Open Access
Add to BibSonomy
Abstract
Histopathology based on formalin-fixed and paraffin-embedded tissues has long been the gold standard for surgical margin assessment (SMA). However, routine pathological practice is lengthy and laborious, failing to guide surgeons intraoperatively. In this report, we propose a practical and low-cost histological imaging method with wide-field optical-sectioning microscopy (i.e., High-and-Low-frequency (HiLo) microscopy). HiLo can achieve rapid and non-destructive imaging of freshly-excised tissues at an extremely high acquisition speed of 5 cm2/min with a spatial resolution of 1.3 µm (lateral) and 5.8 µm (axial), showing great potential as an SMA tool that can provide immediate feedback to surgeons and pathologists for intraoperative decision-making. We demonstrate that HiLo enables rapid extraction of diagnostic features for different subtypes of human lung adenocarcinoma and hepatocellular carcinoma, producing surface images of rough specimens with large field-of-views and cellular features that are comparable to the clinical standard. Our results show promising clinical translations of HiLo microscopy to improve the current standard of care.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Histopathology has long remained the gold standard for surgical margin assessment (SMA). However, routine pathological examinations based on thin tissue slices (typically ∼4–7 µm) sectioned from formalin-fixed and paraffin-embedded (FFPE) tissues cause a significant delay in generating accurate diagnostic reports, failing to guide surgeons intraoperatively. Although frozen section can serve as an intraoperative alternative to FFPE, it still requires a turnaround time of 20 to 30 minutes. Besides, the frozen section is subjected to an inadequate sampling of resection margins as only a small portion of the resected tissue is frozen, or only a few thin tissue slices are sectioned and visualized. Furthermore, freezing artifacts are inevitable when dealing with edematous and fatty tissues, affecting pathological interpretation and diagnostic accuracy [1]. it is reported that over 20% of breast cancer patients treated with breast-conserving surgery will undergo repeated surgeries [2] due to the positive margins observed in the postoperative FFPE histology. Therefore, developing a rapid and accurate tissue assessment tool that can be used by surgeons and pathologists intraoperatively is of great significance.
Lots of efforts have been spent to improve routine clinical practice. Among them, optical imaging has demonstrated significant success in in vivo and ex vivo tissue imaging due to its rapid and non-invasive nature. Advanced microscopy techniques with optical sectioning capability can image a thin layer of a freshly excised tissue without the need for physically sectioning the specimen, greatly simplifying the procedures associated with slide preparation in conventional FFPE histology. Detailed comparisons between the state-of-art non-destructive imaging techniques in terms of imaging throughput, system complexity, and ease of use are shown in Fig. S1. In summary, the scanning-based depth-resolved imaging techniques, including reflectance and fluorescence confocal microscopy [3–5], photoacoustic microscopy [6,7], optical coherence tomography [8–10], stimulated Raman scattering [11,12], and nonlinear microscopy [13,14], have demonstrated promising results in imaging human biopsies. However, the requirement of sequential beam scanning limits the imaging throughput of these methods, posing a challenge to examine large resection specimens within a short diagnostic timeframe. In contrast, wide-field imaging techniques which enable parallel pixel acquisition are advantageous for clinical applications due to their rapid imaging speed and low system complexity. Microscopy with ultraviolet surface excitation (MUSE) [15,16], open-top light-sheet microscopy (OTLS) [17,18], and structured illumination microscopy (SIM) [19,20], are examples under this category. MUSE relies on the limited penetration depth of ultraviolet light to achieve moderate optical sectioning, which exhibits some variations between different types of tissues [21]. OTLS is a versatile microscopy technique that enables rapid surface imaging and volumetric imaging of large specimens with and without tissue clearing, respectively. However, oblique illumination will generate severe aberrations at the air-glass-water interface due to wavefront- and index-mismatching [22], which have to be carefully corrected. Besides, axial resolution is inherently traded with field-of-view (FOV) in Gaussian light-sheet systems. In comparison, SIM enables digital rejection of out-of-focus background by leveraging the fact that only in-focus components can be modulated by the structured pattern. Although SIM is unable to obtain high-quality optical sections deep into the tissue [23] since the pattern’s contrast is rapidly deteriorated due to scattering, it is still a light-efficient imaging technique that has demonstrated promising clinical impact in the diagnosis of prostate cancer [20,24].
Here, we propose a practical tissue scanner for intraoperative SMA with High-and-Low-frequency (HiLo) microscopy [25,26]. HiLo is a double-shot optical sectioning technique that relies on the acquisition of two images, one with speckle illumination and one with uniform illumination, eliminating the need for generating well-defined and controlled patterns as required by conventional SIM. In addition, HiLo is robust to aberrations and scattering in tissues as the fully-developed speckles are statistically invariant [27]. We validate that HiLo can generate high-quality optically-sectioned images at the surface of freshly-excised and rough tissues, enabling rapid, slide-free, and non-destructive imaging at an acquisition speed of 5 cm2/min per fluorescence channel with a spatial resolution of 1.3 µm (lateral) and 5.8 µm (axial), which is competent to reveal subcellular diagnostic features and provide immediate on-site feedback. We demonstrate that HiLo reveals diagnostic features of different subtypes of human lung adenocarcinoma and hepatocellular carcinoma, producing images with remarkably recognizable cellular features comparable to the gold standard histology, showing its great potential as an assistive imaging platform for surgeons and pathologists to provide optimal adjuvant treatment intraoperatively. Our results show promising clinical translations of HiLo microscopy to improve the current clinical practice in pathological examination.
