Globalization of a telepathology network with artificial intelligence applications in Colombia: The GLORIA program study protocol.

In Colombia, cancer is recognized as a high-cost pathology by the national government and the Colombian High-Cost Disease Fund. As of 2020, the situation is most critical for adult cancer patients, particularly those under public healthcare and residing in remote regions of the country. The highest lag time for a diagnosis was observed for cervical cancer (79.13 days), followed by prostate (77.30 days), and breast cancer (70.25 days). Timely and accurate histopathological reporting plays a vital role in the diagnosis of cancer. In recent years, digital pathology has been globally implemented as a technological tool in two main areas: telepathology (TP) and computational pathology. TP has been shown to improve rapid and timely diagnosis in anatomic pathology by facilitating interaction between general laboratories and specialized pathologists worldwide through information and telecommunication technologies. Computational pathology provides diagnostic and prognostic assistance based on histopathological patterns, molecular, and clinical information, aiding pathologists in making more accurate diagnoses. We present the study protocol of the GLORIA digital pathology network, a pioneering initiative, and national grant-approved program aiming to design and pilot a Colombian digital pathology transformation focused on TP and computational pathology, in response to the general needs of pathology laboratories for diagnosing complex malignant tumors. The study protocol describes the design of a TP network to expand oncopathology services across all Colombian regions. It also describes an artificial intelligence proposal for lung cancer, one of Colombia's most prevalent cancers, and a freely accessible national histopathological image database to facilitate image analysis studies.

Regulatory Aspects of AI-ML.

In the realm of healthcare, numerous generative and non-generative artificial intelligence and machine learning (AI-ML) tools have been developed and deployed. Simultaneously, manufacturers of medical devices are leveraging AI-ML. However, the adoption of AI in healthcare raises several concerns, including safety, security, ethical biases, accountability, trust, economic impact, and environmental effects. Effective regulation can mitigate some of these risks, promote fairness, establish standards, and advocate for more sustainable AI practices. Regulating AI tools not only ensures their safe and effective adoption but also fosters public trust. It is important that regulations remain flexible to accommodate rapid advances in this field to support innovation and also not to add additional burden to some of our preexisting and well-established frameworks. This article covers regional and global regulatory aspects of AI-ML including data privacy, Software as a Medical device (SaMD), agency approval and clearance pathways, reimbursement, and laboratory developed tests (LDTs).

Telecytology versus conventional rapid on-site evaluation for endobronchial ultrasound-guided fine needle aspiration: a single institution's experience.

INTRODUCTION: Telecytology (TC) has the advantage of allowing cytopathologists to remotely support multiple sites rapid on-site evaluation (ROSE) concurrently and represents a potential solution for an increased clinical demand for ROSE. In this study, we share our comparative experience of using TC versus conventional (in-person) ROSE for endobronchial ultrasound-guided fine needle aspiration (EBUS-FNA). MATERIALS AND METHODS: We evaluated 475 consecutive cases of EBUS-FNA that underwent TC-ROSE from May 2020 to August 2021 along with 475 consecutive cases which had conventional ROSE from November 2019 to August 2021 at the University of Michigan. Concordant rates of preliminary and final diagnoses were calculated and compared between TC and conventional methods. RESULTS: While there was no significant difference in preliminary diagnostic rates of nondiagnostic, benign, atypical, and malignant categories between the TC and conventional cohorts, a significantly lower proportion of TC cases received a preliminary suspicious for malignancy diagnosis (2%) compared to the conventional group (4%) (P = 0.03). The concordance rate of preliminary and final diagnoses in TC and conventional ROSE was 96% and 94%, respectively. The average total number of passes per procedure did not differ significantly between TC and conventional ROSE (4.9 versus 4.7). While a relatively higher number of TC-ROSE cases collected dedicated passes compared with conventional ROSE (49% versus 40%), the difference was not statistically significant. CONCLUSIONS: The performance of TC-ROSE for EBUS-FNA is comparable to that of conventional ROSE with similar performance metrics and therefore can be used as a feasible substitute.

