histology as the gold standard.

Materials and methods

Optical coherence tomography (Light-CT) system

FF-OCT scanning of fresh ex vivo LN tissue was performed using a CelTivity Biopsy System (AQUYRE Biosciences), which features a Linnik interferometer with incoherent illumination. The system consists of two arms: one holding the sample (object arm) and the other housing a reference mirror. For “en face” image acquisition, a scanning unit measuring 1.24 mm × 1.24 mm is employed, with the process completed in up to 2 s. Sequential scanning of adjacent units is performed in designated areas, and these scans are combined to create a larger field of view. Both the native thickness of the scanned sections and the device resolution are 1 μm. The system allows for “optical slicing” of the tissue at user-selected depths, providing microscopic-scale imaging beneath the surface¹⁴. In addition, the system supports dynamic cell imaging, which offers a dynamic view of intracellular activity that complements the morphological insights gained through FF-OCT¹⁶.

Sample collection, tissue processing, and optical/histological image matching

This observational prospective study was approved by the relevant ethics committee (CE-2021-143) and written informed consent for anonymized data collection was obtained from all patients. The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. From February 2023 to July 2024, patients scheduled for LN dissection as part of their surgical treatment for gynecological cancer were enrolled based on the availability of the FF-OCT machine. Fresh ex vivo LNs were retrieved during surgery. LN samples were carefully freed from surrounding fat and immediately transferred to the FF-OCT laboratory. The samples were sliced with a microtome and placed in the dedicated specimen containers suitable for immediate FF-OCT scanning, without further tissue preparation. Images were acquired at a depth of 15 μm. After FF-OCT imaging, the specimens were transferred into standard containers for histopathological analysis, appropriately marked and oriented to ensure that the histological sections corresponding to the FF-OCT acquisition plane were identified; then, formalin-fixed, paraffin-embedded, hematoxylin-eosin (H&E) stained, 3-4-µm thick sections were prepared for standard histological evaluation. In addition to standard diagnostic sections, sections co-registered with the intraoperative FF-OCT acquisition were obtained for comparative analysis. All histology slides were digitally scanned using either the Ventana DP 200 slide scanner (Roche) or the Pannoramic P480 (3DHistech) and reviewed by an independent pathologist (PLQ). Finally, the optical images were blindly reviewed by a second experienced pathologist (AN). The LNs were classified as normal (N-) or positive (N+, with a distinction between macrometastases measuring > 2 mm and micrometastases measuring 0.2 to 2 mm). Isolated tumor cells (< 0.2 mm) or other malignant LN pathologies (e.g., lymphoma) were not included in the statistical analysis. Figure 1 illustrates the study workflow, sample preparation, and FF-OCT system used.

Fig. 1.

Study design and workflow.

Statistics

The primary objective of this study was to evaluate the sensitivity of FF-OCT in detecting LN metastases compared to histology as the gold standard. Therefore, the number of LNs analyzed rather than the total patient cohort was considered. The sample size was determined based on a 10% risk of metastasis, requiring a total of 70 LNs to ensure a sensitivity of 99% with a 95% confidence interval (CI) and a half-width of 7%, along with a precision of the specificity of 10%. Image interpretation results were evaluated using a 2 × 2 contingency table to calculate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy for the entire cohort.

Results

Sample pathological evaluation

A total of 80 LNs were sampled for the study. However, due to technical errors encountered during the initial phase of the procedural setup, 6 LNs were excluded, resulting in a final analysis of 74 LNs from 20 patients with gynecological malignancies: 46 LNs (62.2%) from endometrial cancer, 17 (23%) from ovarian cancer, and 11 (14.8%) from cervical cancer. The majority of the LNs (75.7%, 56/74) were healthy (N-), while 4 specimens (5.4%) contained only adipose tissue. A total of 13 LNs (17.5%) were found to be metastatic (N+). Of these positive LNs, 12 (92.3%) exhibited macrometastases (i.e., metastases > 2 mm in the larger axis), with 1 of 12 (8.3%) from endometrial cancer, 6 of 12 (50%) from ovarian cancer, and 5 of 12 (41.7%) from cervical cancer. Co-occurring micrometastases (metastases < 2 mm) were observed in 2 of 12 LNs. Besides, 1 of 13 LNs (7.7%) exhibited a solitary micrometastasis in a patient with endometrial cancer. Isolated tumor cells were identified in only 1/70 LNs (1.4%) on H&E stains in a patient with endometrial cancer. Additionally, B-cell lymphocytic lymphoma was diagnosed in 1 of 70 cases (1.4%) in a patient with endometrial cancer. Since our focus was on detecting macro- and micrometastases, both cases were considered negative.