2. Methods
2.1 Theoretical background of HiLo microscopy
Optical sectioning through HiLo microscopy has been reported in detail elsewhere [25,26]. In brief, HiLo requires two images to obtain one optically sectioned image. A uniform-illumination image (${I_u}$) is used to provide high-frequency (Hi) components whereas a speckle-illumination image (${I_s}$) is used to provide low-frequency (Lo) components of the final image. The fusion of these two images will produce a full-resolution optically-sectioned image ${I_{HiLo}}$, which can be calculated as
It is well known that the intensity of high-frequency components is attenuated more rapidly than that of low-frequency components with the increase of defocus. As a result, high-frequency components are imaged with high contrast only at the focal plane. Therefore, high-frequency components are naturally optically sectioned and can be extracted from ${I_u}$ via simple high-pass filtering,
The low-frequency components can be calculated with a complementary low-pass filter LP as
The decay efficiency of ${C_s}$ can be further accelerated by applying an additional band-pass filter to the difference image before contrast evaluation. This band-pass filter can be generated by subtracting two Gaussian low-pass filters, which is calculated as
By setting ${k_c}$ to approximately 0.18σ [26], the axial resolution of HiLo can be controlled by only changing the optical sectioning parameter σ (e.g., Fig. 3(k)). The numerical processing can be done in nearly real-time with a workstation using a Core i9-10980XE CPU @ 4.8 GHz and 8 × 32GB RAM, and 4 NVIDIA GEFORCE RTX 3090 GPUs. The image processing pipeline of HiLo is shown using a freshly excised mouse heart as an example (Fig. S2), and the MATLAB code can be found at https://github.com/TABLAB-HKUST/HiLo.
2.2 Experimental setup of the HiLo imaging system
The HiLo system is configured in epi-illumination mode (Fig. 1) that allows convenient sample loading and accommodates tissues of any size and thickness. Two laser diodes (CPS 532 and CPS635S, Thorlabs Inc.), emitting at 532 nm and 635 nm, are implemented in the system for simultaneous excitation of cytoplasmic and nuclear channels. These two coherent sources are combined by a short-pass dichroic beamsplitter (FF556-SDi01, Semrock Inc.) and propagated through a variable beam expander (BE052-A, 0.5–2× zoom, Thorlabs Inc.). The resulting beam is illuminated on a diffuser (DG10-1500, Thorlabs Inc.) with a beam diameter of ∼3 mm, generating a coarse speckle pattern with an averaged grain size of ∼4 µm to maintain high image contrast in thick and scattering tissues (see section 2.3). The diffuser is inserted in a home-built motor-driven rotating mount which can be rapidly rotated at a speed of 8000 revolutions per minute to generate uniform illumination. The illumination patterns are reflected by a dual-edge dichroic beamsplitter (FF560/659-Di01, Semrock Inc.) and introduced to the objective lens’s back focal plane through a condenser lens (LB1723, f = 60 mm, Thorlabs Inc.), ensuring that the speckle size remains relatively constant along with the axial propagation of light. The specimen is illuminated from the bottom with an excitation power of 4 mW, and the excited fluorescence is detected by an inverted microscope which consists of a plan achromat objective lens (Plan Fluor, 10×/0.3 NA, Olympus) and an infinity-corrected tube lens (TTL180-A, Thorlabs Inc.). The fluorescence from two channels is successively filtered by two band-pass filters (FF01-579/34-25 and BLP01-647R-25, Semrock Inc.), and imaged by a monochrome scientific complementary metal-oxide-semiconductor camera (PCO edge 4.2, 2048 × 2048 pixels, PCO. Inc.) which can theoretically reach an imaging throughput of ∼800 megabytes/s. In our experiments, the specimen is raster-scanned with 10% overlapping by a 2-axis motorized stage (L-509.20SD00, maximum velocity 20 mm/s, PI miCos GmbH) in two scanning cycles, i.e., one with a static diffuser for speckle illumination and the other with a rapidly rotating diffuser for uniform illumination. The captured image mosaics are stitched by our developed image stitching software, and the whole system is synchronized by the lab-designed LabVIEW program (National Instruments Corp.). The exposure time for each raw fluorescence image is set to fill the full dynamic range of 16-bit camera without intensity saturation, which is sample-dependent and generally less than 100 ms in our experiments. The stage settling time between two successive images is set to 10 ms.
Fig. 1. Schematic of the HiLo system for rapid histological imaging. Two laser diodes are combined by a dichroic beamsplitter, propagated through a variable beam expander, and are subsequently illuminated on a rotatable diffuser to generate uniform and speckle illuminations. The resulting illumination patterns are illuminated onto the bottom surface of the specimen through a condenser lens and an objective lens. The excited fluorescence from the cytoplasmic and nuclear channels is successively collected by the same objective lens, refocused by a tube lens, spectrally filtered, and finally imaged by a monochromatic camera. The specimen is supported by a sample holder and raster-scanned by the motorized stages. LD, laser diode; DB, dichroic beamsplitter; BE, beam expander; Dif, diffuser; CL, condenser lens; OL, objective lens; TL, tube lens; SF, spectral filter; SH, sample holder; TS, translational stage.
2.3 Determination of the optimal speckle frequency for thick tissue imaging
It has been reported that strong optical scattering in thick tissues poses a challenge for HiLo imaging due to the reduced in-focus speckle contrast [28], which serves as a weighting function to numerically filter the out-of-focus fluorescence contributions from the specimen. This effect can be alleviated by illuminating with a coarse speckle pattern at the cost of a compromised optical sectioning strength. Since the degradation of speckle contrast is mainly due to the strong tissue scattering, therefore, the highly scattered brain tissue, instead of other organs such as the liver and kidney, was selected to determine the optimal speckle frequency to achieve robust imaging performance in thick and opaque tissues. A formalin-fixed mouse brain was sectioned by a vibratome to obtain a thin tissue slice (∼10-µm thick) and a thick tissue block (∼5-mm thick). Both of them were processed with nuclear staining for HiLo imaging, and subsequently processed by a standard histological protocol to obtain the corresponding clinical standard image for comparison. The tissues were illuminated by a series of speckle patterns with varying frequencies by changing the beam diameter illuminated on the diffuser, and imaged with a 0.3-NA objective lens. The speckle contrast Cs is calculated from Eq. (4) with a window size of 10 µm × 10 µm, which is automatically determined in the algorithm with an input optical sectioning parameter σ (here σ is set to 1). The mean value of Cs within the region of interest (e.g., Fig. 2(a),(b)) is used for the quantitative analysis in Fig. 2. As expected, the speckle contrast degrades quickly with the increase of speckle frequency (${f_s}$), and this effect is worsened in thick tissues which can cause an underestimation of the weighting function even though the specimen is in focus. Besides, for a given sample thickness, ${C_s}$ is low for a high-NA objective since its narrow depth-of-field (DOF) will further exacerbate the out-of-focus fluorescence [28]. To make a good balance between sectioning capability and signal-to-noise ratio (SNR) of the HiLo images, a 10×/0.3NA objective lens is implemented, and the speckle frequency is set to 250 mm−1 in this study, which is corresponding to an averaged speckle size of ∼4 µm. Since the mean value of Cs is reported in Fig. 2, the result is primarily affected by tissue scattering instead of nuclear size and specific nuclear features. This can be also validated from Sections 3.2 and 3.3 that the selected illumination frequency based on the brain tissue can achieve robust performance in different types of tissues with various pathological features.
