Application of artificial intelligence‐generated clinical cases in case‐based learning of histology and embryology

Abstract

To investigate the effectiveness of a case‐based learning (CBL) approach in histology and embryology, integrating instructor‐reviewed, AI‐assisted clinical cases within a Structure–Function–Clinical (SFC) framework. Undergraduate students from Hangzhou City University's 2022 cohort (control, n = 150) and 2023 cohort (experimental, n = 147) participated. The experimental group engaged in an SFC‐oriented CBL approach, utilizing instructor‐reviewed, AI‐generated clinical cases systematically integrated into the curriculum. The control group received conventional lecture‐based instruction without case activities. Effectiveness was evaluated by comparing academic performance (formative assessments, midterms, finals, overall grades) and student satisfaction. The experimental group significantly outperformed the control group in formative assessments, midterms, and overall course grades. While total final exam scores showed no significant difference, the experimental group had a higher proportion of high‐scoring students. Regarding satisfaction surveys, the experimental group reported significantly higher ratings in specific dimensions, including learning engagement, clinical relevance, SFC clarity, and perceived effectiveness (all q < 0.05), although overall satisfaction scores did not differ significantly after correction (q = 0.056). Notably, these improvements were achieved without increasing perceived cognitive load (q > 0.05). An SFC‐oriented CBL approach, supported by instructor‐reviewed, AI‐assisted cases, effectively enhances students' ability to apply histological and embryological knowledge in clinical contexts without increasing cognitive load. This method successfully bridges foundational morphological knowledge with clinical practice, offering a practical, scalable model for integrating AI into basic medical education.

INTRODUCTION

Histology and embryology constitute foundational courses in undergraduate medical education, providing the essential morphological basis for understanding human structure and function. ¹ , ² However, students commonly perceive the content as abstract and disconnected from clinical practice, leading to diminished motivation and difficulty in long‐term retention of knowledge. Although prior educational reforms have incorporated digital slides and online resources—significantly improving access to and convenience of microscopic image study—the persistent gap between morphological knowledge and its clinical application remains inadequately addressed. ³ , ⁴ , ⁵

Case‐based learning (CBL) significantly enhances the relevance, engagement, and applicability of knowledge by contextualizing theoretical concepts and microscopic structures within authentic clinical scenarios. ⁶ , ⁷ , ⁸ Building on this pedagogical approach, our teaching team has revised the laboratory manual by systematically integrating clinical cases into each chapter. This revised textbook has been designated as a provincial “14th Five‐Year Plan” educational resource. Implementation has demonstrated that laboratory instruction enriched with clinical cases has been widely endorsed by students. Extending this strategy, we have further incorporated discipline‐aligned clinical cases into the theoretical lectures of histology and embryology to deepen students' conceptual understanding and facilitate knowledge transfer and application.

In recent years, the application of AI in medicine has become increasingly mature, with its reliability and effectiveness in areas such as diagnostic support, medical image analysis, and clinical decision‐making widely validated. ⁹ , ¹⁰ , ¹¹ , ¹² However, the labor‐intensive development of high‐quality clinical cases remains a major bottleneck for scalable implementation. Recent studies indicate that Large Language Models (LLMs) demonstrate robust accuracy and reliability in histology and embryology. Bolgova et al. confirmed LLMs' efficacy in addressing professional queries on embryonic development and teratogenic mechanisms, while Mavrych et al. reported high consistency and scientific accuracy in LLM‐generated descriptions of microscopic structures and functional explanations. ¹³ , ¹⁴ These findings support leveraging LLMs as auxiliary tools for case generation.

Accordingly, this study employed LLMs integrated with Structure–Function–Clinical (SFC) prompt strategies and rigorous instructor review to efficiently generate scientifically accurate and pedagogically suitable clinical cases for histology and embryology lectures. This innovative approach aims to overcome traditional efficiency constraints, fostering students' clinical integration and deep understanding of morphological knowledge without compromising quality.

METHODS

Study design and participants

This study involved undergraduate medical students from two consecutive cohorts at Hangzhou City University who were enrolled in the Histology and Embryology course during their second year of study. The control group consisted of students from the 2022‐entry cohort (n = 150), who completed the course in 2023 and received conventional didactic instruction. The experimental group comprised students from the 2023‐entry cohort (n = 147), who took the course in 2024 and participated in case‐based teaching during theoretical lectures, in which clinical cases—initially drafted by a large language model and subsequently reviewed and refined by instructors according to the Structure–Function–Clinical (SFC) framework—were systematically integrated. Both groups had completed identical foundational medical courses prior to enrollment, and their admission academic profiles (e.g., national college entrance exam scores) were comparable (Table 1).