FF-OCT image interpretation

FF-OCT images were independently reviewed by a pathologist who was blinded to the conventional pathological evaluations, and LNs were classified as normal (N-) or positive (N+) with a distinction between macrometastases (> 2 mm) and micrometastases (0.2 to 2 mm). Table 1 summarizes the characteristics of the samples and the main results. On FF-OCT images, the adipose tissue appeared as a honeycomb-like black matrix, facilitating the straightforward identification of adipose-only samples (4/74). Normal lymphatic tissue appeared as round and nodular large areas of homogeneous cells with a dark gray color on images, sometimes separated by thin linear white lines corresponding to vessels and connective tissue (Fig. 2). On the other hand, metastatic LNs appeared as patchy, highly cellular areas forming nodular or pseudonodular foci with a heterogeneous and a lighter gray color compared to normal lymphatic tissue, corresponding to infiltrating cells. Metastases of various sizes were observed, ranging from small micrometastases in nodular patterns (Fig. 3) to more advanced metastatic infiltration, sometimes obliterating the entire sample (Fig. 4). Based on these initial observations, the reader noted potential pitfalls, particularly in highly cellular FF-OCT samples, where normal lymphatic tissue may show a similar grayish infiltrate and could potentially be mistaken for cancer. In that case, the presence of irregular fibrosis (which appeared as brighter – whiter – linear trabecular areas on FF-OCT images) surrounding the metastatic tissue was also instrumental in locating cancer foci within the specimens. Recognizing the normal architecture and physiological compartments of the LN was critical, as capsular infiltration, hilum effacement, replacement of lymphoid follicles, and disruptions of the paracortical or interfollicular areas were particularly helpful in detecting areas of neoplastic invasion.

Table 1.

Comparison between FF-OCT and H&E histology.