Fig. 2. Experimental determination of the optimal speckle frequency for HiLo imaging of thick tissues. a,b, Uniformly-illuminated images of thin (∼10-µm thick) and thick (∼5-mm thick) mouse brain tissues, respectively. c,d, Speckle-illuminated images of the thin and thick mouse brain tissues with a speckle frequency of 150 mm−1. e,f, Speckle-illuminated images of the thin and thick mouse brain tissues with a speckle frequency of 500 mm−1. g, The corresponding FFPE histology image sectioned at the surface of b with a thickness of 7 µm. h, The relationship between speckle frequency and speckle contrast. Iu, uniform-illumination image; Is, speckle-illumination image; fs, speckle frequency; Cs, speckle contrast.
Fig. 3. Experimental characterization of HiLo’s axial resolution. a, Maximum intensity projection of the HiLo images of 10-µm-diameter fluorescent microspheres. b–e, Zoomed-in HiLo images of the white dashed region marked in a at different axial depths. f–i, The corresponding uniformly illuminated wide-field images. j, Axial intensity distribution of a selected microsphere (indicated by the orange box in b–e) with an axial step of 0.5 µm. k, The relationship between axial resolution and optical section parameter σ. The experiment was performed with a 10×/0.3NA objective lens at the 532-nm excitation wavelength and 560-nm fluorescent emission wavelength.
2.4 Quantification of the system axial resolution
Fluorescent beads with a diameter smaller than the resolution limit of a microscope are commonly used to experimentally determine the system’s axial resolution. However, this method will collide with the filtering process in HiLo microscopy which achieves optical sectioning by evaluating the speckle contrast over a sampling window containing several imaged grains. Alternatively, we quantify HiLo’s axial resolution by imaging 10-µm-diameter fluorescent microspheres [29], and the resulting axial resolution can be calculated as:
where $FWH{M_{measured}}$ is the full width at half maximum (FWHM) of the measured optical sectioning curve and ${d_{bead}}$ is the diameter of fluorescent microspheres, which is 10 µm according to the specifications. The specimen is axially scanned over a total range of 50 µm with a step of 0.5 µm (the maximum intensity projection image is shown in Fig. 3(a)). The HiLo images (Fig. 3(b)–(e)) and wide-field images (Fig. 3(f)–(i)) at different axial scanning depths are compared. The curve of the axial intensity distribution of a selected fluorescent microsphere (indicated by the orange box in Fig. 3(b)–(e)) is shown in Fig. 3(j), with a measured FWHM of 11.6 µm, corresponding to an axial resolution of 5.8 µm, which is competent to produce an optical section instead of a physical section in slide-based FFPE histology. In addition, different sectioning capabilities can be obtained from the same raw dataset by adjusting the optical section parameter σ during numerical processing [26] (Fig. 3(k)), which is the unique feature of HiLo microscopy. The image is sampled at 0.65 µm/pixel with 10× magnification, leading to a Nyquist-limited lateral resolution of 1.3 µm. Still, HiLo can provide sufficient resolution to reveal subcellular features which are essential for diagnosis.3. Results
3.1 Experimental protocols
3.1.1 Collection of animal and human tissues
The organs were extracted from C57BL/6 mice. The heart, brain, kidney, and liver were harvested immediately after the mice were sacrificed. All experiments were carried out in conformity with a laboratory animal protocol approved by the Health, Safety and Environment Office of the Hong Kong University of Science and Technology (HKUST). The human sample protocol was approved by the Institutional Review Board at the Prince of Wales Hospital. The imaged tissue was considered as leftover tissue, i.e., it represents a portion of a collected specimen that is not needed for assessment of diagnostic, prognostic, and other parameters in the diagnosis and treatment of the patient.
3.1.2 Tissue processing
For imaging experiments, the specimens were stained with a cocktail containing 10 µM of TO-PRO3 (T3605, Thermo Fisher Scientific Inc.) and 2% v/v solution of Eosin Y (E4009, Sigma-Aldrich Inc.) in 1× phosphate-buffered saline (PBS) for 1 minute, and then rinsed in PBS for 30 seconds, blotted dry with laboratory tissues, and immediately imaged by HiLo. TO-PRO3 is selected for nuclear staining since it is a far-red DNA-selective dye that can be paired with other fluorophores for dual-channel fluorescence imaging without spectral crosstalk. After imaging, the specimens were processed following the standard protocol to obtain the hematoxylin and eosin (H&E)-stained histological images. Specifically, the specimens were processed for dehydration, clearing, and infiltration by a tissue processor (Revos, Thermo Fisher Scientific Inc.) for 12 hours, and then paraffin-embedded as block tissues which were subsequently sectioned at the surface with a thickness of 7 µm by a microtome (RM2235, Leica Microsystems Inc.). After that, the thin tissue slices were stained by H&E, and imaged by a digital slide scanner (NanoZoomer-SQ, Hamamatsu Photonics K.K.) to generate the corresponding histological images. Note that although TO-PRO3 and eosin can serve as a fluorescent analog to H&E histological staining, we and other groups found that eosin tends to leak out of the tissue during imaging since it is weakly bound to fresh and hydrated specimens [18,30], generating unwanted background fluorescence which deteriorates the image contrast. For simplicity, we mainly preserve the nuclear channel in this study.
