Context.—

Digital pathology using whole slide images has been recently approved to support primary diagnosis in clinical surgical pathology practices. Here we describe a novel imaging method, fluorescence-imitating brightfield imaging, that can capture the surface of fresh tissue without requiring prior fixation, paraffin embedding, tissue sectioning, or staining.

Objective.—

To compare the ability of pathologists to evaluate direct-to-digital images with standard pathology preparations.

Design.—

One hundred surgical pathology samples were obtained. Samples were first digitally imaged, then processed for standard histologic examination on 4-μm hematoxylin-eosin–stained sections and digitally scanned. The resulting digital images from both digital and standard scan sets were viewed by each of 4 reading pathologists. The data set consisted of 100 reference diagnoses and 800 study pathologist reads. Each study read was compared to the reference diagnosis, and also compared to that reader’s diagnosis across both modalities.

Results.—

The overall agreement rate, across 800 reads, was 97.9%. This consisted of 400 digital reads at 97.0% versus reference and 400 standard reads versus reference at 98.8%. Minor discordances (defined as alternative diagnoses without clinical treatment or outcome implications) were 6.1% overall, 7.2% for digital, and 5.0% for standard.

Conclusions.—

Pathologists can provide accurate diagnoses from fluorescence-imitating brightfield imaging slide-free images. Concordance and discordance rates are similar to published rates for comparisons of whole slide imaging to standard light microscopy of glass slides for primary diagnosis. It may be possible, therefore, to develop a slide-free, nondestructive approach for primary pathology diagnosis.

While in vivo methods for disease detection and classification, including radiology and liquid biopsy, continue to improve with respect to spatial resolution (radiology) and molecular information, they have not, to date, obviated the need for tissue biopsy and/or excision, methods that remain the current standard for all cancer and many noncancer diagnoses in clinical medicine.1,2  However, tissue processing of these biopsies incurs a significant delay, typically overnight, and not infrequently some tissue is lost or discarded in this process. The resulting standard tissue sections mounted on glass slides can then be scanned using automated whole slide microscopy systems, and then reviewed on computer displays; this approach has recently been approved by the US Food and Drug Administration (FDA) to support primary surgical pathology diagnosis in clinical medicine.35  However, while such digital imaging comes with real advantages, including being a necessary prerequisite for computational pathology approaches, they do not solve some major logistical challenges posed by conventional pathology, and in fact contribute to them. This is because the time- and effort-consuming workflow involved in standard slide production is still required, and in fact, scanning, which occurs after the slides are produced, represents another costly and time-consuming step.

It would be desirable if histologic information could be acquired even before slides are generated, cutting hours and expense from the diagnostic process. The current approach for rapid histology involves intraoperative assessment using frozen sections; the process can be performed quickly, but requires skilled input, consumes potentially valuable tissue, and often generates substandard histology results due to artifacts associated with tissue freezing and other steps in the cryotomy process.

In response to a perceived need for rapid, slide-free histology, a number of tissue-imaging technologies are under development, but adoption to date has been primarily in research-only applications. These approaches include microscopy with ultraviolet excitation (MUSE),6  light-sheet microscopy,7,8  structured illumination microscopy,9  optical coherence microscopy,10  stimulated Raman histology,11  and most recently, photoacoustic-based systems,12  all useful and promising approaches, with current implementations presenting some challenges with respect to cost, resolution, and/or throughput.

Here, we describe a new method for rapid slide-free histology, fluorescence-imitating brightfield imaging (FIBI). FIBI is based on the use of absorbing (not necessarily fluorescing) stains applied briefly to the specimen, coupled with 405-nm illumination. The stains penetrate the surface of the tissue to approximately 100 μm (data not shown), but the resulting images demonstrate apparent thickness only slightly greater than the corresponding hematoxylin and eosin (H&E) slides. The 405-nm excitation light generates broad-spectrum autofluorescence diffusely inside the bulk of the thick (unsectioned) sample. This generated autofluorescence then back-illuminates the stained layer to generate an image of the surface layer of the tissue. Conveniently, stains familiar to histopathologists, namely, H&E, perform well in this application. Hematoxylin is a strongly absorbing, nonfluorescing dye that binds predominantly to nuclei. Eosin staining provides a red-pink tint to protein and other non–nucleic acid elements, just as it does in conventional slide-based histology, while contributing additional contrast due to its intrinsic fluorescence.13  An example of a FIBI image of nonneoplastic breast tissue is shown in Figure 1.