				Histology (H&E)			FF-OCT			H&E/FF-OCT
Patient ID	Lymph node ID	Type of tumor	FIGO* final stage	Healthy	Macrometastasis	MicrometastasiS	Healthy	Macrometastasis	MicrometastasiS	Spatial correspondence
1	1	OC	IIIC1	1	0	0	1	0	0	NO
1	2	OC	IIIC1	0	1	0	0	1	0	YES
2	3	EC	IA	1	0	0	1	0	0	YES
2	4	EC	IA	AT	/	/	AT	/	/	YES
2	5	EC	IA	1	0	0	1	0	0	YES
3	6	EC	IA	1	0	0	1	0	0	YES
3	7	EC	IA	1	0	0	1	0	0	YES
4	8	EC	IB	1	0	0	1	0	0	YES
4	9	EC	IB	1	0	0	1	0	0	YES
5	10	CC	IIIC1	1	0	0	1	0	0	YES
5	11	CC	IIIC1	1	0	0	1	0	0	YES
5	12	CC	IIIC1	1	0	0	1	0	0	YES
6	13	EC	IIC	1	0	0	1	0	0	YES
6	14	EC	IIC	2	LYMPHOMA	0	2	LYMPHOMA	0	YES
6	15	EC	IIC	1	0	0	1	0	0	YES
6	16	EC	IIC	1	0	0	1	0	0	YES
6	17	EC	IIC	1	0	0	1	0	0	YES
6	18	EC	IIC	1	0	0	1	0	0	YES
6	19	EC	IIC	1	0	0	1	0	0	YES
7	20	EC	IB	1	0	0	1	0	0	YES
7	21	EC	IB	AT	/	/	/	/	/	YES
7	22	EC	IB	1	0	0	1	0	0	YES
7	23	EC	IB	1	0	0	1	0	0	YES
7	24	EC	IB	1	0	0	1	0	0	YES
7	25	EC	IB	1	0	0	1	0	0	YES
8	26	EC	IA	1	0	0	1	0	0	YES
8	27	EC	IA	1	0	0	1	0	0	YES
8	28	EC	IA	1	0	0	1	0	0	YES
8	29	EC	IA	1	0	0	1	0	0	NO
9	30	EC	IIC	1	0	0	1	0	0	YES
9	31	EC	IIC	1	0	0	1	0	0	NO
10	32	EC	IIIB	1	0	0	1	0	0	YES
10	33	EC	IIIB	1	0	0	1	0	0	YES
10	34	EC	IIIB	1	0	0	1	0	0	NO
10	35	EC	IIIB	1	0	0	1	0	0	YES
10	36	EC	IIIB	1	0	0	1	0	0	YES
10	37	EC	IIIB	1	0	0	1	0	0	NO
10	38	EC	IIIB	1	0	0	1	0	0	YES
10	39	EC	IIIB	1	0	0	1	0	0	YES
11	40	CC	IIIC2	0	1	1	0	1	1	YES
11	41	CC	IIIC2	1	0	0	1	0	0	YES
11	42	CC	IIIC2	1	0	0	1	0	0	YES
11	43	CC	IIIC2	0	1	0	1	1	0	YES
11	44	CC	IIIC2	0	1	0	0	1	0	YES
11	45	CC	IIIC2	0	1	0	0	1	0	YES
11	46	CC	IIIC2	0	1	0	0	1	0	YES
11	47	CC	IIIC2	1	0	0	0	1	0	NO
12	48	OC	IIIC1	AT	/	/	AT	/	/	NO
12	49	OC	IIIC1	1	0	0	1	0	0	YES
12	50	OC	IIIC1	AT	/	/	/	/	/	YES
13	51	EC	IA	1	0	0	1	0	0	YES
13	52	EC	IA	1	0	0	1	0	0	YES
13	53	EC	IA	1	0	0	1	0	0	YES
14	54	OC	IA	1	0	0	1	0	0	YES
14	55	OC	IA	1	0	0	1	0	0	YES
14	56	OC	IA	1	0	0	1	0	0	NO
14	57	OC	IA	1	0	0	1	0	0	YES
14	58	OC	IA	1	0	0	1	0	0	YES
14	59	OC	IA	1	0	0	1	0	0	YES
15	60	EC	IIIC1	1	0	0	1	0	0	NO
15	61	EC	IIIC1	1	0	0	1	0	0	YES
15	62	EC	IIIC1	1	0	0	1	0	0	YES
16	63	EC	IC	1	0	0	1	0	0	YES
16	64	EC	IC	1	0	0	1	0	0	YES
17	65	EC	IIIC1	0	0	1	1	0	0	YES
17	66	EC	IIIC1	1	0	0	1	0	0	YES
18	67	OC	IIIC2	0	1	1	0	1	1	YES
19	68	OC	IIIC2	0	1	0	0	1	0	YES
19	69	OC	IIIC2	0	1	0	0	1	0	YES
19	70	OC	IIIC2	0	1	0	0	1	0	YES
19	71	OC	IIIC2	0	1	0	0	1	0	YES
19	72	OC	IIIC2	1	0	0	1	0	0	YES
20	73	EC	IA	1	0	0	1	0	0	NO
20	74	EC	IA	0	1	0	0	1	0	NO

Cancer types: EC = endometrial cancer, OC = ovarian cancer, CC = cervical cancer.

Fig. 2.

Normal lymph node seen on FF-OCT and histology images. Digital H&E slides, x2 magnification (A), with two close-up views at x40 magnification (red box = B and black box = C), showing normal adipose tissue (large arrow), vessels (thin arrow), and normal lymphoid tissue (star). FF-OCT image of the same regions (D) with two close-up views (red box = E and white box = F), showing adipose tissue (large arrow) with a “black honeycomb” appearance, connective tissue and vessels appearing as white linear areas (thin arrow), and normal lymphoid tissue (star) appearing as homogeneous nodular regions of small regular cells with a dark gray color.

Fig. 3.

Positive lymph nodes (N+) with a macrometastasis seen on FF-OCT and histology images in three patients with cervical cancer^1,2 &³. Digital H&E slides, x2 magnification (A) with a close-up view (black rectangle = C) showing a metastatic infiltration (star) of the lymph node. FF-OCT image (B), similar section, with a close-up view (black rectangle = D) showing a metastatic infiltration of the lymph node presenting as a heterogeneous area of high cellular density (star) and irregular peri-tumoral fibrosis appearing as white irregular pseudo-linear trabecular areas (large arrow).