3.2 Validation on freshly-excised mouse tissues
Freshly excised mouse tissues, including heart (Fig. S3), brain (Fig. 4(a)–(c)), kidney (Fig. 4(d),(e)), and liver (Fig. 4(f),(g)), are manually sectioned with a thickness of ∼5 mm by a scalpel, and stained and imaged by HiLo to validate the system’s performance initially. After imaging, the specimens are histologically processed to obtain the corresponding H&E-stained images for comparisons. It can be observed that the cell nuclei in the hippocampus of the brain (Fig. 4(b)), glomerular capsules in the kidney (Fig. 4(e)), and hepatocytes in the liver (Fig. 4 g) can be clearly resolved with a significantly improved image contrast. The nuclear features, such as cross-sectional area and intercellular distance, are extracted for the quantitative comparison between HiLo and FFPE histology. To extract the distributions of nuclear features, both HiLo and H&E-stained images are segmented by a Fiji plugin (trainable Weka segmentation [31]), and subsequently binarized and analyzed to acquire the cross-sectional area and centroid of each cell nucleus. With the localized center positions of cell nuclei, the intercellular distance is calculated to be the shortest adjacent distance to a neighboring cell nucleus. The statistical results (Fig. 4 h,i), which are calculated from 50 hepatocytes in Fig. 4 g, suggest that the cellular features extracted from HiLo agree fairly well with that from the FFPE histology, validating the accuracy of the information that can be provided by HiLo.
Fig. 4. HiLo imaging of freshly excised mouse tissues. a, HiLo image of a mouse brain. b,c, Zoomed-in widefield images (left), HiLo images (middle), and the corresponding H&E-stained images (right) of the dashed regions marked in a, respectively. d, HiLo image of a mouse kidney. e, Zoomed-in widefield image (left), HiLo image (middle), and the corresponding H&E-stained image (right) of the dashed region marked in d. f, HiLo image of a mouse liver. g, Zoomed-in widefield image (left), HiLo image (middle), and the corresponding H&E-stained image (right) of the dashed region marked in f. h,i, Distributions of nuclear features extracted from g. Wilcoxon rank-sum testing is carried out across groups with n = 50 for each distribution. The significance is defined as p* ≤ 0.05 in all cases. Scale bars: 1 mm (a,d,f) and 100 µm (the remaining).
3.3 Validation on formalin-fixed human cancer tissues
To demonstrate the full potential of HiLo in the application of intraoperative SMA, formalin-fixed and unprocessed human lung tissues (Fig. 5, Visualization 1) and human liver tissues (Fig. 6, Visualization 2) are imaged. The specimens were fixed in formalin immediately after surgical excision to prevent tissue degradation. After HiLo imaging, the specimens were histologically processed to obtain the H&E-stained images for comparison. Note that the H&E-stained thin tissue sections are not able to exactly replicate the surface imaged by HiLo since the tissue might be distorted or differently orientated during the preparation procedures.
Fig. 5. HiLo imaging of human lung cancer tissues. a, HiLo (top) and H&E-stained images (bottom) of a fixed human lung specimen with large cell carcinoma, the dashed lines outline the interface between the normal (right) and tumor (left) regions. The inset at the bottom left of the HiLo image shows the photograph of the specimen. b–d, Zoomed-in HiLo images of the blue, yellow, and green solid regions marked in a, respectively. e–g, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in b–d, respectively. LCN, lymphocyte cell nucleus; TCN, tumor cell nucleus. h, Distribution of nuclear cross-sectional areas of lymphocytes and cancer cells derived from b (N = 50) with a median of 65 µm2 and 240 µm2, respectively. i, Distribution of intercellular distances of lymphocytes and cancer cells derived from b (N = 50) with a median of 9 µm and 16 µm, respectively. j, HiLo (top) and H&E-stained images (bottom) of a fixed human lung specimen with papillary-predominant adenocarcinoma. The inset at the bottom left of the HiLo image shows the photograph of the specimen. k–m, Zoomed-in HiLo images of the blue, yellow, and green solid regions marked in j, respectively. n–p, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in k–m, respectively. q, Combined HiLo and H&E-stained mosaic image of the orange solid region marked in j. r, HiLo (top), surface topology (middle), and H&E-stained images (bottom) of a fixed human lung specimen with acinar-predominant adenocarcinoma. The inset at the bottom right of the HiLo image shows the photograph of the specimen. s,t, Zoomed-in HiLo images of the blue and yellow solid regions marked in r, respectively. u,v, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in s and t, respectively. w,x, Zoomed-in HiLo (left), surface topology (middle), and H&E-stained images (right) of the green and orange solid regions marked in r, respectively. Scale bars: 1 mm (a,j,r), 30 µm (e–g,n–p,u,v), and 100 µm (the remaining). See Visualization 1.
Fig. 6. HiLo imaging of human liver cancer tissues. a, HiLo (top), surface topology (middle), and H&E-stained images (bottom) of a fixed human liver specimen with scirrhous pattern HCC, the dashed lines outline the interface between the normal (left) and tumor (right) regions. The images before and after surface extraction (SE) are separated by the blue solid line in HiLo, and the arrows represent the missing parts due to the specimen’s surface irregularity. The inset at the bottom right of the HiLo image shows the photograph of the specimen. b–d, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in a, respectively. e, HiLo (top) and H&E-stained images (bottom) of a fixed human liver specimen with solid pattern HCC. The inset at the bottom left of the HiLo image shows the photograph of the specimen. f,g, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in e, respectively. h, HiLo (top) and H&E-stained images (bottom) of a fixed human liver specimen with fibrolamellar HCC. The inset at the bottom right of the HiLo image shows the photograph of the specimen. i,j, Zoomed-in HiLo and H&E-stained images of the white dashed regions marked in h, respectively. Unlabeled scale bars: 100 µm. See Visualization 2.