Figure 1

Fluorescence-imitating brightfield imaging image of breast lobules, ducts, stroma, and microvasculature (original magnification ×80).

Figure 1

Fluorescence-imitating brightfield imaging image of breast lobules, ducts, stroma, and microvasculature (original magnification ×80).

Close modal

This report presents a comparison study of pathologist diagnoses using FIBI technology for a variety of surgical pathology specimens compared to paired standard whole slide images (WSIs) of formalin-fixed, paraffin-embedded H&E 4-μm tissue sections.

FIBI Imaging Instrumentation and Procedure

Figure 2 presents a schematic drawing of the instrumentation. The tissue samples are flattened against a glass coverslip inside a custom-built specimen holder with an imaging window of 25 × 35 mm2. Illumination is performed using a 405-nm light-emitting diode (LED; LZ1-00UB00, LED Engin) directed to the sample via a broadband dichroic (Di03-R405-t1-25x36, Semrock). The scanning is performed with an average speed of 1 cm2/min with XY travel range of up to 25 × 50 mm2, along with a Z-positioner (Zaber Technologies, Vancouver, Canada) to provide focus stacking. A ×10, NA = 0.3 Plan Fluor objective (Nikon, Japan) was employed for this study, although both ×10 NA = 0.45 and ×20 objectives can also be used, and images were collected with a 9-megapixel scientific-grade CCD color camera (Ximea, MD091CU-SY) connected via a 160-mm tube lens (Thorlabs TTL 165-A).

Figure 2

Schematic diagram of a fluorescence-imitating brightfield imaging microscope.

Figure 2

Schematic diagram of a fluorescence-imitating brightfield imaging microscope.

Close modal

Images were acquired using a focus-stacking technique to create extended depth of field images required to accommodate slight tissue surface unevenness.14  The captured images were color-corrected using a custom histogram-matching tool15  and processed to improve contrast and sharpness using the AForge.NET open-source library. Finally, the images were stitched and pyramidal svs files created using a stitching library from Microvisioneer (Esslingen am Neckar, Germany). Image acquisition, alternation between light sources, stage movement, and focusing was performed using control software developed in the .NET environment.

In order to determine if the image quality obtained using FIBI was noninferior to conventional formalin-fixed, paraffin-embedded H&E WSIs we designed a study comprising 100 surgical pathology cases, and recruited 4 reading pathologists who had no prior experience viewing FIBI-acquired images. They were asked to provide a diagnosis for each case twice, seeing either the FIBI case first or the H&E case first, and then, after a 30-day washout period, they were presented with the same specimens imaged with the alternative technique. That is, for some cases they saw the FIBI (or H&E) images first, and then, 30 days later, the H&E (or FIBI) images of the same cases. All 100 specimens were read by all study pathologists in both modalities. We compared the concordances of diagnoses for these 800 reads to the reference diagnoses that had been determined by 2 experienced pathologists who were not part of the reading panel. The results were then interpreted in line with analyses used to evaluate performance in consensus recommendations used for WSI versus glass slide comparison studies performed as part of the FDA approval process used in the Leica/Aperio scanner submission.5 

Specimens

One hundred surgical pathology specimens from the University of California Davis Health clinical laboratory (Sacramento, California) in surgical pathology were chosen to span common specimen types from available excess tissues not needed for the initial diagnosis and designated for disposal. In all cases the tissues used were fixed in formalin, but not otherwise processed. Tissues used for the study were all obtained without protected health information (ie, deidentified) and were obtained under a University of California Davis institutional review board determination of “not-human subjects research.” Selection from this cohort biased the cases toward excisions, as much smaller biopsy specimens are typically processed in their entirety (and therefore would not have been available for this study). In addition, because the most morphologically relevant tissue regions may have been mostly or entirely processed as part of the clinical workup, residual discarded tissue might consist only of adjacent benign material. This yielded an array of tissue types and diagnoses including benign normal, hyperplasias, benign neoplasms, cancers, and various inflammatory lesions. See Table 1 for sites of origin and a complete list of diagnoses.