Fig. 4.

Positive lymph node (N+) with a micrometastasis (< 2 mm) seen on FF-OCT and histology images in a patient with cervical cancer. Digital H&E slides, x40 magnification (C) with a close-up view (black rectangle = A) showing a micrometastasis. FF-OCT image (D), similar section, with a close-up view (black rectangle = B) showing a micro-metastasis in the lymph node presenting as a heterogeneous nodular area of high cellular density contrasting with the darker surrounding lymphoid tissue.

The performance metrics were as follows: sensitivity 92.3% (95% CI 63.9–99.8), specificity 98.2% (95% CI 90.6–99.9), PPV 85.3% (95% CI 45.4–97.6), and NPV 99.1% (95% CI 94.5–99.8). Overall accuracy was 97.6% (95% CI 90.3–99.8). Furthermore, a direct comparison between the digital FF-OCT and digital histological images revealed spatial correspondence in 63/74 cases (85.1%). The normal structures of LNs, such as the continuous fibrous capsule, cellular lymphoid follicles, and lymphatic sinuses, were clearly identifiable.

Discussion

We demonstrated excellent concordance between FF-OCT images and definitive histopathological diagnosis in detecting LN macro- and micrometastases in gynecological cancers. Our findings support the usefulness of this novel imaging technique to guide intraoperative decision-making in gynecologic cancer surgery. FF-OCT identified LN macro- and micrometastases with an accuracy of 97.6%, a sensitivity of 92.3%, and a specificity of 98.2%, all within minutes, from fresh, unstained tissue. Interpretation did not require an extensive learning curve from the experienced pathologists, and both machine setup and sample insertion were technically straightforward even for non-pathologists. Although some practice was necessary, only 6 cases were excluded due to insufficient FF-OCT image quality during the pilot phase.

Currently, extemporaneous analysis of frozen sections is the most commonly used method to guide intraoperative decisions, especially in cervical and endometrial cancers^17–19. However, the accuracy of frozen section analysis (excluding isolated tumor cells) is approximately 72%^4,20. Furthermore, when using sentinel LN techniques, histology shows a lack of LN parenchyma in 7–20% of specimens⁷, particularly when adipose tissue is overrepresented⁸, which is a clinical issue as re-intervention or adjuvant treatment are considered in such cases. Additionally, other techniques (e.g., bioimpedance spectroscopy, ultraviolet photoacoustic microscopy, micro-CT, etc.) have been evaluated in preliminary studies, but have not yet achieved extensive clinical validation. Similarly, a one-step nucleic acid amplification method—a rapid assay that detects cytokeratin-19 mRNA in LN tissue by brief tissue homogenization followed by amplification directly from the lysate¹⁹—has recently been proposed for diagnosing LN metastases^21,22. However, it has notable limitations, including high cost, technical complexity, and the destructive nature of the sample processing, which precludes any subsequent definitive histopathological analysis²³.

In 2010, the use of optical coherence tomography (OCT) for perioperative LN analysis was first reported in breast cancer patients, emphasizing the need for alternative approaches to selective LN removal due to the low number of metastatic LNs obtained with extensive lymphadenectomy procedures²⁴. Shortly thereafter, another study demonstrated the ability of OCT to provide detailed morphological descriptions of axillary LN metastases¹⁶. Similarly, the potential of FF-OCT to detect LN metastasis in breast cancer was confirmed, with a sensitivity of 81.3% and a specificity of 90.3%²⁵, similar to the sensitivity and specificity we obtained in gynecological malignancies. Based on OCT use for guidance in oncologic surgery¹¹, we compared FF-OCT for intraoperative microscopic extemporaneous LN assessment in a pilot between normal and metastatic LNs in gynecologic cancer⁹.