The Anatomic Pathology Hospitalist Model.

CONTEXT.­: Challenges to staffing a high-quality frozen section service include consolidation of health systems and pathology practices, off-campus relocation of some pathology offices, growing numbers of stand-alone surgery centers, and subspecialization among pathologists and surgeons. To address these challenges, we developed a novel anatomic pathology hospitalist model with explicit emphasis in frozen section. OBJECTIVE.­: To evaluate our anatomic pathology hospitalist program's impact on (1) frozen section staffing, (2) frozen-permanent diagnostic concordance, and (3) turnaround time. DESIGN.­: Frozen section staffing and performance data were collected for the 28-month period spanning July 1, 2021, to October 31, 2023. Outcomes were compared between hospitalists, nonhospitalists, and fellows. RESULTS.­: Hospitalists performed more frozen sections per month than nonhospitalists (median, 87 versus 17, respectively; P = .002). After implementation, nonhospitalists' average frozen section staffing obligation fell from 3.7 (30%) of 12.3 total service days per month to 2.8 (22%) of 12.6 total service days per month (P = .005), compared with hospitalists' average of 9.5 frozen section days (69%) of 13.7 total service days per month. Frozen-permanent concordance was marginally but significantly higher for hospitalists (4701 of 4744 blocks, 99.1%) than nonhospitalists (7259 of 7362 blocks, 98.6%; P = .02). Concordance did not correlate with pathologists' academic rank or subspecialization. Turnaround times were comparable for hospitalists, nonhospitalists, and fellows across multiple metrics. CONCLUSIONS.­: Our anatomic pathology hospitalists significantly reduced the frozen section obligations of nonhospitalist faculty, with a small but significant increase in frozen-permanent concordance and no substantial change in turnaround time.

Validation of AI-assisted ThinPrep® Pap test screening using the GeniusTM Digital Diagnostics System.

Advances in whole-slide imaging and artificial intelligence present opportunities for improvement in Pap test screening. To date, there have been limited studies published regarding how best to validate newer AI-based digital systems for screening Pap tests in clinical practice. In this study, we validated the Genius™ Digital Diagnostics System (Hologic) by comparing the performance to traditional manual light microscopic diagnosis of ThinPrep® Pap test slides. A total of 319 ThinPrep® Pap test cases were prospectively assessed by six cytologists and three cytopathologists by light microscopy and digital evaluation and the results compared to the original ground truth Pap test diagnosis. Concordance with the original diagnosis was significantly different by digital and manual light microscopy review when comparing across: (i) exact Bethesda System diagnostic categories (62.1% vs 55.8%, respectively, p = 0.014), (ii) condensed diagnostic categories (76.8% vs 71.5%, respectively, p = 0.027), and (iii) condensed diagnoses based on clinical management (71.5% vs 65.2%, respectively, p = 0.017). Time to evaluate cases was shorter for digital (M = 3.2 min, SD = 2.2) compared to manual (M = 5.9 min, SD = 3.1) review (t(352) = 19.44, p < 0.001, Cohen's d = 1.035, 95% CI [0.905, 1.164]). Not only did our validation study demonstrate that AI-based digital Pap test evaluation had improved diagnostic accuracy and reduced screening time compared to light microscopy, but that participants reported a positive experience using this system.

Utilization of an artificial intelligence-enhanced, web-based application to review bile duct brushing cytologic specimens: A pilot study.