Table 1

Case Distribution by Primary Tissue Type

Case Distribution by Primary Tissue Type
Case Distribution by Primary Tissue Type

FIBI Imaging

In preparation for imaging, the tissue available for each case was examined in conjunction with the deidentified pathology report, and a selected area corresponding to the size of a typical medium-to-large specimen (ie, up to about 2.0 × 2.5 cm2) was manually dissected using a razor blade to achieve a tissue surface as flat as possible.

FIBI imaging was performed by first staining the tissue with 0.5 mg/ml Mayer hematoxylin (H3136, MilliporeSigma) for 30 seconds. The tissue was then rinsed in deionized water before staining with 0.25% eosin Y (E4382, MilliporeSigma) in 70% ethanol for 30 seconds. Following staining, the tissue was rinsed twice with deionized water for 30 seconds. The tissue was placed in a holder designed to gently compress the specimen against an imaging coverslip, and the whole tissue specimen was scanned with the FIBI instrument described above, with extended depth of field enabled.

H&E Imaging and Case Curation

After FIBI imaging, the tissue sample was submitted for standard histology involving paraffin embedding and microtome sectioning at 4 μm, mounting on glass slides, H&E staining, and coverslipping; the resulting slides were then scanned on an Aperio/Leica AT2 slide scanner at ×20. The resulting FIBI and H&E WSIs were uploaded to PathPresenter (www.pathpresenter.net) for viewing. The supervising study pathologist ensured that each image captured the study reference diagnosis. In some cases, this reflected the actual clinical diagnosis (obtained from the deidentified pathology report), but if the study pathologist determined that the sections included in the study series did not contain tissue regions that reflected that diagnosis, an alternative reference diagnosis was recorded. For example, the study tissue may have been comprised of benign tissue even if the primary case diagnosis recorded the presence of cancer. Minimal clinical data were provided to the reviewing pathologists, consisting of patient age and sex and tissue sampling origin. If, in the opinion of the lead study pathologist, additional clinical metadata were needed to arrive at the reference diagnosis (as would be available in a true clinical context), they were added to the available metadata.

Case Reading Methodology

Four “reader” pathologists were recruited to the study. All are board certified in anatomic pathology, with varying experience in general and specific surgical pathology (gastrointestinal and pediatric), with 2 in academic and 2 in private surgical pathology practice.

Case order and modality (FIBI versus H&E) was randomized for each pathologist to view only those images on PathPresenter, a cross-platform digital image viewer. (Note that in the discussion that follows, for convenience and familiarity, we refer to the conventionally sectioned slide-based images as “H&E” even though the thick [FIBI] and thin-cut specimens were both stained with H&E.) The study pathologists were asked to provide the best or most likely diagnosis they could arrive at without additional information or ancillary studies such as immunohistochemistry and were also asked to comment about adequacy. They were advised that any resource, textbook, medical literature, or online database that they would use clinically to help arrive at the best diagnosis was permitted. They were not permitted, however, to consult with a colleague even though this is common in clinical practice. If they would obtain immunohistochemistry in clinical practice, or if they would send a case for expert consultation, they could add a note or comment, but these steps were not permitted in the study.

After a 30-day washout period, the same 100 cases were again provided—but with the modality switched for each case. If the observer saw the H&E for a case in the first set, in the second round they had access only to the FIBI image, and vice versa. They were not able to view their prior diagnoses or the previous set of slides.

Data Collection and Adjudication

The diagnoses were compared to the reference and scored as concordant, minorly discordant, or majorly discordant. “Minor discordance” was defined as a difference in the diagnosis that would not lead to a different clinical treatment or outcome, whereas “major discordance” implies that the difference in the diagnosis would lead to a different treatment recommendation or outcome.5  All diagnoses and concordance scores were reviewed independently by 2 pathologists, and any differences in concordance assignment were discussed to reach a consensus.

