Utility of whole‐slide imaging for rapid evaluation of thyroid FNA: A multireader prospective study

Abstract

Background

Rapid on‐site evaluation (ROSE) of thyroid fine‐needle aspiration biopsy (FNAB) improves diagnostic adequacy and facilitates ancillary molecular testing. In this prospective, multireader study, the authors evaluated the feasibility of using whole‐slide images (WSIs) for ROSE to determine specimen adequacy and preliminary categorization (according to The Bethesda System for Reporting Thyroid Cytopathology [Bethesda]) of image‐guided thyroid FNABs compared with conventional light‐microscopic (LM) examination of the same specimens in a referral cancer center.

Methods

The authors evaluated 98 ultrasound‐guided thyroid FNAB cases. Smears were stained with Papanicolaou and Diff‐Quik and were scanned at ×20 magnification using a Leica Aperio CS2 scanner. Five cytopathologists evaluated specimen adequacy and Bethesda categorization using WSI followed by LM assessment after a 2‐week washout. Intraobserver and interobserver agreements were calculated using Cohen and Fleiss kappa (κ) statistics. Scan time, interpretation time, and the need for ×40 magnification or z stacking were recorded.

Results

In total, 463 slides were scanned, with mean scan time of 5.48 minutes. WSI quality was acceptable in most cases. Z stacking and ×40 magnification were requested in 23% and 14% of reviews, respectively. Intrareader agreement between WSI and LM examination was excellent (κ = 0.86–0.95). Inter‐reader agreement was moderate for both WSI (κ = 0.48) and LM examination (κ = 0.56). Concordance was highest for Bethesda categories I and VI and lowest for categories III–V. Interpretation with WSI took significantly longer than with LM examination (p < .0001).

Conclusions

WSI is a feasible alternative to LM examination for ROSE of thyroid FNABs, with high intrareader agreement and comparable inter‐reader agreement. The limited need for high magnification and z stacking supports its practical utility.

Short abstract

Whole‐slide imaging is feasible for the rapid evaluation of specimen adequacy and preliminary Bethesda categorization of thyroid fine‐needle aspiration biopsies, with diagnostic agreement similar to that of light microscopic examination and a minimal need for ×40 high magnification or z stacking.

INTRODUCTION

Fine‐needle aspiration biopsy (FNAB), with or without ultrasound (US) image guidance, is well established as a safe, simple, and cost‐effective technique that is commonly used for the initial evaluation of thyroid nodules. ¹ , ² Rapid on‐site evaluation (ROSE) of thyroid FNABs can enable assessment of specimen adequacy, thereby improving the procurement of diagnostic material. ³ , ⁴ , ⁵ In addition, preliminary cytologic categorization of thyroid FNABs in real time also ensures the collection of additional material for potential ancillary molecular testing if the aspirate is categorized according to The Bethesda System for Reporting Thyroid Cytopathology (Bethesda) as belonging to the indeterminate category, including Bethesda categories III and IV. ⁶ , ⁷

Thyroid FNABs, with or without image guidance, are very often performed in endocrinology clinics and radiology suites in different locations within the same institution or in satellite centers away from the main hospital, making it challenging to support ROSE of image‐guided thyroid FNABs in real time using conventional light‐microscopic (LM) examination. ³ Telecytology, including the electronic transmission of digital images of glass slides with aspirated cytology material, provides a potentially viable strategy to overcome issues related to performing ROSE of FNABs using LM examination in both academic and community cytology practices. ⁸

There are several digital options that can be considered for using telecytology for ROSE of thyroid FNABs, including static image transmission using a camera attached to a microscope, which is a very simple option for transmitting static images of selected areas of microscopic slides using e‐mail, smartphone applications, or cloud‐based secure sites. ⁹ , ¹⁰ , ¹¹ Alternatively, the on‐site FNA operator can share their computer screen using conferencing software, such as Microsoft Teams (Microsoft Corporation), Zoom (Zoom Communications Inc.), or Google Meet (Google LLC), allowing the remote viewer to view the smear as it appears under LM examination.