The lung specimen from the first patient (Fig. 5(a)) is a pathologically confirmed case of large cell carcinoma (LCC). The size of the specimen is 11 mm (length) × 6 mm (width) × 1 mm (thickness), and the imaging of the whole tissue surface is completed within 10 seconds. Both the HiLo image (top of Fig. 5(a)) and H&E-stained image (bottom of Fig. 5(a)) outline a clear interface between normal (right) and tumor (left) regions. The zoomed-in HiLo images (Fig. 5(b),(c),e,f) show the large tumor cells with vesicular nuclei growing with abundant lymphocytic infiltration, which is the feature of LCC. Leucocytes close to the tumor margin (Fig. 5(d),(g)) are well identified. The distributions of nuclear features are extracted from Fig. 5(b). The statistical results (Fig. 5 h,i) indicate that the cancer cells and lymphocytes can be well distinguished based on the cellular features. In contrast, the specimens from the second and third patients are confirmed cases of lung adenocarcinoma which presents a different growth pattern compared with LCC. Specifically, the specimen from the second patient is categorized as papillary-predominant adenocarcinoma (Fig. 5(j)), which is characterized by the finger-like papillary architecture with tumor cells lining the surface of branching fibrovascular cores (Fig. 5(k),(n)). The vessels (Fig. 5 l,o) and densely packed tumor cells in micropapillary clusters (Fig. 5 m,p) are revealed with remarkably recognizable features comparable to the H&E-stained images. The specimen from the third patient is diagnosed with acinar-predominant adenocarcinoma (Fig. 5(r)), in which tumor cells are arranged in an acinar pattern spreading on a fibrotic stroma. The specimen has a size of 11 mm (length) × 5 mm (width) × 5 mm (thickness), with a surface irregularity of ∼120 µm. To extract the intact surface profile of the specimen (middle of Fig. 5(r)), the sample is axially scanned with a 10-µm interval and the total imaging time is within 80 seconds. The generated HiLo images at different axial depths are processed by an extended DOF plugin in ImageJ [32] to produce the surface topology of the specimen. The glandular structures are well recognized in Fig. 5(s)–(x), which are well distinguished from papillary adenocarcinoma in terms of the growth pattern. Note that lung adenocarcinoma, which is the most common case of non-small cell lung cancer, is usually found with a mixture of different histologic subtypes. Various diagnostic features in lung adenocarcinoma specimens, including micropapillary cluster (Fig. S4e), pseudostratified columnar cells (Fig. S4f), respiratory bronchiole (Fig. S4g), and alveolar space (Fig. S4i), are shown in Fig. S4 to further validate the effectiveness of the proposed method.
Finally, three liver specimens which are pathologically confirmed cases of hepatocellular carcinoma (HCC) are imaged to show the effectiveness of ex vivo fluorescence histology by HiLo microscopy (Fig. 6). The specimen from the first patient (Fig. 6(a)) is categorized as scirrhous pattern HCC, in which tumor cell nets are separated by rich fibrous connective tissues. The specimen has a size of 12 mm (length) × 4 mm (width) × 5 mm (thickness), with a surface irregularity of ∼200 µm. Similarly, variable focusing with an axial step of 10 µm is performed to obtain the specimen’s surface topology (middle of Fig. 6(a)). The acquisition process is completed within 120 seconds. Several FFPE thin slices are sectioned from the tissue surface to obtain the corresponding features between HiLo and FFPE histology (the slice that is the most representative of the imaged surface is shown at the bottom of Fig. 6(a)). The interface between the normal (left) and tumor (right) regions is easily outlined in the HiLo image (the yellow dashed line). Zoomed-in HiLo images with their corresponding histological images are shown in Fig. 6(b)–(d). The morphology of normal hepatocytes can be well identified (Fig. 6(b)). It can be observed that hepatocytes are replaced by scar tissues with intratumoral collagen in the tumor region (Fig. 6(c)), although it is not intuitive with only nuclear contrast. With 1.3-µm lateral resolution, small lymphocytes accumulated at the circumference of hepatic vasculature are resolved individually (Fig. 6(d)). In contrast, the specimen from the second patient (Fig. 6(e)) is categorized as solid pattern HCC, which comprises sheets of tumor cells with uniform nuclei and conspicuous nucleoli. Lymphocytic infiltration (Fig. 6(f)) and monomorphic tumor cells lacking hepatocyte plates are well identified (Fig. 6 g). The specimen from the third patient (Fig. 6 h) is a confirmed case of fibrolamellar HCC, in which clusters of tumor cells are separated by parallel fibrous lamellae. We use both nuclear and cytoplasmic staining to better show the structures. The tumor cell cords separated by the dense fibrous septa are well identified (Fig. 6(i)), and spindle-shaped fibroblasts lining the thick band of lamellar fibrosis can be clearly distinguished (Fig. 6(j)). In addition, the hepatocytes with nuclear vacuolation are visualized in liver specimens with fatty liver disease (Fig. S5). These results validate that HiLo enables the non-destructive evaluation of large specimens in a realistic time frame, showing its great potential as an intraoperative SMA tool that can be used by surgeons and pathologists to provide the optimal adjuvant treatment.
4. Discussion
HiLo is a promising and transformative imaging technology that enables rapid and non-destructive imaging of large clinical specimens, demonstrating great potential as an assistive imaging platform that can guide surgeons intraoperatively. To the best of our knowledge, this work demonstrates the first application of using speckle illumination microscopy (i.e., HiLo) for rapid diagnosis of different subtypes of human lung adenocarcinoma and hepatocellular carcinoma, producing images with remarkably recognizable cellular features comparable to the gold standard FFPE histology (Figs. 5 and 6). Although formalin-fixed tumor specimens were imaged in this study, fresh tissues are also expected to achieve similar results (e.g., Fig. 4) by ex vivo fluorescence microscopy [31]. Since eosin is weakly bound to cytoplasmic proteins in fresh and hydrated tissues, only nuclear fluorescence channel is implemented in this study to ensure a high image contrast. An optimized dual-channel fluorescence staining through the combination of SYBR Gold (λEx/Em = 495/537 nm, nuclear stain) and ATTO 655 NHS ester (λEx/Em = 655/680 nm, cytoplasmic stain) might be a better option [18], which can be adapted to this work to facilitate robust intraoperative histopathology in the future.