For the purposes of the primary endpoint, major discordance rates were compared to the combined rates of concordance and minor discordance. In addition, intraobserver variation was determined by comparing a reading pathologist’s diagnosis from one modality to the other. In this way, although a reader may have been in disagreement with the reference diagnosis, discrepancies due solely to the impact of the modality used could be assessed.

To determine if the FIBI method was noninferior to H&E, a binomial regression with an identity link was fit to model the percentage of major discordance versus method. Because observations by the same pathologist and on the same specimen were correlated, we estimated model parameters with generalized estimating equations and used a robust covariance matrix to estimate the variance. The FIBI method was deemed to be noninferior to H&E if the upper bound of a 1-sided 95% confidence limit for the difference between the FIBI and H&E method was less than the prespecified noninferiority margin of 4.0%. PROC GENMOD in SAS Version 9.4 was used to fit this model.

Images were acquired directly from the surface of intact fixed specimens, which were then submitted for standard histology, attention being paid to having the same face of the tissue available for sectioning and staining that was previously captured during the original FIBI scan. An example of a whole specimen image, similar to what was presented to the reviewers via PathPresenter, is shown in Figure 3, A, with the standard H&E on fixed, processed, and stained tissue in Figure 3, B. Reviewers could control orientation (rotation) and zooming, as well as simple image display parameters (brightness and contrast).

Figures 3

Comparison of fluorescence-imitating brightfield imaging image of the cervix (A) with whole slide image scanning (B), the latter following standard hematoxylin-eosin staining after fixation and processing (original magnification ×4). Screen shots from PathPresenter are shown here.

Figures 3

Comparison of fluorescence-imitating brightfield imaging image of the cervix (A) with whole slide image scanning (B), the latter following standard hematoxylin-eosin staining after fixation and processing (original magnification ×4). Screen shots from PathPresenter are shown here.

Close modal

The distribution of tissues of origin and corresponding diagnoses of all the cases in the study set is shown in Table 1. Overall, 54% (54 of 100) were malignant, 26% showed non–malignancy-related pathology (eg, proliferative, inflammatory), and 20% showed no evidence of disease.

Concordance

Determination of concordance was performed by comparing reader diagnoses to reference diagnoses, and also by comparing the reader diagnosis from FIBI to the same-reader diagnosis from H&E. Adjudication followed criteria established with FDA guidance and used in similar studies.5  The rates of concordance, minor discordance, and major discordance were tabulated for each reader pathologist, and for each image modality.

For the 800 reader diagnoses, the overall rates for both modalities combined were 91.8% concordance, 6.1% minor discordance, and 2.1% major discordance. The rates for H&E-only were 93.8% concordance, 5.0% minor discordance, and 1.2% major discordance. The rates for FIBI-only were 89.8% concordance, 7.2% minor discordance, and 3.0% major discordance. The difference in the major discordance rates between FIBI and H&E was 1.8% with a 1-sided upper 95% CI of 3.1%, meeting the predetermined criterion for acceptance of less than 4.0%.

Intraobserver discrepancies were assessed by first comparing each reader’s H&E-to-FIBI diagnosis. Considering the relative contribution of discordances in each modality, versus identical discordance in both modalities for a given case from a given observer, we audited all major discordances with respect to the reference diagnosis to determine if the discordances were modality-specific. For all readers combined, the H&E-to-FIBI rate of intraobserver no major discordance was 98.2% (1.8% or 14 of 17 major discordances) with 4.0% minor differences. Of the H&E majorly discordant diagnoses (n = 5), 3 out of 5 had the same discordant diagnosis from that reader using FIBI (60% internal concordance in the H&E-to-FIBI direction). Of the FIBI major-discordant diagnoses (n = 12) 3 of 12 had the same discordant diagnosis from that reader using H&E (25% internal concordance in the FIBI to H&E direction). A summary of the discordance rates is found in Table 2, and further delineation of the cases showing major discordances is found in Table 3.