BACKGROUND: The authors previously developed an artificial intelligence (AI) to assist cytologists in the evaluation of digital whole-slide images (WSIs) generated from bile duct brushing specimens. The aim of this trial was to assess the efficiency and accuracy of cytologists using a novel application with this AI tool. METHODS: Consecutive bile duct brushing WSIs from indeterminate strictures were obtained. A multidisciplinary panel reviewed all relevant information and provided a central interpretation for each WSI as being "positive," "negative," or "indeterminate." The WSIs were then uploaded to the AI application. The AI scored each WSI as positive or negative for malignancy (i.e., computer-aided diagnosis [CADx]). For each WSI, the AI prioritized cytologic tiles by the likelihood that malignant material was present in the tile. Via the AI, blinded cytologists reviewed all WSIs and provided interpretations (i.e., computer-aided detection [CADe]). The diagnostic accuracies of the WSI evaluation via CADx, CADe, and the original clinical cytologic interpretation (official cytologic interpretation [OCI]) were compared. RESULTS: Of the 84 WSIs, 15 were positive, 42 were negative, and 27 were indeterminate after central review. The WSIs generated on average 141,950 tiles each. Cytologists using the AI evaluated 10.5 tiles per WSI before making an interpretation. Additionally, cytologists required an average of 84.1 s of total WSI evaluation. WSI interpretation accuracies for CADx (0.754; 95% CI, 0.622-0.859), CADe (0.807; 95% CI, 0.750-0.856), and OCI (0.807; 95% CI, 0.671-0.900) were similar. CONCLUSIONS: This trial demonstrates that an AI application allows cytologists to perform a triaged review of WSIs while maintaining accuracy.

Are we ready to bridge classification systems? A comprehensive review of different reporting systems in thyroid cytology.

The evaluation of thyroid lesions is common in the daily practice of cytology. While the majority of thyroid nodules are benign, in recent decades, there has been increased detection of small and well-differentiated thyroid cancers. Combining ultrasound evaluation with fine-needle aspiration cytology (FNAC) is extremely useful in the management of thyroid nodules. Furthermore, the adoption of specific terminology, introduced by different thyroid reporting systems, has helped effectively communicate thyroid FNAC diagnoses in a clear and understandable way. In 1996, the Papanicolaou Society thyroid cytological classification was introduced. This was followed in 2005 by the first Japanese and then in 2007 by the Bethesda System for Reporting Thyroid Cytopathology, which subsequently underwent two revisions. Other international thyroid terminology classifications include the British, Italian, Australasian and other Japanese cytology systems. This review covers similarities and differences among these cytology classification systems and highlights key points that unify these varied approaches to reporting thyroid FNAC diagnoses.

Synthetic Data and Its Utility in Pathology and Laboratory Medicine.

In our rapidly expanding landscape of artificial intelligence, synthetic data have become a topic of great promise and also some concern. This review aimed to provide pathologists and laboratory professionals with a primer on the role of synthetic data and how it may soon shape the landscape within our field. Using synthetic data presents many advantages but also introduces a milieu of new obstacles and limitations. This review aimed to provide pathologists and laboratory professionals with a primer on the general concept of synthetic data and its potential to transform our field. By leveraging synthetic data, we can help accelerate the development of various machine learning models and enhance our medical education and research/quality study needs. This review explored the methods for generating synthetic data, including rule-based, machine learning model-based and hybrid approaches, as they apply to applications within pathology and laboratory medicine. We also discussed the limitations and challenges associated with such synthetic data, including data quality, malicious use, and ethical bias/concerns and challenges. By understanding the potential benefits (ie, medical education, training artificial intelligence programs, and proficiency testing, etc) and limitations of this new data realm, we can not only harness its power to improve patient outcomes, advance research, and enhance the practice of pathology but also become readily aware of their intrinsic limitations.

Cytologic features of differentiated high-grade thyroid carcinoma: A multi-institutional study of 40 cases.