Dynamic telecytology options include the use of a high‐definition video camera to stream live images directly from a light microscope to a remote viewer that enables continuous transmission of the microscopic field of view in real time. ¹² , ¹³ , ¹⁴ , ¹⁵ The video camera option may or may not be associated with a robotic microscope. The glass slide on the microscope can be controlled remotely on video while it is driven by an on‐site operator in remote viewing. ¹² , ¹³ , ¹⁴ , ¹⁵

Hybrid telecytology platforms, including digital microscopes and whole‐slide scanners, are currently available for live image streaming or acquiring whole‐slide images (WSIs). Whole‐slide scanners for real‐time scanning of limited glass slides may be suitable for telecytology. ¹⁶

A very limited number of studies have used telecytology for the evaluation of thyroid FNABs, as summarized in a recent systematic review. ¹⁷ Thirteen studies were identified that used digital options ranging from static images in four studies, dynamic real‐time image streaming in five, WSIs in three, and both WSIs and real‐time microscopy in one. ¹⁷ Essentially, only four studies reported the use of WSIs for the evaluation of thyroid FNABs. ¹⁸ , ¹⁹ , ²⁰ , ²¹ The main objective of these studies was using WSIs of thyroid FNABs for primary diagnosis or remote international consultation. To our knowledge, no reported studies have prospectively investigated the feasibility of using WSIs for ROSE of thyroid FNABs.

We evaluated the feasibility of using WSIs for ROSE to determine specimen adequacy and preliminary Bethesda categorization of US‐guided thyroid FNABs compared with conventional LM examination of the same specimens in a prospective, multi‐reader study at a referral cancer center.

MATERIALS AND METHODS

After we obtained approval from the Institutional Review Board of The University of Texas MD Anderson Cancer Center with waiver of informed consent, in total, 98 US‐guided thyroid FNAB cases were prospectively collected after the completion of standard‐of‐care ROSE using LM examination and final reporting. The cases were selected from all consecutive patients who underwent US‐guided FNAB at the University of Texas MD. Anderson Cancer Center and received by the Cytopathology laboratory between September 8, 2023, and November 9, 2023. The US‐guided thyroid FNABs were performed by radiologists who specialized in US imaging of the head and neck. Two passes were performed as the standard protocol, and direct smears were prepared and fixed in Carnoy fixative for Papanicolaou (Pap) staining and air‐dried for Diff‐Quik staining. The FNAB rinse was collected in Roswell Park Memorial Institute medium and prepared as either a cytospin or a cell block based on the size of the pellet resulting from centrifugation of the rinse. Immediate assessment of specimen adequacy and preliminary Bethesda categorization were performed by a cytopathologist using light microscopy that included a review of both Pap‐stained and Diff‐Quik–stained smears.

Patients’ demographic information, including age and sex, the site and size of thyroid nodules, and relevant clinical and oncologic histories were recorded. We also collected the American College of Radiology Thyroid Imaging Reporting and Data System (TI‐RADS) scores of US‐guided thyroid FNABs from patients when available. The glass slides were scanned in a tissue scanner (CT2 scanner; Leica Biosystems) at ×20 magnification. The scanner virtual slide files of the WSIs were stored at a dedicated site on a network drive that was accessible only to authorized users. Aperio Image scope software (Leica Biosystems) was used for viewing the images on desktop computers (Dell Inc.) using 24‐inch monitors. The time taken for scanning each glass slide was recorded.