Table 2

Summary of Results

Summary of Results
Summary of Results
Table 3

Summary of Discordant Cases

Summary of Discordant Cases
Summary of Discordant Cases

Example Image Pairs

A pair of FIBI and corresponding H&E images from a benign breast specimen is shown in Figure 4, A through F. Even at low digital magnification, the lobules are clearly visible in both presentations. In general, looking at higher-power magnification, there is a close resemblance between the modes, but there are discernable differences in image properties. FIBI images can display contrast not in evidence in standard H&Es. For example, lipid is still present (and imageable) in the intact tissues imaged with FIBI, but as lipids are removed during the solvent-based paraffin-embedding process, lipid within adipose regions becomes empty “white” areas on H&E-stained slides, as can be seen in the lower areas of whole specimen images (top row), which also highlights presence of adipose blood capillaries not readily visible in the H&E version.

Figure 4

Normal breast. Fluorescence-imitating brightfield imaging (FIBI) image (A). Higher power of boxed areas in A (B and C). Standard hematoxylin-eosin–stained section (D). Higher power of boxed areas in D (E and F). Note lipid and capillaries in adipose tissue visible in the FIBI specimens (original magnifications ×4 [A and D], ×40 [B and E], and ×20 [C and F]).

Figure 4

Normal breast. Fluorescence-imitating brightfield imaging (FIBI) image (A). Higher power of boxed areas in A (B and C). Standard hematoxylin-eosin–stained section (D). Higher power of boxed areas in D (E and F). Note lipid and capillaries in adipose tissue visible in the FIBI specimens (original magnifications ×4 [A and D], ×40 [B and E], and ×20 [C and F]).

Close modal

Figure 5, A through D, illustrates a case of invasive lobular carcinoma imaged via FIBI (Figure 5, A and C) and H&E (Figure 5, B and D). Zoomed-in regions in each mode show a duct surrounded by invasive lobular carcinoma cells. An example of a tongue invasive squamous cell carcinoma with necrosis is present in Figure 6, A through D. Resolution in the FIBI image is sufficient to demonstrate chromatin appearance and the presence of mitotic figures, as shown in the both the FIBI (Figure 6, C) and standard H&E (Figure 6, D) preparations.

Figure 5

Fluorescence-imitating brightfield imaging (A and C) and corresponding standard hematoxylin-eosin–stained images (B and D) of invasive lobular carcinoma of the breast. Whole specimen (A and B) and zoomed-in regions (C and D) showing nonneoplastic ducts surrounded by invading lobular carcinoma cells (original magnifications ×4 [A and B] and ×160 [C and D]).

Figure 5

Fluorescence-imitating brightfield imaging (A and C) and corresponding standard hematoxylin-eosin–stained images (B and D) of invasive lobular carcinoma of the breast. Whole specimen (A and B) and zoomed-in regions (C and D) showing nonneoplastic ducts surrounded by invading lobular carcinoma cells (original magnifications ×4 [A and B] and ×160 [C and D]).

Close modal
Figure 6

Fluorescence-imitating brightfield imaging (FIBI) (B and C) and corresponding standard hematoxylin-eosin–stained (H&E) images (A and D) of invasive squamous cell carcinoma of the tongue with necrosis. Whole specimen (A and B) and zoomed-in regions (C and D). Note mitotic figures (circles) visible in both FIBI (C) and standard H&E (D) images (original magnifications ×6 [A and B] and ×300 [C and D]).

Figure 6

Fluorescence-imitating brightfield imaging (FIBI) (B and C) and corresponding standard hematoxylin-eosin–stained (H&E) images (A and D) of invasive squamous cell carcinoma of the tongue with necrosis. Whole specimen (A and B) and zoomed-in regions (C and D). Note mitotic figures (circles) visible in both FIBI (C) and standard H&E (D) images (original magnifications ×6 [A and B] and ×300 [C and D]).

Close modal

For more than a century, diagnostic pathology has relied upon bright-field transillumination microscopy of thinly sliced and stained tissues. The preparation of these specimens for microscopic analysis entails a sequence of steps (fixation, dehydration, infusion with paraffin, oriented block mounting, microtome sectioning at 4 μm, transfer to a glass slide, rehydrating and staining with histology dyes, and coverslip mounting). In aggregate, this process is time- and resource-demanding.