BACKGROUND: Differentiated high-grade thyroid carcinoma (DHGTC) is recently recognized by the World Health Organization (WHO) as a subgroup of thyroid carcinomas with high-grade features while retaining the architectural and/or cytologic features of well-differentiated follicular-cell-derived tumors. The cytomorphology of DHGTC is not well documented despite potential implications for patient triage and management. METHODS: The pathology archives of six institutions were searched for cases diagnosed on resection as "high-grade thyroid carcinoma" using WHO criteria. The fine-needle aspiration (FNA) cohort represents a 10-year period (2013-2023); all were reviewed to confirm DHGTC classification. The corresponding FNAs were assessed for 32 cytomorphologic features. RESULTS: Forty cases of DHGTC with prior FNA were identified. The mean patient age was 64.2 years. The average lesion size was 4.9 cm, and the majority demonstrated a TI-RADS score of 4 or 5 (95.2%). Three main high-grade subsets of DHGTC based on corresponding histology included papillary thyroid carcinoma (65%), follicular carcinoma (22.5%), and oncocytic carcinoma (12.5%). Over 97% of FNA cases were classified as Bethesda category IV or above. Approximately 25% of DHGTC showed cytologic features that included marked cytologic atypia, increased anisonucleosis, large oval nuclei, mitotic activity, or necrosis (p < .05); 68% of DHGTC cases were associated with high-risk molecular alterations. TERT mutations occurred in 41%, of which 89% of these were associated with a second mutation, usually RAS or BRAF p.V600E. CONCLUSIONS: Cytology has a low sensitivity for DHGTC, although a subset of DHGTCs have cytologic features raising the possibility of a high-grade thyroid carcinoma. Other findings include high-risk molecular changes and clinicopathologic features such as older patient age and larger lesion size.

Metastases to the parotid glands: An institutional experience.

BACKGROUND: Metastatic lesions to the salivary gland are rare and mostly affect the parotids. Metastases represent 8% of all malignant lesions of the parotid gland. Around 80% originate from squamous cell carcinomas (SCC) of the head and neck region. Fine needle aspiration (FNA) plays a crucial role in distinguishing primary salivary gland lesions from metastases. Herein we describe our series of metastases to the parotid glands. MATERIALS AND METHODS: We analysed 630 parotid gland FNAs over a decade including conventional and liquid-based cytology specimens. Ancillary techniques such as immunocytochemistry (ICC) were conducted on cell blocks. RESULTS: Eighty (12.4%) cases were malignant lesions, of which 53 (63.75%) were metastases including 24% melanoma, 22.6% SCC, 19% renal carcinomas, 7.5% breast carcinomas, 11.3% lung, 9% intestinal and 1.8% testicular, malignant solitary fibrous tumour and Merkel cell carcinoma. The 53 cases, classified according to the Milan system for salivary cytopathology, belonged to 5 Suspicious for malignancy (SFM) and 48 malignant (M) categories. Forty had a known history of primary malignancy (75.4%), while 13 were suspicious to be a metastatic localisation (24.5%), distributed as 5SFM (2SCC and 3Melanoma) and 8 M. A combination of clinical history, cytomorphology and ICC identified 100% of them. CONCLUSIONS: Fine needle aspiration plays a central role in the diagnostic workup of patients with metastatic lesions to their parotid glands, thereby defining the correct management. Diagnostic accuracy may be enhanced by applying ICC. Although melanoma and SCC are the most common histological types, several other malignancies may also metastasize to the parotid glands and should be kept into consideration.

The current state of digital cytology and artificial intelligence (AI): global survey results from the American Society of Cytopathology Digital Cytology Task Force.

INTRODUCTION: The integration of whole slide imaging (WSI) and artificial intelligence (AI) with digital cytology has been growing gradually. Therefore, there is a need to evaluate the current state of digital cytology. This study aimed to determine the current landscape of digital cytology via a survey conducted as part of the American Society of Cytopathology (ASC) Digital Cytology White Paper Task Force. MATERIALS AND METHODS: A survey with 43 questions pertaining to the current practices and experiences of WSI and AI in both surgical pathology and cytology was created. The survey was sent to members of the ASC, the International Academy of Cytology (IAC), and the Papanicolaou Society of Cytopathology (PSC). Responses were recorded and analyzed. RESULTS: In total, 327 individuals participated in the survey, spanning a diverse array of practice settings, roles, and experiences around the globe. The majority of responses indicated there was routine scanning of surgical pathology slides (n = 134; 61%) with fewer respondents scanning cytology slides (n = 150; 46%). The primary challenge for surgical WSI is the need for faster scanning and cost minimization, whereas image quality is the top issue for cytology WSI. AI tools are not widely utilized, with only 16% of participants using AI for surgical pathology samples and 13% for cytology practice. CONCLUSIONS: Utilization of digital pathology is limited in cytology laboratories as compared to surgical pathology. However, as more laboratories are willing to implement digital cytology in the near future, the establishment of practical clinical guidelines is needed.

