Digital transformation and artificial intelligence integration in cytopathology: From standardization to emerging semi-quantitative paradigms

Dear Editor,

Cytopathology is entering a transformative era, driven by rapid advances in digital pathology and artificial intelligence (AI). Once dependent on manual screening and subjective interpretation, it is now starting to evolve toward standardized, data-rich, and reproducible diagnostic workflows. Digitization, initially introduced for efficiency and archiving, is increasingly becoming the foundation for quantitative cellular analysis and AI-assisted interpretation.^[1,2]

The earliest milestones of this shift include automated imaging systems such as the ThinPrep Imaging System^® (TIS), which combine algorithm-guided field review with human expertise. In our previous study, TIS improved sensitivity for detecting atypical glandular cells compared with manual screening, underscoring the capability of digital augmentation to enhance cytologic interpretation.^[3] These systems established a foundation for objective triage, reproducibility, and structured review, cornerstones of AI readiness.^[4]

Over the past decade, digital cytopathology research has begun to extend beyond automated screening toward AI-assisted standardization and interpretive analytics, with several proof-of-concept studies exploring quantitative surrogates such as staining intensity, nuclear morphology, and subcellular localization. Although cytopathology has not yet fully adopted quantitative frameworks, lessons from histopathologic AI models, such as H-score-based scoring paradigms, suggest that similar approaches could eventually be adapted for cellular preparations. As AI-based paradigms continue to evolve, they have the potential and emerging tendency to integrate morphological assessment with numerical precision, moving cytopathology toward a semi-quantitative framework. To contextualize this transformation, the following section defines the concept of “emerging semi-quantitative paradigms,” which underpins the manuscript’s central theme. In this context, the term “emerging semi-quantitative paradigms” refers to algorithmic or AI-assisted frameworks that generate composite numerical indices (such as H-score-like metrics) combining staining intensity and cellular proportions. These paradigms bridge qualitative interpretation and quantitative analysis, enhancing reproducibility while retaining diagnostic context. This editorial outlines the evolution from early imaging systems to emerging AI-based interpretive and quantitative models, explores validation needs, and envisions a future where cytopathology operates at the intersection of morphology, mathematics, and molecular data.^[5]

DIGITAL TRANSFORMATION IN CYTOPATHOLOGY

Cytopathology entered the digital era gradually, beginning with traditional microscopy and fully manual screening, which shaped the early diagnostic workflow. This early phase established the foundation of routine practice and set the stage for later advances that would standardize preparation and enable digital review. The transition became more pronounced with the introduction of liquid-based cytology (LBC), which facilitated monolayer preparations and high-resolution scanning.^[2,4] Subsequent innovations, such as BD FocalPoint^™ and the TIS, enabled automated field selection and quality control, demonstrating how machine assistance can reduce human error and interobserver variation.^[4,6] Key milestones in this digital evolution are summarized in Table 1.

Table 1:

Evolution of cytopathology from qualitative morphology to emerging semi-quantitative and integrated digital paradigms.

Stage	Era/period (key milestones)	Milestone	Core features	Conceptual shift
1. Conventional microscopy	Pre-2000 (pre-digital era)	Manual screening using optical microscopes	Subjective interpretation, limited reproducibility	Qualitative morphology
2. Liquid-based cytology	Late 1990s–2000s (ThinPrep® FDA approval, 1996; SurePath®, 1999)	Introduction of monolayer preparations enabling clearer, standardized samples	Enhanced visualization, artifact reduction, and digital readiness	Standardization
3. Automated imaging systems	2000s–2010s (TIS FDA approval, 2003; BD FocalPoint™ integration, mid-2000s)	ThinPrep Imaging System®, BD FocalPoint™	Algorithm-guided field review, quality control systems, and reproducibility	Semi-digital transition
4. AI-based screening	Early 2020s (Genius Digital Diagnostics launch, 2021)	AI-assisted triage platforms, deep learning-based slide review	Improved throughput, interobserver concordance, assisted screening	AI integration
5. Semi-quantitative/quantification paradigms	Mid-2020s (AI-derived intensity-based scoring models, 2023–2024)	H-score-like models incorporating intensity×proportion	Composite indices, reduced subjectivity, reproducibility	Semi-quantitative transition
6. Integrated digital cytopathology	Future (beyond 2025)	Synergy of morphology, molecular diagnostics, and AI	Predictive modeling, data-driven precision pathology	Integration and precision

FDA: Food and Drug Administration, AI: Artificial intelligence

The table outlines key milestones in the digital transformation of cytopathology, combining chronological landmarks with conceptual shifts, from manual microscopy and liquid-based cytology to AI-assisted standardization, interpretive analytics, and emerging semi-quantitative scoring frameworks. This hybrid timeline reflects the discipline’s gradual transition toward data-informed and precision-oriented practice.