TABLE 1.

Participant characteristics.

Number of participants (percentage)
Gender	2022 (n = 150)	2023 (n = 147)
Male	86 (57.33%)	84 (57.14%)
Female	64 (42.67%)	63 (42.86%)
Admission scores/750	597.8 ± 4.31	599.7 ± 3.82
Human anatomy scores/100	57.00 ± 17.79	62.06 ± 9.62

This study was approved by the Medical Ethics Committee of Hangzhou City University (Approval No. [2024] Consent Letter 13, September 3, 2024). Informed consent was implied upon questionnaire submission, following a notice detailing the study's purpose, exclusive academic use of data, voluntary participation, and the right to withdraw without academic penalty. To ensure privacy, no direct identifiers were collected, and academic performance data were anonymized prior to analysis.

AI‐based clinical case generation and review workflow

This study employed the Qwen2 large language model to generate histology clinical cases within a closed‐loop quality control system featuring “triple‐prompt engineering” and a “three‐tier audit.” Generation was guided by prompts enforcing framework constraints (clinical elements and Structure–Function–Clinical [SFC] modules), knowledge anchoring (core curriculum concepts), and cognitive adaptation (sophomore‐level difficulty). All 48 AI‐generated cases (three per topic) underwent rigorous dual‐expert review assessing scientific accuracy, SFC logical integrity, and pedagogical suitability. Inter‐rater reliability was quantified via the Intraclass Correlation Coefficient (ICC); only cases achieving strong consensus (ICC > 0.80) across all dimensions were selected for instruction, ensuring strict alignment with learning objectives (Supplementary File 1).

Teaching process design

In this study, the Histology and Embryology course for both groups was delivered by the same core teaching team. Comprising two instructors, each with over five consecutive years of experience teaching this course, the team was responsible for identical teaching chapters across both student cohorts. This arrangement effectively ensured fundamental consistency in teaching content, instructional style, and evaluation criteria between the two groups.

In the instructional design (Figure 1), the control group received standard teacher‐centered, lecture‐based instruction, in which core histological concepts were systematically delivered, supplemented by demonstrations using digital slides to illustrate microscopic structures and help students acquire fundamental morphological knowledge. In contrast, the experimental group participated in an integrated, AI‐assisted case‐based instructional approach. Each lecture session began with an AI‐generated clinical case closely aligned with the histological topic of that session. Students worked in small groups to analyze the case, drawing upon prior knowledge to discuss the histological basis of the pathological changes and present their reasoning process in class. Instructors facilitated the discussion as needed, guiding students toward key learning objectives and promoting horizontal integration between foundational biomedical knowledge and clinical practice. This instructional design aimed to enhance the contextual relevance, interactivity, and cognitive depth of learning, thereby shifting students' focus from rote memorization of morphological features toward understanding functional correlates and applying knowledge in clinically meaningful contexts.

FIGURE 1.

Implementation plan for case‐based learning in histology and embryology.

Academic assessment and instructional feedback

To comprehensively evaluate student learning outcomes and systematically collect instructional feedback, this study employed a multimodal assessment approach combining multidimensional academic evaluation with a structured survey. Student academic performance was assessed across four components: (1) formative assessment (30%, comprising class participation and quizzes), (2) midterm examination (20%), (3) final examination (50%), and (4) overall course grade. In accordance with institutional academic regulations, if a student scored below 50% on the final examination, their overall course grade was recorded as the final examination score alone, thereby ensuring adherence to the minimum competency threshold for core knowledge.

Strict blinded assessment and standardized scoring protocols were employed to minimize bias. Summative exam papers were anonymized and randomly coded for independent blind grading. The temporal separation of the 2022 and 2023 cohorts naturally masked group allocation, preventing expectancy bias by obscuring ‘experimental’ versus ‘control’ labels. In formative assessments, objective items were auto‐graded, while subjective tasks (e.g., presentations) followed structured rubrics to ensure standardization. Additionally, test instruments were validated for equivalence in format and difficulty to maintain inter‐group comparability.

At the conclusion of the instructional cycle, student feedback was collected using a structured questionnaire (Supplementary File 2). 105 valid responses were recovered from each group. The instrument comprised 17 items organized into six domains: learning engagement, clinical relevance, clarity of the SFC framework, perceived learning effectiveness, overall satisfaction, and cognitive load. Responses were recorded on a 5‐point Likert scale (1 = strongly disagree, 5 = strongly agree). Domain scores were calculated as the mean of the corresponding item scores, and an overall experiential score was derived by taking the arithmetic mean of the six domain scores.