The superiority of HiLo over the gold standard histology is that large-scale surgical specimens can be sufficiently sampled with HiLo microscopy whereas only spot-checked with conventional histology. This is of great significance to improve diagnostic accuracy. HiLo should be able to operate at half of the camera frame rate (100 frames per second in our settings) theoretically. The acquisition speed of the current system is mainly limited by the exposure time for each raw fluorescence image, which is ∼100 ms with an illumination power of 4 mW, corresponding to an imaging speed of 5 cm2/min per channel. Increasing the excitation power can further accelerate the acquisition speed by an order of magnitude. For patients who undergo hepatectomy for HCC, the surface area of the specimens can easily reach dozens of square centimeters, which could be imaged within several minutes with the increased imaging speed.
As shown in Fig. 2, thick specimens will affect the optical sectioning strength of HiLo by reducing the in-focus speckle contrast. This can be alleviated by using coarser speckle patterns, which, however, will lead to a compromised axial resolution. Even though, the current axial resolution of ∼6 µm is sufficient to produce an optical section instead of a physical section in conventional slide-based histology. The lateral resolution, which is currently limited by the sampling pixel size on the detector, will be ultimately determined by the NA of the employed objective lens. Although a lateral resolution of 1.3 µm in this study can resolve the densely packed tumor cells as in lung adenocarcinomas (Fig. 5, Fig. S4), it is not sufficient to reveal nucleolar structures which are associated with the degrees of malignancy. Super-resolution microscopy via synthetic aperture with speckle illumination [32–34] could potentially provide a solution to break this resolution limit to further boost the accuracy of tissue analysis. Another limitation of HiLo is the imaging depth. Similar to SIM, the image SNR will be significantly decreased when imaging deep into the tissue since the shot noise is recorded along with the in-focus signal [35]. This will limit the effectiveness of HiLo for volumetric imaging applications compared with the beam-scanning approaches. Despite this, HiLo is still a cost-effective and light-efficient tool that enables rapid surface imaging of large-scale clinical specimens, particularly favoring the applications such as SMA.
The practicality of the HiLo system can be further improved for clinical translations. First, the narrow DOF (<10 µm with a 0.3-NA objective lens) is not able to accommodate large irregular surfaces, causing the surgical margins to come in and out of focus during the HiLo imaging. Variable focusing is required to obtain the intact tissue surface (Figs. 5(r),6(a)), which will sacrifice the overall imaging throughput. Recently proposed methods for extended DOF via dynamic remote focusing [36] or deep learning [37] hold great promise to speed up the whole imaging process by eliminating the need for axial scanning. Second, the current system can be developed as an endoscope through a flexible fiber bundle [38–40], which could enable in vivo assessment of remaining tissue after tumor excision, facilitating image-guided surgeries which can provide real-time feedback to surgeons for intraoperative decision-making. This will also improve the quality of the treatment with reduced invasiveness.
In summary, we have experimentally demonstrated that HiLo enables rapid diagnosis of different subtypes of human lung adenocarcinoma and hepatocellular carcinoma, producing images with remarkably recognizable cellular features comparable to the gold standard FFPE histology. As a proof of concept, this study is limited by the small number of specimens. Future studies will involve more specimens to quantify the diagnostic metrics (i.e., sensitivity and specificity). Moreover, a computer-aided diagnosis could be incorporated with HiLo to further improve the efficiency of the current pathological workflow.
Funding
Research Grants Council, University Grants Committee (1620862026203619).
Disclosures
V. T. C. T. and T. T. W. W. have a financial interest in PhoMedics Limited, which, however, did not support this work. The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. A. Preeti, G. Sameer, S. Kulranjan, S. A. Abhinav, R. Preeti, Y. Sunita, and G. M. Mati, “Intra-operative frozen sections: experience at a tertiary care centre,” Asian Pac. J. Cancer Prev. 17, 5057 (2016). [CrossRef]
2. F. Fleming, A. D. Hill, E. Mc Dermott, A. O’Doherty, N. O’Higgins, and C. Quinn, “Intraoperative margin assessment and re-excision rate in breast conserving surgery,” Eur. J. Surg. Oncol. 30(3), 233–237 (2004). [CrossRef]
3. D. S. Gareau, Y. Li, B. Huang, Z. Eastman, K. S. Nehal, and M. Rajadhyaksha, “Confocal mosaicing microscopy in Mohs skin excisions: feasibility of rapid surgical pathology,” J. Biomed. Opt. 13(5), 054001 (2008). [CrossRef]