Digital pathology, the analysis and evaluation of tissue sections that have been converted to a digital image using commercial slide scanners and then examined and evaluated on a computer or monitor screen, has several theoretical advantages as compared to standard analog pathology based on direct viewing of glass slides. Some of these advantages include the ability to share images instantaneously with other pathologists in different locations, and, when coupled with image analysis and machine learning, the potential for greater accuracy, reproducibility, and standardization of pathology diagnoses and prediction of patient outcome.16 

Two large seminal studies conclusively demonstrated the “noninferiority” of digital-image–based diagnoses compared with microscopic slide evaluation.5,17  In the study by Mukhopadhaya et al,17  1992 cases were studied, corresponding to 15 925 reads, the major discordance rate with the reference standard diagnosis was 4.9% for digital images and 4.6% for microscopy. The overall major discrepancy rate with respect to reference standard diagnoses,5  in which 2045 cases and 5849 slides were examined, was 3.6% for digital images, and 3.2% for manual slide review.

In this report, we have described FIBI, a novel slide-free digital histology system that permits rapid acquisition of digital images, taking minutes instead of hours or days as required by traditional slide-based pathology followed by whole slide imaging. Furthermore, we have demonstrated the ability of pathologists to render accurate diagnoses based on their evaluation of these FIBI-acquired digital images, with an accuracy within the range of existing standard histology and WSI systems. Results of the present study examining the suitability of FIBI-enabled histology demonstrated an accuracy of 97.0%, well within the ranges reported for conventional WSI approaches.

Staining is reproducible and relatively resilient to variations in stain concentration and staining time. The dyes, H&E, stain the tissues with consistent partitioning between different tissue components, hematoxylin preferentially staining nuclei, eosin the bulk of the other constituents. What varies, depending on staining parameters, is the amount of the dyes actually binding to the specimen, so that, just as with standard histology slides, the nuclei, for example, can be darkly or lightly stained. Since the FIBI images are intrinsically digital, simple color histogram adjustments can be employed to maintain a consistent appearance between specimens and staining sessions.

We have seen very little effect on fixation time and FIBI image properties. Spatial and staining features appear to be essentially identical. The only practical difference is that the brightness of the autofluorescent backlighting does vary with fixation status. Fixation can considerably enhance the intensity of the diffuse, thick-tissue–generated backlight, resulting in typical exposure times that can be longer for fresh (lower backlight intensity) than fixed. Even so, exposure times per frame in fresh specimens can be around 25 milliseconds (versus 5 milliseconds with fixed specimens) as captured with recent, brighter-LED–equipped prototypes.

The development of FIBI comes at a time when advances in optics, light sensors, microcomputer power, and algorithmic methodologies have dramatically increased the available strategies for optical microscopy in recent years. But while technology may permit development of slide-free optical microscopy, any compelling slide-free microscopy system that has potential for introduction into diagnostic surgical pathology laboratories ideally should be based on simple and affordable hardware, produce images within minutes not hours, not be destructive of tissue, generate a wide field of view, and be able to produce images that closely simulate those of scanned-in images of conventional slide-based H&E stained tissue, permitting accurate diagnostic assessment by the pathologist.

A recent review highlights a number of slide-free histopathology imaging systems,18  some of which can produce images that approximate the appearance of slide-based H&E-stained tissue. These include confocal microscopy,19  nonlinear microscopy,20  structured illumination microscopy phase detection,21  microscopy with ultraviolet surface excitation (MUSE),6,22  full-field optical coherence tomography,23  light-sheet microscopy,7,8  and photoacoustic microscopy.12  Among this group, some approaches have an ability to generate contrast without requiring exogenous stains (for example, multiphoton24  and photoacoustic techniques12) or to create 3-dimensional data that provide information beyond what is visible on 2-dimensional thin sections (for example, light sheet systems7). However, these methods can involve relatively complex optics, large data set generation and computationally dependent image display.