We recruited five board‐certified cytopathologists at our institution to evaluate the WSIs of thyroid FNABs. The readers were tasked to evaluate the WSIs of Pap‐stained and Diff‐Quik–stained smears of thyroid FNABs for specimen adequacy and Bethesda categorization for thyroid cytology. In addition, the cytopathologists were required to record the time taken to evaluate each case, the image quality, the need for ×40 magnification or z stacking for optimal interpretation, and any additional comments related to the WSIs. After a washout period of at least 2 weeks, the cytopathologists were required to evaluate the thyroid FNABs using LM examination and describe the specimen adequacy, Bethesda categorization, and time taken for reviewing each case.

Statistical analysis

We evaluated intrareader agreement in Bethesda categorization of US‐guided thyroid FNABs between WSIs and LM examination separately within each reader using the Cohen kappa (κ) statistic. We used the weighted version of the κ statistic, which considers the order of responses and gives higher weight to responses (agreement or disagreement) that are closer in value and lower in weight compared with responses that are further apart.

Inter‐reader agreement of Bethesda categorization between WSI and LM examination was determined using the Fleiss extension of the Cohen κ statistic. Inter‐reader agreement was determined for the overall evaluation of all cases as well as for the individual Bethesda categories. The time taken for interpreting WSIs was compared with that for LM examination using the Wilcoxon signed‐rank test.

All statistical analyses were performed using R version 4.3.1 (The R Project for Statistical Computing). All statistical tests used a significance level of 5%. No adjustments for multiple testing were made. For the κ statistics, the interpretation of values followed the Landis and Koch classification: from no agreement to slight agreement (0.00–0.20), fair agreement (0.21–0.40), moderate agreement (0.41–0.60), substantial agreement (0.61–0.80), and excellent agreement (≥0.81).

RESULTS

The mean age of the patients in the study was 61 years (range, 23–94 years), and the female‐to‐male ratio was approximately 2:1. The thyroid nodules evaluated by US‐guided FNAB ranged in size from 0.4 to 9.4 cm in greatest dimension, with a mean size of 2.4 cm. The TI‐RADS scores were distributed as follows: TI‐RADS category 2 (TR2) in one case, TR3 in 21 cases, TR4 in 18 cases, and TR5 in 11 cases. In addition, one case was reported with a range of TR2–T3, and two had a range of TR3–TR4. The TI‐RADS score was not reported for the remaining 44 cases.

In total, 463 slides were scanned, with an average of 4.72 slides per case (range, from two to 10 slides per case). The average scanning time per slide was 5.48 minutes (range, from 45 seconds to 15 minutes). Five readers rated the WSI quality as acceptable in most cases for evaluating specimen adequacy and Bethesda classification. A weighted average of 379 images (82%) were rated as acceptable to good, and 83 (18%) were rated as poor. Figure 1 illustrates representative WSIs of thyroid FNABs acquired at ×20 magnification. Figure 2 shows a representative case of papillary thyroid carcinoma, including WSIs of Pap‐stained and Diff‐Quik–stained smears acquired at ×20 magnification demonstrating thick fragments with associated thinner areas that allowed interpretation. The need for z‐stacked images to evaluate three‐dimensional and thick fragments was indicated for 74 of 338 slides (23%), whereas the need for ×40 magnification scanned WSIs was mentioned for 47 of 336 slides (14%).

FIGURE 1.

Representative whole‐slide images of thyroid fine‐needle aspiration biopsies acquired at ×20, including: (A) A Diff‐Quik–stained smear showing thin colloid, (B) a Papanicolaou‐stained smear showing inspissated colloid and a loose group of benign thyroid follicular cells, (C) a Diff‐Quik–stained smear of recurrent papillary thyroid carcinoma in the thyroid bed, and (D) a Papanicolaou‐stained smear from a case of poorly differentiated thyroid tumor. The specimen adequacy and Bethesda categorization of these four cases were the same for all five readers in the study.

FIGURE 2.

Representative whole‐slide images of (A) Diff‐Quik–stained and (B) Papanicolaou–stained smears from a case of papillary thyroid carcinoma in the study that was acquired at ×20 magnification. Note that, although there were thick fragments that compromised optimal interpretation, the thinner portions of the smear (arrows) allowed accurate evaluation of the aspirate by all the readers.