Statistical analysis and data reliability

Data management and primary statistical analyses were performed using GraphPad Prism version 10.0 (GraphPad Software, San Diego, CA, USA), while reliability and agreement analyses were conducted with IBM SPSS Statistics version 27.0 (IBM Corp., Armonk, NY, USA). Continuous variables were expressed as mean ± standard deviation (SD). The normality of data distribution was assessed via the Shapiro–Wilk test combined with visual inspection of Q‐Q plots. Based on the confirmed normal distribution, independent‐samples t‐tests were applied to compare intergroup differences in academic performance (formative assessment, midterm exam, final exam, and overall course grade) and subjective questionnaire scores across domains.

Statistical significance was initially set at p < 0.05. To quantify the magnitude of intergroup differences, Hedges' g effect sizes with 95% confidence intervals (CIs) were calculated and interpreted according to conventional criteria: small effect (|g| ≈ 0.2), medium effect (|g| ≈ 0.5), and large effect (|g| ≥ 0.8). To address the risk of Type I errors arising from multiple comparisons across distinct outcome measures, the Benjamini‐Hochberg procedure was used to adjust raw p‐values for false discovery rate (FDR) control. Statistical significance was ultimately defined as an adjusted p‐value (q) < 0.05 for all analyses.

For data reliability evaluation, the Cronbach's α coefficient was used to test the internal consistency of each domain in the student satisfaction questionnaire, with a coefficient of α > 0.7 considered acceptable reliability and α > 0.8 indicating good reliability. In addition, to assess the inter‐rater reliability of the AI‐generated clinical case quality evaluation (including scientific accuracy, SFC logical integrity, and pedagogical suitability), the intraclass correlation coefficient (ICC) was computed. Given that two fixed instructors independently rated all cases, a two‐way mixed‐effects model for absolute agreement [ICC (1, 3)] was adopted. ICC values were interpreted as follows: <0.50 (poor), 0.50–0.75 (moderate), 0.75–0.90 (good), and >0.90 (excellent). Only cases with strong inter‐rater consensus (ICC > 0.80) across all evaluation dimensions were selected for instructional application.

RESULTS

Student performance

To evaluate the impact of clinical case‐based instruction on student learning outcomes, this study conducted a systematic analysis of performance across multiple assessment components between the experimental and control groups (Figure 2A) (Tables S1 and S2). In terms of formative assessment scores, the experimental group performed significantly higher than the control group (91.34 ± 7.11 vs. 86.42 ± 8.94; p < 0.000001, q < 0.000001, Hedges' g = 0.61), indicating that AI‐supported clinical case‐based instruction effectively enhanced students' daily engagement and quality of task completion. The score distribution (Figure 2B) showed that no students in the experimental group scored below 60, whereas 2% of the control group remained in this range. In the high‐score brackets (≥80), the proportions of students in the experimental group achieving scores of 80–89 and 90–100 were 19.05% and 71.43%, respectively—significantly higher than the corresponding proportions in the control group (37.33% and 50.67%), suggesting that AI‐supported clinical case‐based teaching substantially increased the proportion of students attaining high achievement levels.

FIGURE 2.

Comparison of academic performance between the experimental group (n = 147) and the control group (n = 150). (A) Performance of the two groups on formative assessments, midterm examination, final examination, and overall course grade. (B) Distribution of overall course grades across score ranges (<60, 60–69, 70–79, 80–89, ≥90). (C) Comparison of performance during the formative learning phase (maximum score = 50 points, comprising formative assessments and midterm examination).

In the midterm examination, the experimental group significantly outperformed the control group (79.37 ± 12.17 vs. 63.30 ± 10.91; p < 0.000001, q < 0.000001, Hedges' g = 1.39), reflecting a substantial positive impact of the instructional intervention on students' mastery of course content at the midpoint of the semester. The failure rate (score < 60) decreased from 27.33% in the control group to 10.20% in the experimental group, while the proportion of students achieving high scores (≥80) rose to 62.59% in the experimental group—markedly higher than the 6.67% observed in the control group. These results indicate that students who received case‐based instruction during the initial phase of the course had already demonstrated a clear advantage in their early academic performance.