Among the slide-free approaches, MUSE and FIBI are differentiated because of their use of simple, affordable hardware that produce images that pathologists can interpret much as they would standard H&E-stained tissue sections. FIBI differs from MUSE mostly in that the contrast used to generate the images is based on staining with the familiar dyes, H&E; induced broad-band “virtual” backlighting then generates images that directly resemble standard histopathology—no image intensity inversion or major recoloring steps are required. Moreover, FIBI offers an imaging rate that is an order of magnitude faster, due to achievable high brightness, and because of epi-illumination (through-the-lens excitation), it is straightforward to include an objective turret with a variety of lenses with different magnifications. A current limitation of FIBI is that the acquired images resemble standard microscopy of physical sections. So far, the optics produce images that would be directly comparable to a slightly thicker formalin-fixed, paraffin-embedded section, perhaps 6–8 mm. Comparing to a 4-μm standard section, about 30% more nuclei can be seen in dense cellular areas (data not shown). Optical and computational methods to further “pseudo-thin” the FIBI images are feasible in future implementations.

The study was designed as a 100-case pilot, without previous knowledge of expected discordance rates, and without an intentional design to prove noninferiority via FDA standards. With that said, the number of specimens and the number of reader pathologists would need to be expanded to meet FDA guidance for approval for use in primary clinical diagnosis.

In summary, in this pilot validation study, we have demonstrated the ability of FIBI to produce direct-to-digital images within minutes that permit pathologist assessment with an accuracy well within College of American Pathologists and US FDA guidance for approvals of digital pathology WSIs for primary diagnosis.4,5  In contrast to traditional WSI digital pathology, where an extra step—scanning the slide once it is complete—is needed, FIBI produces an intrinsically digital image directly from the tissue with brief staining but without requiring additional processing. It may be possible, therefore, to develop a slide-free, nondestructive diagnostic work flow for primary pathology diagnosis. Such a method promises improved speed, reduced cost, and better conservation of tissue for advanced ancillary studies.