4. M. Ragazzi, S. Piana, C. Longo, F. Castagnetti, M. Foroni, G. Ferrari, G. Gardini, and G. Pellacani, “Fluorescence confocal microscopy for pathologists,” Mod. Pathol. 27(3), 460–471 (2014). [CrossRef]
5. S. Krishnamurthy, A. Cortes, M. Lopez, M. Wallace, S. Sabir, K. Shaw, and G. Mills, “Ex vivo confocal fluorescence microscopy for rapid evaluation of tissues in surgical pathology practice,” Arch. Pathol. Lab. Med. 142(3), 396–401 (2018). [CrossRef]
6. R. Li, P. Wang, L. Lan, F. P. Lloyd, C. J. Goergen, S. Chen, and J.-X. Cheng, “Assessing breast tumor margin by multispectral photoacoustic tomography,” Biomed. Opt. Express 6(4), 1273 (2015). [CrossRef]
7. T. T. W. Wong, R. Zhang, P. Hai, C. Zhang, M. A. Pleitez, R. L. Aft, D. V. Novack, and L. V. Wang, “Fast label-free multilayered histology-like imaging of human breast cancer by photoacoustic microscopy,” Sci. Adv. 3(5), e1602168 (2017). [CrossRef]
8. F. T. Nguyen, A. M. Zysk, E. J. Chaney, J. G. Kotynek, U. J. Oliphant, F. J. Bellafiore, K. M. Rowland, P. A. Johnson, and S. A. Boppart, “Intraoperative evaluation of breast tumor margins with optical coherence tomography,” Cancer Res. 69(22), 8790–8796 (2009). [CrossRef]
9. E. Auksorius, Y. Bromberg, R. Motiejūnaitė, A. Pieretti, L. Liu, E. Coron, J. Aranda, A. M. Goldstein, B. E. Bouma, A. Kazlauskas, and G. J. Tearney, “Dual-modality fluorescence and full-field optical coherence microscopy for biomedical imaging applications,” Biomed. Opt. Express 3(3), 661 (2012). [CrossRef]
10. C. Kut, K. L. Chaichana, J. Xi, S. M. Raza, X. Ye, E. R. McVeigh, F. J. Rodriguez, A. Quiñones-Hinojosa, and X. Li, “Detection of human brain cancer infiltration ex vivo and in vivo using quantitative optical coherence tomography,” Sci. Transl. Med. 7(292), 292ra100-292ra100 (2015). [CrossRef]
11. D. A. Orringer, B. Pandian, Y. S. Niknafs, T. C. Hollon, J. Boyle, S. Lewis, M. Garrard, S. L. Hervey-Jumper, H. J. L. Garton, C. O. Maher, J. A. Heth, O. Sagher, D. A. Wilkinson, M. Snuderl, S. Venneti, S. H. Ramkissoon, K. A. McFadden, A. Fisher-Hubbard, A. P. Lieberman, T. D. Johnson, X. S. Xie, J. K. Trautman, C. W. Freudiger, and S. Camelo-Piragua, “Rapid intraoperative histology of unprocessed surgical specimens via fibre-laser-based stimulated Raman scattering microscopy,” Nat. Biomed. Eng. 1(1), 0001–13 (2017). [CrossRef]
12. T. C. Hollon, B. Pandian, A. R. Adapa, E. Urias, A. V. Save, S. S. S. Khalsa, D. G. Eichberg, R. S. D’Amico, Z. U. Farooq, S. Lewis, P. D. Petridis, T. Marie, A. H. Shah, H. J. L. Garton, C. O. Maher, J. A. Heth, E. L. McKean, S. E. Sullivan, S. L. Hervey-Jumper, P. G. Patil, B. G. Thompson, O. Sagher, G. M. McKhann, R. J. Komotar, M. E. Ivan, M. Snuderl, M. L. Otten, T. D. Johnson, M. B. Sisti, J. N. Bruce, K. M. Muraszko, J. Trautman, C. W. Freudiger, P. Canoll, H. Lee, S. Camelo-Piragua, and D. A. Orringer, “Near real-time intraoperative brain tumor diagnosis using stimulated Raman histology and deep neural networks,” Nat. Med. 26(1), 52–58 (2020). [CrossRef]
13. Y. K. Tao, D. Shen, Y. Sheikine, O. O. Ahsen, H. H. Wang, D. B. Schmolze, N. B. Johnson, J. S. Brooker, A. E. Cable, J. L. Connolly, and J. G. Fujimoto, “Assessment of breast pathologies using nonlinear microscopy,” Proc. Natl. Acad. Sci. U. S. A. 111(43), 15304–15309 (2014). [CrossRef]
14. H. Tu, Y. Liu, D. Turchinovich, M. Marjanovic, J. K. Lyngsø, J. Laegsgaard, E. J. Chaney, Y. Zhao, S. You, W. L. Wilson, B. Xu, M. Dantus, and S. A. Boppart, “Stain-free histopathology by programmable supercontinuum pulses,” Nat. Photonics 10(8), 534–540 (2016). [CrossRef]
15. F. Fereidouni, Z. T. Harmany, M. Tian, A. Todd, J. A. Kintner, J. D. McPherson, A. D. Borowsky, J. Bishop, M. Lechpammer, S. G. Demos, and R. Levenson, “Microscopy with ultraviolet surface excitation for rapid slide-free histology,” Nat. Biomed. Eng. 1(12), 957–966 (2017). [CrossRef]
16. W. Xie, Y. Chen, Y. Wang, L. Wei, C. Yin, A. K. Glaser, M. E. Fauver, E. J. Seibel, S. M. Dintzis, J. C. Vaughan, N. P. Reder, and J. T. C. Liu, “Microscopy with ultraviolet surface excitation for wide-area pathology of breast surgical margins,” J. Biomed. Opt. 24(02), 1 (2019). [CrossRef]
17. A. K. Glaser, N. P. Reder, Y. Chen, E. F. McCarty, C. Yin, L. Wei, Y. Wang, L. D. True, and J. T. C. Liu, “Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens,” Nat. Biomed. Eng. 1(7), 0084 (2017). [CrossRef]
18. Y. Chen, W. Xie, A. K. Glaser, N. P. Reder, C. Mao, S. M. Dintzis, J. C. Vaughan, and J. T. C. Liu, “Rapid pathology of lumpectomy margins with open-top light-sheet (OTLS) microscopy,” Biomed. Opt. Express 10(3), 1257 (2019). [CrossRef]
19. T. C. Schlichenmeyer, M. Wang, K. N. Elfer, and J. Q. Brown, “Video-rate structured illumination microscopy for high-throughput imaging of large tissue areas,” Biomed. Opt. Express 5(2), 366 (2014). [CrossRef]