The ability to image the entire LN without the need for tissue-specific preparation, despite some loss of resolution at greater depths (> 100 μm), provides a rapid alternative to frozen section pathology. Although direct comparisons between FF-OCT imaging and traditional histology are challenged by inherent tissue processing and imaging artifacts, this study demonstrates that such comparisons are easy to implement. FF-OCT provides high-resolution imaging with sufficient detail for rapid screening for the presence or absence of metastases. Identifying lesions < 0.2 mm remains a challenge in FF-OCT interpretation. However, the clinical significance of isolated tumor cells in LNs continues to be debated in gynecological cancers. The pathologist involved in FF-OCT image reading had no prior experience with this imaging modality yet was able to interpret the images accurately after a brief familiarization phase, documenting a rapid learning curve.

This study has some limitations. As per the pilot study design, the FF-OCT acquisition was limited to a section of the LNs that was subject to comparative H&E examination. A full optical ultrastaging approach was not performed, which may affect the detection of isolated tumor cells and micrometastases. Moreover, implementing a complete volumetric scan to replicate histological ultrastaging would result in a proportional increase in acquisition time, potentially exceeding the 15-minute per-node.

Additionally, six LNs were excluded due to suboptimal image quality caused by early-stage calibration issues and suboptimal sample positioning during the initial setup phase, reflecting a short technical learning curve.

Recent advances in the FF-OCT technique, such as the addition of dynamic cell imaging (in which a time series is captured for each voxel of interest over a few seconds, followed by processing to extract distinctive features indicative of subcellular activity), have significant potential to provide greater cellular detail and highlight the intense metabolic activity of tumor cells, as demonstrated on breast cancer samples^14,15. Added dynamic cell imaging within the same sample assessment may address some of the limitations of pure FF-OCT in identifying isolated tumor cells, although a clear correlation between image features and cellular characteristics has yet to be established¹⁵. While our analysis was based on qualitative interpretation of FF-OCT images, ongoing work explores quantitative and automated approaches to enhance diagnostic precision. Methods such as the speckle-contrast coefficient²⁶ and the signal attenuation coefficient²⁷ have shown strong potential in OCT-based tissue characterization, including in gynecologic oncology. Applying these metrics to FF-OCT could support objective classification of lymph node metastases and facilitate integration with artificial intelligence-based diagnostic tools. Similarly, the recent integration of artificial intelligence into digital pathology, such as in the Camelyon dataset (limited to H&E-stained images)²⁸ or on preclinical images of dermatological and ovarian cancers^28–32 has shown promising results, and such a technique applied to FF-OCT images could potentially transform intraoperative LN assessment. Ideally, such a tool should be implemented intraoperatively by non-pathologist operators, whose time and expertise are more appropriately reserved for definitive pathological assessment, including labor-intensive procedures such as ultrastaging. This consideration underscores the importance of minimizing the learning curve and highlights the promising role of deep learning in facilitating interpretation.

Conclusion

Comparative analysis of optical images obtained using FF-OCT on fresh ex vivo LNs and their corresponding definitive histological evaluation revealed a high degree of accuracy in identifying metastatic foci measuring ≥ 0.2 mm. This strong correlation supports the use of FF-OCT as a rapid and reliable intraoperative diagnostic tool to assess the metastatic status in LNs, thereby enhancing decision making in surgical oncology.

Author contributions

A.V. and D.Q. developed the study design. M.P., L.R., A.V. and D.Q. developed the study methodology. M.P., L.L., E.C., and C.A. were in charge of specimen retrieval. M.P., L.R., and B.S. performed specimen processing and FF-OCT image acquisition. P.L.Q, C.T., and A.N. performed the histopathological examination. D.G. performed the statistical analysis. L.R. was in charge of visualization. M.P. and L.R. drafted the manuscript. P.L.Q, T.L., A.N., B.S., N.B., G.S., D.Q., and A.V. revised the manuscript. All authors reviewed and approved the final manuscript for submission.

Data availability

All data generated or analyzed in this work are included in this article and/or its figures. Further enquiries can be directed to the corresponding author.

Declarations

Competing interests

Barbara Seeliger declares that she is the recipient of a grant from the French National Agency for Research (Agence Nationale de la Recherche (ANR), 86 rue Regnault, 75013 Paris, France) within the framework of the project AI-DIAL—Diagnostic Imaging of Adrenal Lesions (ANR-22-CE17-0019-01 and ANR-23-IACL-0004) and has a Consulting agreement with Intuitive Surgical. All other authors declare no competing interests.