Intrareader agreement of Bethesda categorization of US‐guided thyroid FNABs between WSIs and LM examination

The Bethesda categorization for WSIs and LM examinations were category I (nondiagnostic) in 13–21 and 13–18 cases, respectively; category II (follicular nodular disease) in 30–49 and 32–48 cases, respectively; category III (atypia of undetermined significance [AUS]) in 5–21 and 11–17 cases, respectively; category IV (follicular neoplasm/suspicious for follicular neoplasm) in one of six and one of four cases, respectively; category V (suspicious for malignancy) in zero of seven and zero of five cases, respectively; and category VI (papillary thyroid carcinoma) in 15–24 and 17–27 cases, respectively. Overall, the Bethesda categorization of each reader was nearly similar between WSI and LM examination.

The intra‐reader agreements, determined by the Cohen weighted κ for readers 1–5, were 0.95, 0.92, 0.93, 0.86, and 0.87, respectively, indicating excellent agreement. Table 1 illustrates the Bethesda classification of the 98 US‐guided thyroid FNABs, as assessed by the five readers using both WSI and LM examination methods, and the intrareader agreement, determined by the Cohen weighted κ statistic.

TABLE 1.

Bethesda categorization and intrareader agreement by Cohen κ statistic using whole‐slide images and light‐microscopic examination of ultrasound‐guided thyroid fine‐needle aspiration biopsies.

Bethesda category	Reader 1, n = 92		Reader 2, n = 98		Reader 3, n = 98		Reader 4, n = 98		Reader 5, n = 98
	WSI	LM	WSI	LM	WSI	LM (n = 94)	WSI	LM	WSI	LM
I	13	13	20	18	16	13	21	18	21	16
II	30	31	49	45	27	34	47	48	42	41
III	14	12	10	12	21	17	5	5	11	11
IV	4	3	1	1	5	4	2	1	0	1
V	7	4	3	5	1	0	5	5	4	8
VI	24	29	15	17	28	26	18	21	20	21
Intrareader agreement: Cohen κ statistic	0.95		0.92		0.93		0.88		0.87

Abbreviations: Bethesda, The Bethesda System for Reporting Thyroid Cytopathology; LM, light‐microscopic examination; WSI, whole‐slide image.

Inter‐reader agreement of Bethesda categorization of image‐guided thyroid FNABs using WSI compared with LM examination

The inter‐reader agreement for Bethesda categorization of thyroid FNABs on WSIs was 0.48 on the basis of the available data from 92 cases, as calculated by using the Fleiss κ statistic. The corresponding inter‐reader agreement was 0.56 for LM examination of 88 cases. Therefore, the inter‐reader agreement of Bethesda categorization was moderate for both WSI and LM examination, although it was more accurate for LM examination than for WSI.

The inter‐reader Fleiss κ agreements for the individual Bethesda categories I–VI, established by reviewing WSIs of thyroid FNABs, were 0.65, 0.47, 0.05, 0.13, 0.22, and 0.73, respectively (lowest for category III and highest for category VI). The corresponding agreements using conventional LM examination were 0.75, 0.55, 0.25, 0.05, 0.14, and 0.80, respectively (lowest for category IV and, similar to WSIs, highest for category VI). The inter‐reader agreement results are shown in Table 2.

TABLE 2.

Comparison of inter‐reader agreement for Bethesda categorization of ultrasound‐guided thyroid fine‐needle aspiration biopsies, determined with the Fleiss κ statistic using whole‐slide imaging and light‐microscopic examination.

Bethesda category	Fleiss κ: Readers 1–5
	WSI	LM
I	0.65	0.75
II	0.47	0.55
III	0.05	0.25
IV	0.13	0.05
V	0.22	0.14
VI	0.73	0.80
Overall	0.48	0.56

Abbreviations: Bethesda category, The Bethesda System for Reporting Thyroid Cytopathology; LM, light‐microscopic examination; WSI, whole‐slide image.

