The DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index.
Accuracy of GPT-4 in histopathological image detection and classification of colorectal adenomas
Abstract
AIMS: To evaluate the accuracy of Chat Generative Pre-trained Transformer (ChatGPT) powered by GPT-4 in histopathological image detection and classification of colorectal adenomas using the diagnostic consensus provided by pathologists as a reference standard. METHODS: A study was conducted with 100 colorectal polyp photomicrographs, comprising an equal number of adenomas and non-adenomas, classified by two pathologists. These images were analysed by classic GPT-4 for 1 time in October 2023 and custom GPT-4 for 20 times in December 2023. GPT-4's responses were compared against the reference standard through statistical measures to evaluate its proficiency in histopathological diagnosis, with the pathologists further assessing the model's descriptive accuracy. RESULTS: GPT-4 demonstrated a median sensitivity of 74% and specificity of 36% for adenoma detection. The median accuracy of polyp classification varied, ranging from 16% for non-specific changes to 36% for tubular adenomas. Its diagnostic consistency, indicated by low kappa values ranging from 0.06 to 0.11, suggested only poor to slight agreement. All of the microscopic descriptions corresponded with their diagnoses. GPT-4 also commented about the limitations in its diagnoses (eg, slide diagnosis best done by pathologists, the inadequacy of single-image diagnostic conclusions, the need for clinical data and a higher magnification view). CONCLUSIONS: GPT-4 showed high sensitivity but low specificity in detecting adenomas and varied accuracy for polyp classification. However, its diagnostic consistency was low. This artificial intelligence tool acknowledged its diagnostic limitations, emphasising the need for a pathologist's expertise and additional clinical context.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“The study also revealed that the model is dependent on the expert’s advice, which is a drawback in fully automated clinical scenarios.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“Laohawetwanit et al. (26) used GPT-4 to detect colorectal adenomas in histopathological images. GPT-4 had some potential with sensitivity of 74% and specificity of 36%.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“The findings on diagnostic inconsistency (κ = 0.06–0.11), reliance on expert prompting, and constraints in image-based reasoning directly reveal the difficulties associated with implementing matching LLM systems for skin biopsies and dermatopathology procedures.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“The findings on diagnostic inconsistency (κ = 0.06–0.11), reliance on expert prompting, and constraints in image-based reasoning directly reveal the difficulties associated with implementing matching LLM systems for skin biopsies and dermatopathology procedures.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“In another study by Laohawetwanit et al. (26), GPT-4 was evaluated for diagnosing histopathological images, achieving a sensitivity of 74% and a specificity of 36%.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“Collectively, these two studies demonstrate a continuous pattern in the literature: models exhibit robust performance under optimum or curated settings but experience significant declines when subjected to real-world variability or whole-slide complexity.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“A study by Laohawetwanit et al. (26) evaluated GPT-4 for the diagnosis of colorectal adenomas, rather than skin conditions.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“However, this model still suffers from a lack of applicability and is heavily dependent on expert input, making it unsuitable for fully automated systems.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“The findings on diagnostic inconsistency (κ = 0.06–0.11), reliance on expert prompting, and constraints in image-based reasoning directly reveal the difficulties associated with implementing matching LLM systems for skin biopsies and dermatopathology procedures.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“The findings on diagnostic inconsistency (κ = 0.06–0.11), reliance on expert prompting, and constraints in image-based reasoning directly reveal the difficulties associated with implementing matching LLM systems for skin biopsies and dermatopathology procedures.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
In the citing paper
“This study was selected due to its unique and pertinent evidence on GPT-4’s performance in histopathological image interpretation, a modality also utilized in dermatopathology.”
From Revolutionizing dermatopathology using AI in skin diagnostics: scoping review (2026) · cites this paper as [26]
Citations
Cites
Rudin, C. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0048-x · Cited by 10,561
Sallam, M. (2023) · Healthcare · doi:10.3390/healthcare11060887 · Cited by 3,126 · cited 1×
“22 Pathologists must be aware of these potential limitations and act as gatekeepers to maximise benefits and minimise harm from AI tool use.”