Second opinion for pulmonary and pleural cytology is valuable for patient care.

INTRODUCTION: Thoracic cytology can be challenging due to limited procured material or overlapping morphology between benign and malignant entities. In such cases, expert consultation might be sought. This study aimed to characterize all pulmonary and pleural cytology consult cases submitted to our practice and provide recommendations on approaching difficult cases. MATERIALS AND METHODS: All thoracic (pulmonary and pleural) cytology cases submitted for expert consultation to the University of Michigan (MLabs) from 2013 to mid-2022 were reviewed. Cases where cytology was only part of a hematopathology or surgical pathology consult were excluded. Patient demographics, specimen location, procedure performed, referring diagnosis, and our diagnoses were recorded for each case. Diagnoses were categorized according to the Papanicolaou Society of Cytopathology recommendations for pulmonary and effusion cytology. Discordant diagnoses were stratified as major or minor. Data was analyzed using chi-square analysis and logistic models. RESULTS: We received 784 thoracic cytology consult cases, including 530 exfoliative samples and 307 fine-needle aspirations. The most common anatomic locations sampled were the bronchial wall (n = 194, 23%), lung nodule (n = 322, 38%), and pleura (n = 296, 35%). 413 cases had a diagnostic discrepancy (48.3%), with 274 (66%) minor and 139 (34%) major discrepancies. By location, pleural effusion specimens had the highest probability of a discrepant diagnosis (P = 0.003). By specimen type, fine-needle aspiration samples were significantly more likely to have a discrepant diagnosis (P = 0.06). CONCLUSION: Nearly half of the thoracic cytology cases submitted for expert second opinion had diagnostic discrepancies. Consequently, consulting a tertiary medical care center with cytopathology expertise for challenging thoracic cytology diagnoses is beneficial.

On image search in histopathology.

Pathology images of histopathology can be acquired from camera-mounted microscopes or whole-slide scanners. Utilizing similarity calculations to match patients based on these images holds significant potential in research and clinical contexts. Recent advancements in search technologies allow for implicit quantification of tissue morphology across diverse primary sites, facilitating comparisons, and enabling inferences about diagnosis, and potentially prognosis, and predictions for new patients when compared against a curated database of diagnosed and treated cases. In this article, we comprehensively review the latest developments in image search technologies for histopathology, offering a concise overview tailored for computational pathology researchers seeking effective, fast, and efficient image search methods in their work.

Automated Quantitative CD8+ Tumor-Infiltrating Lymphocytes and Tumor Mutation Burden as Independent Biomarkers in Melanoma Patients Receiving Front-Line Anti-PD-1 Immunotherapy.

BACKGROUND: CD8+ tumor-infiltrating lymphocyte (TIL) predicts response to anti-PD-(L)1 therapy. However, there remains no standardized method to assess CD8+ TIL in melanoma, and developing a specific, cost-effective, reproducible, and clinically actionable biomarker to anti-PD-(L)1 remains elusive. We report on the development of automatic CD8+ TIL density quantification via whole slide image (WSI) analysis in advanced melanoma patients treated with front-line anti-PD-1 blockade, and correlation immunotherapy response. METHODS: Seventy-eight patients treated with PD-1 inhibitors in the front-line setting between January 2015 and May 2023 at the University of Pittsburgh Cancer Institute were included. CD8+ TIL density was quantified using an image analysis algorithm on digitized WSI. Targeted next-generation sequencing (NGS) was performed to determine tumor mutation burden (TMB) in a subset of 62 patients. ROC curves were used to determine biomarker cutoffs and response to therapy. Correlation between CD8+ TIL density and TMB cutoffs and response to therapy was studied. RESULTS: Higher CD8+ TIL density was significantly associated with improved response to front-line anti-PD-1 across all time points measured. CD8+ TIL density ≥222.9 cells/mm2 reliably segregated responders and non-responders to front-line anti-PD-1 therapy regardless of when response was measured. In a multivariate analysis, patients with CD8+ TIL density exceeding cutoff had significantly improved PFS with a trend toward improved OS. Similarly, increasing TMB was associated with improved response to anti-PD-1, and a cutoff of 14.70 Mut/Mb was associated with improved odds of response. The correlation between TMB and CD8+ TIL density was low, suggesting that each represented independent predictive biomarkers of response. CONCLUSIONS: An automatic digital analysis algorithm provides a standardized method to quantify CD8+ TIL density, which predicts response to front-line anti-PD-1 therapy. CD8+ TIL density and TMB are independent predictors of response to anti-PD-1 blockade.