Time taken for interpreting WSIs compared with LM examination of US‐guided thyroid FNABs

The five readers recorded interpretation times in 92, 72, 72, 95, and 98 cases, respectively. The average time for interpreting WSIs ranged from 3.3 to 10.9 minutes; whereas for LM examination, it ranged from 2.4 to 7.0 minutes. One of the five readers took 3.3 minutes to review WSIs compared with 3.2 minutes for LM examination; all remaining four readers took significantly longer to interpret WSIs than LM slides (p < .0001).

Issues related to WSI acquisition were identified in 30 of the 98 cases, in which one or more slides required rescanning. In two cases, a single slide had to be rescanned four times. The most common reasons for rescanning included blurry initial images caused by focus point errors resulting from issues such as improperly coverslipped slides with the coverslip extending beyond the slide, mounting medium present at the edges or on top of the coverslip, the presence of air bubbles, and physical damage in the slides, such as cracks. Most of the issues were resolved with re‐focusing and scanning the slides again. However, autofocusing proved to be a challenge in some instances, particularly when staining was very faint in smears that were paucicellular, in which case they remained blurry.

These technical challenges were partially reflected in comments from the readers interpreting the slides, who noted inconsistent slide quality and reported issues ranging from focal blurring to overall poor images in occasional cases. Additional difficulties were encountered when evaluating nuclear features on Pap‐stained smears and excessively dark‐stained smears. No technical issues were reported with the Aperio software. However, the overall interpretation of the cases was not compromised despite encountering selected images of suboptimal quality.

DISCUSSION

Our multi‐reader prospective feasibility study using WSIs for ROSE of US‐guided thyroid FNABs, including Pap‐stained and Diff‐Quik–stained smears, demonstrated excellent intrareader agreement for Bethesda categorization (κ ranging from 0.86 to 0.95), thereby indicating the noninferiority of WSIs compared with LM examination. Although the Bethesda categorization was nearly similar for each reader using either WSIs or LM examination, there was significant variation between readers, which resulted in moderate inter‐reader agreement (WSI, κ = 0.48; LM examination, κ = 0.56).

The literature assessing the concordance between WSIs and LM examination is limited, particularly in the context of thyroid FNABs, but available studies report findings that are consistent with our results. A recent systematic review analyzing 19 studies that directly compared WSI versus LM diagnoses reported a concordance rate of 84.1% between WSI and the original LM‐rendered diagnoses. ²² The mean intra‐observer concordance was 92.5%, with a κ coefficient of 0.66, indicating substantial agreement. ¹⁸ Among these 19 heterogeneous studies, only two specifically evaluated thyroid FNABs. In these, the intra‐observer concordance between WSIs and LM examination were 77.5% and 62%–100%. ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ In our current study, the average time for interpreting WSIs ranged from 3.3 to 10.9 minutes, whereas interpretation using LM examination ranged from 2.4 to 7 minutes. Four of the five readers took significantly longer to interpret WSIs than LM slides. In the systematic review reported by Girolami et al., 11 studies reported the time required to screen a digital slide, and seven compared it directly with an LM slide. ²² The mean diagnostic time was reported to be generally longer for WSIs than for LM examination. Unlike most of the previous studies, the WSIs in our study were acquired at ×20 magnification and not at ×40 magnification. ⁸ , ¹⁹ , ²¹ , ²³