No significant difference was observed in final examination scores between the two groups (experimental group: 65.24 ± 14.81 vs. control group: 64.75 ± 15.11; p = 0.776, q = 0.776, Hedges' g = 0.03) (Figure 2A), possibly because the final exam primarily assessed factual knowledge rather than higher‐order cognitive skills, thereby limiting its capacity to detect differences in applied competence. However, an analysis of score distributions revealed that, although the proportions of students in the low‐score range (<60) were comparable between groups (32.67% in the control group vs. 32.65% in the experimental group), the experimental group maintained an advantage in the higher‐score brackets: 17.01% versus 12.67% in the 80–89 range and 4.08% versus 1.33% in the 90–100 range (Figure 2B). This suggests that a subset of students in the experimental group retained stronger integrative and analytical capabilities despite the limitations of the assessment format.

Despite the similarity in final examination scores, the experimental group achieved a significantly higher overall course grade than the control group (73.31 ± 14.96 vs. 68.79 ± 14.56; p = 0.009, q = 0.012, Hedges' g = 0.31) (Figure 2A), reflecting the cumulative impact of the instructional intervention on long‐term learning outcomes.

Analysis of the overall grade distribution (Figure 2B) further revealed a more pronounced advantage for the experimental group in the high‐score ranges: 38.10% of students scored 80–89 (vs. 19.33% in the control group) and 4.08% scored 90–100 (vs. 7.33% in the control group), resulting in a combined proportion of 42.18% of students achieving grades ≥80—higher than that of the control group (26.66%). Additionally, the proportion of students in the low‐score range (<60) decreased from 14.67% in the control group to 12.24% in the experimental group, further indicating that this teaching approach not only enhances overall academic performance but also mitigates the risk of academic underachievement.

Further analysis of the combined score from formative assessments and the midterm examination—which together account for approximately 50% of the overall course grade—revealed a rightward shift in the score distribution for the experimental group, with a stronger central tendency and a markedly higher concentration of students in the high‐score ranges (Figure 2C). This pattern reflects greater stability in learning performance, deeper engagement, and superior knowledge integration throughout the formative phase of the course. These findings suggest that case‐based instruction enhances students' consistency, depth of participation, and coherence in knowledge acquisition during formative learning.

Satisfaction questionnaire

To evaluate students' subjective experiences, a satisfaction survey was administered (Table S3). Prior to analysis, internal consistency was verified, yielding Cronbach's α coefficients ≥0.74 for all domains. Specifically, Engagement (0.841), Clinical Relevance (0.842), SFC Clarity (0.821), and Perceived Learning (0.821) demonstrated good reliability (α > 0.80), while Cognitive Load (0.740) and Satisfaction (0.754) fell within the acceptable range. These results confirm the instrument's reliability for subsequent statistical analysis.

Based on these reliable data, the experimental group reported significantly higher scores than the control group across multiple domains (Figure 3A). Specifically, regarding learning engagement, the experimental group reported markedly higher scores (4.38 ± 0.61 vs. 3.55 ± 0.63; p < 0.000001, q < 0.000001, Hedges' g = 1.33), indicating that clinical cases effectively stimulated students' interest and active involvement. Similarly, students in the experimental group demonstrated a significantly stronger perception of clinical relevance (4.33 ± 0.65 vs. 3.52 ± 0.48; p < 0.000001, q < 0.000001, Hedges' g = 1.43), reflecting an enhanced appreciation of the applicability of histological and embryological knowledge within authentic clinical contexts.

FIGURE 3.

Comparison of student learning experiences between the experimental group (n = 105) and the control group (n = 105). (A) Comparison of mean scores between the two groups across six dimensions: Learning engagement, clinical relevance, clarity of the SFC framework, perceived learning effectiveness, cognitive load, and overall satisfaction. (B) Comparison of overall composite scores. (C) Effect size analysis (Hedges' g) for each individual dimension.

With regard to the perceived clarity of the Structure–Function–Clinical (SFC) relationships, the experimental group scored significantly higher than the control group (4.21 ± 0.57 vs. 3.74 ± 0.56; p < 0.000001, q < 0.000001, Hedges' g = 0.82). Students in the experimental group also reported a more positive perception of their own learning effectiveness (4.20 ± 0.58 vs. 3.87 ± 0.55; p = 0.000015, q = 0.000021, Hedges' g = 0.61), suggesting that AI‐supported clinical case‐based instruction facilitated the construction of a coherent cognitive framework linking morphological structures to their physiological functions and clinical relevance, thereby enhancing knowledge integration. Overall satisfaction showed a nominal increase in the experimental group compared to the control group (4.36 ± 0.60 vs. 4.20 ± 0.58; p = 0.048, q = 0.056, Hedges' g = 0.27), although this difference did not reach statistical significance after FDR correction. More importantly, no significant difference was observed between the two groups in perceived cognitive load (3.85 ± 0.50 vs. 3.98 ± 0.51; p = 0.085, q = 0.085, Hedges' g = −0.24), indicating that the instructional innovation enhanced the learning experience without imposing additional mental burden on students.