20. M. Wang, D. B. Tulman, A. B. Sholl, H. Z. Kimbrell, S. H. Mandava, K. N. Elfer, S. Luethy, M. M. Maddox, W. Lai, B. R. Lee, and J. Q. Brown, “Gigapixel surface imaging of radical prostatectomy specimens for comprehensive detection of cancer-positive surgical margins using structured illumination microscopy,” Sci. Rep. 6(1), 1–16 (2016). [CrossRef]
21. T. Yoshitake, M. G. Giacomelli, L. M. Quintana, H. Vardeh, L. C. Cahill, B. E. Faulkner-Jones, J. L. Connolly, D. Do, and J. G. Fujimoto, “Rapid histopathological imaging of skin and breast cancer surgical specimens using immersion microscopy with ultraviolet surface excitation,” Sci. Rep. 8, 1–12 (2018). [CrossRef]
22. A. K. Glaser, N. P. Reder, Y. Chen, C. Yin, L. Wei, S. Kang, L. A. Barner, W. Xie, E. F. McCarty, C. Mao, A. R. Halpern, C. R. Stoltzfus, J. S. Daniels, M. Y. Gerner, P. R. Nicovich, J. C. Vaughan, L. D. True, and J. T. C. Liu, “Multi-immersion open-top light-sheet microscope for high-throughput imaging of cleared tissues,” Nat. Commun. 10(1), 2781 (2019). [CrossRef]
23. J. Mertz, “Optical sectioning microscopy with planar or structured illumination,” Nat. Methods 8(10), 811–819 (2011). [CrossRef]
24. M. Wang, H. Z. Kimbrell, A. B. Sholl, D. B. Tulman, K. N. Elfer, T. C. Schlichenmeyer, B. R. Lee, M. Lacey, and J. Q. Brown, “High-resolution rapid diagnostic imaging of whole prostate biopsies using video-rate fluorescence structured illumination microscopy,” Cancer Res. 75(19), 4032–4041 (2015). [CrossRef]
25. D. Lim, K. K. Chu, and J. Mertz, “Wide-field fluorescence sectioning with hybrid speckle and uniform-illumination microscopy,” Opt. Lett. 33(16), 1819 (2008). [CrossRef]
26. D. Lim, T. N. Ford, K. K. Chu, and J. Mertz, “Optically sectioned in vivo imaging with speckle illumination HiLo microscopy,” J. Biomed. Opt. 16(1), 016014 (2011). [CrossRef]
27. J. W. Goodman, Speckle Phenomena in Optics: Theory and Applications, 2nd Edition (SPIE, 2020).
28. J. Mazzaferri, D. Kunik, J. M. Belisle, K. Singh, S. Lefrançois, and S. Costantino, “Analyzing speckle contrast for HiLo microscopy optimization,” Opt. Express 19(15), 14508 (2011). [CrossRef]
29. K. Philipp, A. Smolarski, N. Koukourakis, A. Fischer, M. Stürmer, U. Wallrabe, and J. W. Czarske, “Volumetric HiLo microscopy employing an electrically tunable lens,” Opt. Express 24(13), 15029 (2016). [CrossRef]
30. K. N. Elfer, A. B. Sholl, M. Wang, D. B. Tulman, S. H. Mandava, B. R. Lee, and J. Q. Brown, “DRAQ5 and eosin (‘D&E’) as an analog to hematoxylin and eosin for rapid fluorescence histology of fresh tissues,” PLoS One 11(10), e0165530 (2016). [CrossRef]
31. I. Arganda-Carreras, V. Kaynig, C. Rueden, K. W. Eliceiri, J. Schindelin, A. Cardona, and H. S. Seung, “Trainable Weka segmentation: a machine learning tool for microscopy pixel classification,” Bioinformatics 33(15), 2424–2426 (2017). [CrossRef]
32. F. Aguet, D. Van De Ville, and M. Unser, “Model-based 2.5-D deconvolution for extended depth of field in brightfield microscopy,” IEEE Trans. on Image Process. 17(7), 1144–1153 (2008). [CrossRef]
33. S. Krishnamurthy, J. Q. Brown, N. Iftimia, R. M. Levenson, and M. Rajadhyaksha, “Ex vivo microscopy a promising next-generation digital microscopy tool for surgical pathology practice,” Arch. Pathol. Lab. Med. 143(9), 1058–1068 (2019). [CrossRef]
34. T. Dertinger, R. Colyera, G. Iyer, S. Weiss, and J. Enderlein, “Fast, background-free, 3D super-resolution optical fluctuation imaging (SOFI),” Proc. Natl. Acad. Sci. U. S. A. 106(52), 22287–22292 (2009). [CrossRef]
35. E. Mudry, K. Belkebir, J. Girard, J. Savatier, E. Le Moal, C. Nicoletti, M. Allain, and A. Sentenac, “Structured illumination microscopy using unknown speckle patterns,” Nat. Photonics 6(5), 312–315 (2012). [CrossRef]
36. L.-H. Yeh, S. Chowdhury, N. A. Repina, and L. Waller, “Speckle-structured illumination for 3D phase and fluorescence computational microscopy,” Biomed. Opt. Express 10(7), 3635 (2019). [CrossRef]
37. N. Hagen, L. Gao, and T. S. Tkaczyk, “Quantitative sectioning and noise analysis for structured illumination microscopy,” Opt. Express 20, 403–413 (2011). [CrossRef]
38. W. J. Shain, N. A. Vickers, B. B. Goldberg, T. Bifano, and J. Mertz, “Extended depth-of-field microscopy with a high-speed deformable mirror,” Opt. Lett. 42(5), 995 (2017). [CrossRef]
39. L. Jin, Y. Tang, Y. Wu, J. B. Coole, M. T. Tan, X. Zhao, H. Badaoui, J. T. Robinson, M. D. Williams, A. M. Gillenwater, R. R. Richards-Kortum, and A. Veeraraghavan, “Deep learning extended depth-of-field microscope for fast and slide-free histology,” Proc. Natl. Acad. Sci. U. S. A. 117(52), 33051–33060 (2020). [CrossRef]
40. S. Santos, K. K. Chu, D. Lim, N. Bozinovic, T. N. Ford, C. Hourtoule, A. C. Bartoo, S. K. Singh, and J. Mertz, “Optically sectioned fluorescence endomicroscopy with hybrid-illumination imaging through a flexible fiber bundle,” J. Biomed. Opt. 14(3), 030502 (2009). [CrossRef]