Changing digital and telecytology practices post COVID-19 comparing ASC survey results from 2016 to 2023.

INTRODUCTION: During the COVID-19 pandemic, the need for digital pathology tools became more urgent. However, there needs to be more knowledge of the use in cytology. We aimed to evaluate current digital cytology practices and attitudes and compare the results with a pre-COVID-19 American Society of Cytopathology (ASC) survey. MATERIALS AND METHODS: Fourteen survey questions assessing current attitudes toward digital cytology were developed from a 2016 ASC Digital Pathology Survey. Ten new survey questions were also created to evaluate telecytology use. The survey was e-mailed to ASC members over 6 weeks in 2023. RESULTS: A total of 123 individuals responded (116 in 2016). Attitudes toward digital cytology were unchanged; most participants stated digital cytology is beneficial (87% 2023 versus 90% 2016). The percentage of individuals using digital cytology was unchanged (56% in 2016 and 2023). However, telecytology for rapid onsite assessment (ROSE) is now considered the best application (55% 2023 versus 31% 2016). Forty-three institutions reported using digital and telecytology tools; 40% made implementations after 2020; most did not feel that COVID-19 affected digital cytology (56%). Telecytology for ROSE is the most common application now (78%) compared with education (30%) in 2016. Limitations for implementing digital imaging in cytology included inability to focus (38%) and expense (33%). CONCLUSIONS: General attitudes toward digital tools by the cytology community have essentially remained the same between 2016 and now. However, telecytology for ROSE is increasingly being used, which supports a need for validation and competency guidelines.

Benchmarking cytology support for ROSE during endoscopic and bronchoscopic procedures.

INTRODUCTION: There has been an increase in endoscopic and bronchoscopic biopsies as minimally invasive methods to obtain specimens from gastrointestinal (GI) or pancreatobiliary lesions and thoracic or mediastinal lesions, respectively. As hospitals undertake more of these procedures, it is important to consider the staffing implications that this has on cytopathology laboratories with respect to support for rapid on-site evaluation (ROSE). MATERIALS AND METHODS: Volume and time data from endoscopic ultrasound and bronchoscopic procedures (including endobronchial ultrasound-guided transbronchial needle aspirations and small biopsies with touch preparation) in the GI suite, bronchoscopy suite, or operating room were reviewed for 2 months at 2 different medical centers with ROSE services provided by cytologists or fellows physically present at the procedure and cytopathologists located remotely using telecytology. Statistical analysis was performed to investigate significant trends based on the location of the biopsies and other factors. RESULTS: A total of 16 proceduralists performed 159 procedures and submitted 276 different specimens during 16 total weeks at 2 institutions. The total ROSE time for the on-site personnel to cover these procedures was 109.3 hours (bronchoscopy, 62.3 hours [57%]; GI, 29.8 hours [27%]; OR, 17.2 hours [16%]), which represents an average of 0.69 hour (41.4 minutes) per procedure or 0.40 hour (24.0 minutes) per part, with the shortest procedure times per sample recorded during bronchoscopy. When stratified by practice volume for individual proceduralists, the average time per specimen sample submitted was shorter for proceduralists with high volume practices and was most pronounced during bronchoscopy procedures. CONCLUSIONS: Endoscopic and bronchoscopic procedures account for an increasing amount of the ROSE time for the cytology team. On average, each ROSE procedure takes 0.69 hour (41.4 minutes) or approximately 0.40 hour (24.0 minutes) per specimen, with shorter time requirements for specimens obtained in bronchoscopy procedures and for operators with high volume practices for endobronchial ultrasound-guided transbronchial needle aspirations. This provides important benchmarking data to calculate staffing needs for cytology to provide ROSE support for different proceduralists.