All readers in our study observed that the image quality of WSIs, acquired by scanning the glass slides at ×20 magnification, was at least acceptable for evaluation. One of the commonly reported limitations of using WSIs in cytology is the challenge of evaluating thick fragments scanned in a single focal plane, thereby necessitating z stacking to assess three‐dimensional cell clusters in greater detail. ⁸ , ²² , ²³ Because such tissue fragments and three‐dimensional clusters are more often encountered in FNABs, z stacking is reported to be needed for optimal interpretation of FNABs. In the case of liquid‐based cytology preparations, z stacking is reported to be beneficial by the majority of the investigators because small clusters requiring deep focus can also be present in liquid‐based cytology specimens. ²⁶ , ²⁷ However, Kim et al. recently observed that z‐stacked images of thin preparations from urine specimens provided minimal diagnostic benefit. ²⁸ In contrast to the findings from previously published literature using WSIs for final reporting of FNABs, all readers in our current study observed that scanning WSIs in a single plane was sufficient for ROSE in most cases. Z stacking was needed in only 23% of cases. The ability to adequately assess thinner portions of the smear, even in slides containing thick fragments and three‐dimensional clusters, enabled comprehensive evaluation of the FNABs. Previous studies of the use of WSIs for interpreting thyroid FNABs have similarly concluded that z stacking is not essential for satisfactory review. It is worth noting that the whole‐slide scanner used in our study (Leica CS2), as well as the hybrid platforms currently available for telecytology directed towards ROSE of FNABs, have the capability to perform z stacking when needed. This feature can facilitate the optimal evaluation of thick clusters and three‐dimensional groups in selected cases. However, it should be recognized that performing z stacking increases the overall scanning time and thus may not be ideal for telecytology applications pertaining to ROSE.

The moderate inter‐reader agreement for Bethesda categorization of US‐guided thyroid FNABs using WSIs or LM examination resulted from variations in categorization across both modalities. The lowest inter‐reader agreement among the five readers using WSIs was observed for category III (AUS). The κ values for both category III (AUS) and category IV (follicular neoplasm/suspicious for follicular neoplasm) indicated no agreement or only slight agreement. Category V (suspicious for malignancy) showed fair agreement, whereas categories I (nondiagnostic) and VI (papillary thyroid carcinoma) showed substantial agreement. When using LM examination, the inter‐reader κ values for individual Bethesda categories showed both similarities and differences compared with those observed with WSIs: similar to WSIs, substantial agreement was observed for categories I and VI, and moderate agreement was observed for category II (benign). The lowest agreement was noted for category IV, with fair agreement for category III and no or slight agreement for categories IV and V.

Our findings of less‐than‐optimal interobserver agreement for the indeterminate and suspicious Bethesda categories (III, IV, and V), regardless of modality, are consistent with those of previous studies that used LM examination. ²⁹ , ³⁰ , ³¹ Several reports have highlighted poor reproducibility for these categories. For example, Kocjan et al. assessed interobserver agreement for 200 thyroid FNABs stained with May–Grunwald–Giemsa and interpreted by six readers using the UK Royal College of Pathologists' classification system, which is comparable to the Bethesda system. ²⁹ Those readers reported good agreement for the nondiagnostic and malignant categories, moderate agreement for the benign category, and poor agreement for the AUS and suspicious for malignancy categories—findings that were closely aligned with ours. ²⁴ , ²⁵ , ^{26 , 29} , ³⁰ , ³¹ Very recently, Slowinska‐Klencka et al. reported low reproducibility of Bethesda categories III, IV, and V using smears obtained from 213 thyroid nodules that were read by five experienced readers from three centers, with poor overall interobserver agreement. ^{27 , 32} Olson et al. conducted one of the largest reported studies to date that clearly elucidates and quantifies the frequency and magnitude of interobserver variability in Bethesda categorization; they compared the Bethesda classification of the submitting pathologist with that rendered in the tertiary referral center for 3885 thyroid FNABs and observed that the classification changed 32% of the time. ^{28 , 33}