1.
Cardiff
RD,
Gregg
JP,
Miller
JW,
Axelrod
DE,
Borowsky
AD.
Histopathology as a predictive biomarker: strengths and limitations
.
J Nutr
.
2006
;
136
(10)
:
2673S
5S
.
2.
De Las Casas
LE,
Hicks
DG.
Pathologists at the leading edge of optimizing the tumor tissue journey for diagnostic accuracy and molecular testing
.
Am J Clin Pathol
.
2021
;
155
(6)
:
781
792
.
3.
US Food and Drug Administration
.
FDA news release: FDA allows marketing of first whole slide imaging system for digital pathology; 2017
. .
4.
Mukhopadhyay
S,
Feldman
MD,
Abels
E,
et al
Whole slide imaging versus microscopy for primary diagnosis in surgical pathology: a multicenter blinded randomized noninferiority study of 1992 cases (pivotal study)
.
Am J Surg Pathol
.
2018
;
42
(1)
:
39
52
.
5.
Borowsky
AD,
Glassy
EF,
Wallace
WD,
et al
Digital whole slide imaging compared with light microscopy for primary diagnosis in surgical pathology
.
Arch Pathol Lab Med
.
2020
;
144
(10)
:
1245
1253
.
6.
Fereidouni
F,
Harmany
ZT,
Tian
M,
et al
Microscopy with ultraviolet surface excitation for rapid slide-free histology
.
Nat Biomed Eng
.
2017
;
1
(12)
:
957
966
.
7.
Liu
Y,
Rollins
AM,
Jenkins
MW.
CompassLSM: axially swept light-sheet microscopy made simple
.
Biomed Opt Express
.
2021
;
12
(10)
:
6571
6589
.
8.
Glaser
AK,
Reder
NP,
Chen
Y,
et al
Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens
.
Nat Biomed Eng
.
2017
;
1
(7)
:
0084
.
9.
Brown
JQ,
Sholl
AB.
Applications of structured light microscopy in clinical pathology, microscopy, histopathology, and analytics
. In:
Biophotonics Congress: Biomedical Optics Congress 2018 (Microscopy/Translational/Brain/OTS)
.
Hollywood, FL
:
OSA Technical Digest (Optica Publishing Group)
;
2018
:
MF1A 1
.
10.
Gabriele
ML,
Wollstein
G,
Ishikawa
H,
et al
Optical coherence tomography: history, current status, and laboratory work
.
Invest Ophthalmol Vis Sci
.
2011
;
52
(5)
:
2425
2436
.
11.
Ji
M,
Lewis
S,
Camelo-Piragua
S,
et al
Detection of human brain tumor infiltration with quantitative stimulated Raman scattering microscopy
.
Sci Transl Med
.
2015
;
7
(309)
:
309ra163
.
12.
Shi
J,
Wong
TTW,
He
Y,
et al
High-resolution, high-contrast mid-infrared imaging of fresh biological samples with ultraviolet-localized photoacoustic microscopy
.
Nat Photonics
.
2019
;
13
:
609
615
.
13.
Lev
R,
Stoward
PJ.
On the use of eosin as a fluorescent dye to demonstrate mucous cells and other structures in tissue sections
.
Histochemie
.
1969
;
20
(4)
:
363
377
.
14.
Goldsmith
NT.
Deep focus; a digital image processing technique to produce improved focal depth in light microscopy
.
Image Analysis Stereol
.
2000
;
19
:
163
167
.
15.
Jia
W,
Zhang
X,
He
Q,
Wu
A.
A comparison on histogram based image matching methods
. In:
2006 IEEE International Conference on Video and Signal Based Surveillance
.
Sydney, NSW, Australia
,
2006
;
97
97
. .
16.
Bauer
TW,
Slaw
RJ.
Validating whole-slide imaging for consultation diagnoses in surgical pathology
.
Arch Pathol Lab Med
.
2014
;
138
(11)
:
1459
1465
.
17.
Mukhopadhaya
A,
Mendecki
J,
Dong
X,
et al
Localized hyperthermia combined with intratumoral dendritic cells induces systemic antitumor immunity
.
Cancer Res
.
2007
;
67
(16)
:
7798
7806
.
18.
Liu
Y,
Levenson
RM,
Jenkins
MW.
Slide over: advances in slide-free optical microscopy as drivers of diagnostic pathology
.
Am J Pathol
.
2022
;
192
(2)
:
180
194
.
19.
Ragazzi
M,
Piana
S,
Longo
C,
et al
Fluorescence confocal microscopy for pathologists
.
Mod Pathol
.
2014
;
27
(3)
:
460
471
.
20.
Tao
YK,
Shen
D,
Sheikine
Y,
et al
Assessment of breast pathologies using nonlinear microscopy
.
Proc Natl Acad Sci U S A
.
2014
;
111
(43)
:
15304
15309
.
21.
Wang
M,
Kimbrell
HZ,
Sholl
AB,
et al
High-resolution rapid diagnostic imaging of whole prostate biopsies using video-rate fluorescence structured illumination microscopy
.
Cancer Res
.
2015
;
75
(19)
:
4032
4041
.
22.
Pekmezci
M,
Morshed
RA,
Chunduru
P,
et al
Detection of glioma infiltration at the tumor margin using quantitative stimulated Raman scattering histology
.
Sci Rep
.
2021
;
11
(1)
:
12162
.
23.
Apelian
C,
Harms
F,
Thouvenin
O,
Boccara
AC.
Dynamic full field optical coherence tomography: subcellular metabolic contrast revealed in tissues by interferometric signals temporal analysis
.
Biomed Opt Express
.
2016
;
7
(4)
:
1511
1524
.
24.
You
S,
Tu
H,
Chaney
EJ,
et al
Intravital imaging by simultaneous label-free autofluorescence-multiharmonic microscopy
.
Nat Commun
.
2018
;
9
(1)
:
2125
.

Author notes

This work was supported by NIH grant 1R01EB028635 (Levenson and Fereidouni) and in part by Histolix.

Competing Interests

Borowsky, Levenson, and Fereidouni are cofounders of Histolix Inc; Gown is a consultant to Histolix Inc; and Morningstar was a paid full-time employee of Histolix Inc for part of the time of the study. Fleury has a financial interest in a capital group that has invested in Histolix. The other authors have no relevant financial interest in the products or companies described in this article.