Performance of preferentially expressed antigen in melanoma (PRAME) immunohistochemistry for metastatic melanoma in cytology specimens.

BACKGROUND: Preferentially expressed antigen in melanoma (PRAME) has been introduced as a new melanoma marker and potential target for immunotherapy. While PRAME immunohistochemistry (IHC) is well documented in surgical pathology, similar data in cytology are limited. Metastatic melanoma is frequently diagnosed via cytology samples in which IHC plays an important role. We aimed to accordingly evaluate the performance of PRAME IHC in diagnosing metastatic melanoma in cytology samples relative to other commonly used melanoma markers. MATERIALS AND METHODS: The study included 156 archival cytology cases, of which 93 were melanoma cases and 63 nonmelanoma cases (controls). All cases underwent PRAME IHC staining on cell blocks. Nuclear staining of PRAME was evaluated using a quantitative and qualitative scale. Other melanocytic IHC stain results (SOX10, S-100, Melan-A, and HMB45) were also documented. RESULTS: PRAME was detected in tumor cells in 86% of melanoma cases, which was significantly lower than SOX10 (100%) (p < .01), and similar to HMB45 (84%) and Melan-A (82%). S-100 had the lowest sensitivity of 71%. In comparison to other types of melanomas, spindle cell melanoma exhibited higher negativity for PRAME IHC (4/10 = 40%). PRAME was also expressed in some nonmelanocytic malignancies including carcinoma (5/22 = 23%), sarcoma (5/15 = 33%), and hematologic malignancies (1/9 = 11%). Overall, PRAME showed a sensitivity of 86%, specificity of 82%, positive predictive value of 70%, and negative predictive value of 92% for metastatic melanoma. CONCLUSIONS: PRAME is a useful marker for the diagnosis of melanoma in cytology material, but it is less sensitive than SOX10. PRAME is also expressed in other nonmelanocytic tumors which limits its specificity.

Utility of artificial intelligence in a binary classification of soft tissue tumors.

Soft tissue tumors (STTs) pose diagnostic and therapeutic challenges due to their rarity, complexity, and morphological overlap. Accurate differentiation between benign and malignant STTs is important to set treatment directions, however, this task can be difficult. The integration of machine learning and artificial intelligence (AI) models can potentially be helpful in classifying these tumors. The aim of this study was to investigate AI and machine learning tools in the classification of STT into benign and malignant categories. This study consisted of three components: (1) Evaluation of whole-slide images (WSIs) to classify STT into benign and malignant entities. Five specialized soft tissue pathologists from different medical centers independently reviewed 100 WSIs, representing 100 different cases, with limited clinical information and no additional workup. The results showed an overall concordance rate of 70.4% compared to the reference diagnosis. (2) Identification of cell-specific parameters that can distinguish benign and malignant STT. Using an image analysis software (QuPath) and a cohort of 95 cases, several cell-specific parameters were found to be statistically significant, most notably cell count, nucleus/cell area ratio, nucleus hematoxylin density mean, and cell max caliper. (3) Evaluation of machine learning library (Scikit-learn) in differentiating benign and malignant STTs. A total of 195 STT cases (156 cases in the training group and 39 cases in the validation group) achieved approximately 70% sensitivity and specificity, and an AUC of 0.68. Our limited study suggests that the use of WSI and AI in soft tissue pathology has the potential to enhance diagnostic accuracy and identify parameters that can differentiate between benign and malignant STTs. We envision the integration of AI as a supportive tool to augment the pathologists' diagnostic capabilities.