Overall (Figure 3B), the experimental group demonstrated a significantly higher composite satisfaction score than the control group (4.26 ± 0.50 vs. 3.84 ± 0.46; p < 0.000001, q < 0.000001, Hedges' g = 0.86). As shown in Figure 3C, the most pronounced improvements were observed in learning engagement and perceived clinical relevance, while clarity of the SFC framework and perceived learning effectiveness also exhibited medium to large effect sizes. The score distribution for the experimental group was more tightly clustered and shifted rightward compared to that of the control group, indicating that the instructional intervention yielded systematic and widespread enhancements in the learning experience, benefiting the majority of students.

DISCUSSION

Despite the widespread adoption of digital slides, which has significantly improved access to microscopic images, a persistent gap remains between morphological knowledge and clinical reasoning in histology and embryology education. Students often struggle to connect static structures with dynamic functions and clinical applications. To address this challenge, our study implemented a novel instructional approach that systematically integrates instructor‐reviewed, AI‐generated clinical cases into theoretical lectures within a Structure–Function–Clinical (SFC) framework. Unlike traditional case development which is time‐consuming and difficult to scale, our “AI‐generation + instructor oversight” model leverages the efficiency of large language models while ensuring scientific accuracy and pedagogical appropriateness through rigorous faculty validation. This approach was designed to bridge the gap between basic science and clinical practice without overwhelming students or instructors.

A key innovation of this study lies in the effective utilization of AI to overcome the bottleneck of high‐quality case production. Our results demonstrate that AI, when guided by specific SFC‐oriented prompts and constrained by expert review, can efficiently generate diverse and clinically authentic cases that align precisely with learning objectives. The efficiency of this collaborative workflow was evident in our time‐motion analysis: the initial review of AI drafts averaged approximately 10 min per case, while the subsequent finalization required only about 4 min. This represents a substantial reduction in preparation time compared to developing cases from scratch. However, efficiency did not come at the cost of rigor. The instructor review process played a critical role in defining the cognitive boundaries of the curriculum. For instance, several AI‐generated cases on Celiac disease were excluded during validation. While scientifically accurate, these cases elaborated excessively on cellular and molecular mechanisms—such as the activation mechanism of intraepithelial lymphocytes in the intestinal tract, the abnormal expression of tight junction proteins, and the immune pathway mediated by tissue transglutaminase—which exceeded the macroscopic pathological associations required for this undergraduate course. This highlights a crucial finding: AI tends to maximize informational density, whereas educators must curate content to match the learners' developmental stage. The high inter‐rater reliability among instructors during the review process confirms that AI‐generated content can meet strict academic standards when subjected to such a structured quality control workflow. This finding suggests that AI serves not as a replacement for educators, but as a powerful collaborative tool that amplifies their capacity to create rich, context‐based learning materials at scale. By automating the initial drafting of cases, AI allows instructors to focus on higher‐order tasks such as refining clinical logic and filtering out content that falls outside the intended scope, thereby making CBL feasible for large undergraduate cohorts.

The most compelling evidence of the intervention's success comes from the students' subjective learning experiences. Our survey results indicate that the experimental group reported significantly higher levels of learning engagement and perceived clinical relevance compared to the control group, with large effect sizes (Hedges' g > 1.3). This suggests that the AI‐assisted cases successfully transformed abstract histological concepts into meaningful clinical narratives, thereby stimulating student interest and motivation. Furthermore, students in the experimental group perceived the SFC framework as clearer and reported greater learning effectiveness, indicating that the cases helped them construct a coherent cognitive map linking structure, function, and disease. Crucially, these improvements in engagement and perceived utility were achieved without increasing perceived cognitive load. This finding addresses a common concern in educational innovation: that adding complexity (e.g., clinical cases) might overwhelm learners. Our data suggest that the scaffolded nature of the AI‐generated cases, combined with instructor guidance, provided an optimal level of challenge that enhanced learning without inducing cognitive overload. The more tightly clustered distribution of satisfaction scores in the experimental group further implies that this benefit was equitable, reaching students across different proficiency levels.