The results of our study highlight the persistent issue of significant variation in the interpretation of thyroid FNABs, regardless of the modality used. Despite the widespread adoption of the standardized Bethesda system, the inherent subjectivity in interpretation and categorization can lead to discordant results among readers. A consensus review of thyroid FNABs may offer a viable solution to address the low interobserver agreement associated with Bethesda categorization. Jing et al. demonstrated that a group consensus review reduced the number of AUS/follicular lesion of uncertain significance diagnoses, improved interobserver agreement, and promoted excellent cytohistologic correlation, resulting in a diagnostic accuracy of 89.2%. ^{29 , 34} The use of WSIs can facilitate consensus review, either in real time or asynchronously, which may help improve interobserver variation in the evaluation of thyroid FNABs. Another promising strategy for enhancing interobserver concordance involves the use of decision‐support tools developed using deep learning–based artificial intelligence. ³⁰ , ^{31 , 35} , ³⁶ These tools are advancing rapidly, and their integration into standard of care will depend on their performance metrics and the extent of validation required for regulatory approval.

Although the overall quality of WSIs was deemed suitable for the interpretation of thyroid FNABs, all readers noted inconsistencies in smear quality that affected optimal interpretation. These inconsistencies were attributed to suboptimal smear preparation and variation in staining quality. Standardizing FNAB specimen preparation through methods that promote the uniform distribution of cellular material, combined with automated staining protocols, may help achieve more reliable results, particularly when using telecytology for ROSE. Recently, Duke et al. reported the results of a prospective study using an automated specimen‐preparation and staining device based on spray technology. ^{32 , 37} They reported a concordance of 96.8% with conventionally prepared smears. Techniques are currently under investigation that will be optimal for standardizing specimen preparation and staining if telecytology is used for ROSE. The time taken for scanning is directly proportional to the number and size of the smears. Preparing a limited number of high‐quality slides may contribute significantly toward the successful use of WSIs for telecytology of thyroid FNABs.

In summary, the findings of our prospective study comparing WSIs with LM examination for the rapid assessment of US‐guided thyroid FNABs demonstrated substantial intrareader agreement, indicating the noninferiority of WSIs compared with LM examination. The moderate inter‐reader agreement using either WSIs or LM examination reiterates the well recognized issue related to subjectivity in thyroid FNAB interpretation, particularly for the indeterminate categories. Our study highlights the suitability of WSIs scanned at ×20 magnification for ROSE without the need for z stacking or ×40‐magnification scans in all cases. The use of limited slides of high quality may reduce the time taken to complete ROSE within an acceptable timeline in clinical practice. A significant advantage of using WSIs for ROSE of thyroid FNABs includes the possibility of obtaining additional opinions from experts in challenging cases.

In conclusion, WSI is a feasible alternative to LM examination for ROSE of thyroid FNABs, with high intrareader agreement and comparable inter‐reader agreement. The limited need for high magnification and z stacking supports its practical utility.

AUTHOR CONTRIBUTIONS

Mohammed S. Ahmed: Investigation, writing–original draft, formal analysis, data curation, methodology, writing–review and editing. Dianna Klippel Almaraz: Investigation, formal analysis, data curation, methodology, writing–review and editing. Sara E. Amin: Investigation, methodology, data curation, and writing–review and editing. Gloria H. Sura: Writing–review and editing, methodology, data curation, and investigation. Uma Kundu: Investigation, writing–review and editing, methodology, and data curation. Wendong Yu: Investigation, writing–review and editing, data curation, and methodology. John M. Stewart: Investigation, writing–review and editing, data curation, and methodology. Qiong Gan: Investigation, writing–review and editing, methodology, and data curation. Savitri Krishnamurthy: Conceptualization, investigation, funding acquisition, writing–review and editing, methodology, formal analysis, project administration, supervision, and data curation.

CONFLICT OF INTEREST STATEMENT

The authors disclosed no conflicts of interest.

Ahmed MS, Klippel‐Almaraz D, Amin SE, et al. Utility of whole‐slide imaging for rapid evaluation of thyroid FNA: A multireader prospective study. Cancer Cytopathol. 2025;e70046. doi: 10.1002/cncy.70046