Beyond automation, AI integration has extended digitization from visualization to interpretation. The CytoJournal 2025 review highlighted how LBC now intersects with multi-omics and machine learning, positioning cytology at the core of computational pathology.^[2] Similarly, international white papers emphasize that AI-based screening augments rather than replaces human expertise, enhancing consistency, triage accuracy, and workflow efficiency.^[5,7]

Telecytology represents another critical pillar of digital transformation. Validation guidelines from the American Society of Cytopathology provide frameworks for ensuring diagnostic concordance across remote systems.^[8] Recent reviews confirm that cloud-based telecytology enhances real-time consultation, education, and access to subspecialty expertise, especially in resource-limited settings.^[9] Collectively, these developments illustrate a discipline transitioning from microscope-bound observation to algorithm-informed interpretation.^[4,9]

AI-DRIVEN STANDARDIZATION AND EMERGING INTERPRETIVE PARADIGMS

The current phase of digital cytopathology emphasizes AI-driven standardization, aiming to minimize observer variability and enhance diagnostic reproducibility.^[5,7] Beyond field selection and automation, deep learning models increasingly assist in cellular segmentation, nuclear classification, and pattern recognition, transforming digitized slides into structured, analyzable datasets.^[2,4] Such interpretive algorithms, supported by curated image repositories and consensus-labeled datasets, foster consistency across laboratories and pave the way for observer-independent evaluations.^[5,7]

As these interpretive frameworks mature, explainable AI tools are being introduced to visualize decision pathways and strengthen trust in automated outputs.^[2,5] This paradigm marks a shift from static visualization toward analytical interpretation. AI currently serves as a decision-support tool that assists cytopathologists in highlighting diagnostically relevant regions and quantitatively evaluating subtle morphological differences; in the future, it may also contribute to generating probabilistic differential diagnoses.^[7] Together, these developments position AI not as a replacement but as a cognitive partner. Importantly, the irreplaceable role of the cytologist remains central, especially in morphological integration, contextual interpretation, quality assurance, and clinical decision-making, while AI primarily reinforces transparency, reproducibility, and clinical confidence.^[2,5,7]

TOWARD QUANTITATIVE AND SEMI-QUANTITATIVE FRAMEWORKS

Although cytopathology has not yet adopted fully quantitative frameworks, insights from histopathologic AI models offer a roadmap for future integration.^[5] In tissue-based studies, deep learning-assisted scoring paradigms inspired by the H-score have successfully combined staining intensity and positive cell proportions into composite numerical indices, improving objectivity and interobserver agreement.^[10] This histopathologic progress underscores a broader potential: If analogous algorithms are adapted to cytologic preparations, AI-augmented quantification could enhance cellular assessment by translating morphological patterns into reproducible metrics.^[5,10]

In our recently published study on meningiomas,^[11] digital quantification of hormone receptor expression, performed through a percentage- and intensity-weighted H-score approach integrating weak, moderate, and strong staining tiers, demonstrated high concordance with manual evaluation and reduced interobserver variability, illustrating how validated, semi-quantitative systems in histopathology can inform analogous cytologic applications in the near future.

The progression from early digital imaging to AI-based interpretive and quantitative paradigms sheds light on how cytopathology may evolve from descriptive morphology toward measurable, evidence-based precision.

FROM MORPHOLOGY TO DATA INTEGRATION: THE ROAD AHEAD

Looking forward, the convergence of standardization, interpretive analytics, and quantitative modeling is reshaping the foundation of cytopathologic practice. As multi-omics integration, telecytology, and cloud-based computational pipelines become increasingly accessible,^[2,8,9] AI systems are increasingly poised to unify morphological, mathematical, and molecular dimensions into a single diagnostic continuum.^[1,2] The evolution from descriptive observation to measurable precision reflects not only a technological advance but also a conceptual redefinition that envisions cytopathology operating at the intersection of morphology, mathematics, and molecular data.^[5,9]

AVAILABILITY OF DATA AND MATERIALS

Not applicable. No datasets were generated or analyzed for this editorial.

ABBREVIATIONS

AI: Artificial intelligence

FDA: Food and Drug Administration

LBC: Liquid-based cytology

TIS: ThinPrep Imaging System

AUTHOR CONTRIBUTIONS

The author solely conceived, drafted, and approved the final version of the manuscript. Author meets the ICMJE author qualifications.

Funding Statement

FUNDING: Not applicable.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

Not applicable. This editorial does not involve human or animal subjects.

CONFLICTS OF INTEREST

The author declares no conflict of interest.

EDITORIAL/PEER REVIEW

To ensure the integrity and highest quality of CytoJournal publications, the review process of this manuscript was conducted under a double-blind model (authors are blinded for reviewers and vice versa) through an automatic online system.