Consistent with the positive shifts in student engagement and perception, we observed tangible improvements in academic performance. The experimental group significantly outperformed the control group in formative assessments and midterm examinations, reflecting enhanced daily engagement and earlier mastery of course content. Although no significant difference was found in the final examination scores—likely due to the exam's focus on factual recall rather than higher‐order reasoning—the experimental group still demonstrated a higher proportion of students achieving top grades (≥80). More importantly, the overall course grade was significantly higher in the experimental group, driven by the cumulative advantages gained throughout the semester. driven by the cumulative advantages in formative assessment and midterm examination (accounting for 50% of the overall course grade) gained throughout the semester. This pattern reinforces the notion that the primary value of AI‐assisted CBL lies in fostering sustained, deep learning and consistent participation, rather than merely boosting short‐term memorization for a final test.

In summary, our study provides empirical support for a scalable model of AI integration in medical education. By prioritizing student engagement and clinical relevance through carefully curated AI‐generated cases, we effectively enhanced the learning experience and academic outcomes in histology and embryology. The synergy between AI efficiency and instructor expertise offers a practical pathway for reforming foundational medical courses, ensuring that technological advancements serve to deepen, rather than dilute, the educational experience. Future work should explore the long‐term retention of knowledge and the transfer of these competencies to clinical rotations, but the current findings strongly advocate for the responsible adoption of AI as a catalyst for active, context‐rich learning.

LIMITATIONS AND FUTURE DIRECTIONS

This study has several limitations. First, the non‐randomized, quasi‐experimental design across consecutive cohorts introduces potential temporal confounders (e.g., instructor drift or post‐pandemic behavioral shifts) that, despite baseline comparability, may affect causal inference, warranting future cross‐over or multi‐year randomized studies to isolate intervention effects. Second, while AI‐generated cases underwent rigorous review, their quality remains dependent on manual curation in the absence of a standardized automated protocol, and the specific prompt engineering strategies may require adaptation for replication in settings with different resources. Third, the discrepancy between improved formative engagement and unchanged summative exam scores suggests a misalignment between assessment tools and learning outcomes; specifically, final exams focusing on factual recall likely lacked the sensitivity to fully capture the potential for developing higher‐order clinical reasoning skills scaffolded by the intervention, highlighting the need for future assessments such as clinical vignettes or OSCEs. Finally, generalizability is limited by the single‐institution setting and the specific Chinese linguistic and clinical context, necessitating multicenter studies across diverse systems to validate the framework's global applicability. Despite these constraints, this study provides empirical support for a responsible “AI‐generation + instructor review” model, offering a practical framework for fostering an environment conducive to clinical reasoning in foundational science education while identifying key areas for future methodological refinement.

CONCLUSION

As artificial intelligence reshapes medical education, the central challenge lies not merely in generating content but in ensuring its pedagogical fidelity. This study validates a practical dual‐track model where the efficiency of AI generation is balanced by rigorous instructor curation. Our findings demonstrate that this approach significantly elevates student engagement and fosters deep conceptual integration—enabling learners to better connect structural knowledge with functional applications—while crucially maintaining cognitive load at manageable levels. It offers a concrete reference for modernizing basic medical education, ensuring that technological innovation serves to deepen, rather than dilute, educational quality.

AUTHOR CONTRIBUTIONS

SSL: Conceptualization, writing – original draft, writing – review & editing. XDZ: Investigation, data visualization, writing – original draft, writing – review & editing. WDS: Writing – review & editing. DYZ: Conceptualization, data curation, funding acquisition, writing – original draft, writing – review & editing.

FUNDING INFORMATION

This study was supported by the Zhejiang Province “14th Five‐Year” Undergraduate Education Textbook Planning Project – Histology, Embryology and Pathology Laboratory Manual (2nd edition) [no grant number]; the 2020 Zhejiang Province Offline First‐Class Course Construction Project (Histology and Embryology) [no grant number]; the Zhejiang Provincial Undergraduate Teaching Reform Project [Grant No. JGCG2025380]; and the Hangzhou City University Digital Intelligence Course Construction Project (Grant No. 2024TP021 and SZKC2510).

CONFLICT OF INTEREST STATEMENT

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Supporting information

Supplementary File 1:

Supplementary File 2:

Table S1. Comparison of academic performance between control and experimental groups in the histology and embryology course.

Table S2. Score distribution percentage of academic performance in different grade ranges between the control and experimental groups.

Table S3. Comparison of student subjective evaluation scores between control and experimental groups.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.