BackgroundLong-term colorectal-cancer incidence and mortality after lower endoscopy
Nishihara, R., Wu, K., Lochhead, P., Morikawa, T., Liao, X., Qian, Z. R., Inamura, K., Kim, S. A., Kuchiba, A., Yamauchi, M., et al. (2013) · New England Journal of Medicine · doi:10.1056/nejmoa1301969 · Cited by 1,483
Digital pathology and artificial intelligence
Niazi, M. K. K., Parwani, A. V., & Gurcan, M. N. (2019) · The Lancet Oncology · doi:10.1016/s1470-2045(19)30154-8 · Cited by 1,232
Screening for Colorectal Cancer
Lin, J. S., Perdue, L. A., Henrikson, N. B., Bean, S., & Blasi, P. R. (2021) · JAMA · doi:10.1001/jama.2021.4417 · Cited by 661
Artificial intelligence in diagnostic pathology
Shafi, S., & Parwani, A. V. (2023) · Diagnostic Pathology · doi:10.1186/s13000-023-01375-z · Cited by 371
Suárez, A., García, V. D., Algar, J., Sánchez, M. G., de Pedro, M. L., & Freire, Y. (2023) · International Endodontic Journal · doi:10.1111/iej.13985 · Cited by 234
Applicability of ChatGPT in Assisting to Solve Higher Order Problems in Pathology
Sinha, R. K., Deb Roy, A., Kumar, N., & Mondal, H. (2023) · Cureus · doi:10.7759/cureus.35237 · Cited by 187
Schukow, C. P., Smith, S. C., Landgrebe, E., Parasuraman, S., Folaranmi, O. O., Paner, G. P., & Amin, M. B. (2023) · Advances in Anatomic Pathology · doi:10.1097/pap.0000000000000406 · Cited by 87
Weak Human Preference Supervision for Deep Reinforcement Learning
Cao, Z., Wong, K., & Lin, C. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3084198 · Cited by 69
Research works printed but not linked: 1
- OpenAI . GPT-4 technical report. 2023. Available: https://arxiv.org/pdf/2303.08774.pdf · arXiv
Cited by
Sorted by citations
Large Language Models in Medical Diagnostics: Scoping Review With Bibliometric Analysis
Su, H., Sun, Y., Li, R., Zhang, A., Yang, Y., Xiao, F., Duan, Z., Chen, J., Hu, Q., Yang, T., et al. (2025) · Journal of Medical Internet Research · doi:10.2196/72062 · Cited by 69
Applications of Large Language Models in Pathology
Cheng, J. (2024) · Bioengineering · doi:10.3390/bioengineering11040342 · Cited by 56
ChatGPT‐4 Consistency in Interpreting Laryngeal Clinical Images of Common Lesions and Disorders
Maniaci, A., Chiesa-Estomba, C. M., & Lechien, J. R. (2024) · Otolaryngology · doi:10.1002/ohn.897 · Cited by 26
Assessing the accuracy of the GPT-4 model in multidisciplinary tumor board decision prediction
Erdat, E. C., Yalçıner, M., Örüncü, M. B., Ürün, Y., & Şenler, F. Ç. (2025) · Clinical & Translational Oncology · doi:10.1007/s12094-025-03905-1 · Cited by 24
Evaluating ChatGPT’s diagnostic potential for pathology images
Ding, L., Fan, L., Shen, M., Wang, Y., Sheng, K., Zou, Z., An, H., & Jiang, Z. (2025) · Frontiers in Medicine · doi:10.3389/fmed.2024.1507203 · Cited by 25
Metadata
- Authors
- Laohawetwanit, T., Namboonlue, C., & Apornvirat, S.
- Affiliations
- Thammasat University (TH)
- Dr Pong Clinic, Bangkok, Thailand
- Corresponding author
- Laohawetwanit, Thiyaphat
- Venue
- Journal of Clinical Pathology
- Publisher
- BMJ
- Published
- 2024-01-10
- Volume
- 78
- Issue
- 3
- Pages
- 202-207
- ISSN
- 0021-9746 (ISSN-L)
- 1472-4146
- OpenAlex
- W4390708138
- Cited by
- 35
- References
- 20
- Citations by year
- 2026: 4 · 2025: 14 · 2024: 17
- FWCI
- 5.17
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Journal of Clinical Pathology (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Health Sciences › Medicine › Oncology › Colorectal Cancer Screening and Detection
- Keywords
- GPT-4, large language models, colorectal adenoma, colorectal polyps, histopathological image classification, adenoma detection, polyp classification, diagnostic accuracy, diagnostic consistency, artificial intelligence in pathology
- MeSH terms
- Adenoma, Colonic Polyps, Humans, Image Interpretation, Computer-Assisted, Sensitivity and Specificity, Colorectal Neoplasms, Reproducibility of Results, Observer Variation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
