List of Papers Browse by Subject Areas Author List
Abstract
Gastroscopic Lesion Detection (GLD) is one of the critical tasks within computer-assisted gastroscopic diagnostics. Each category of gastroscopic lesions have several sub-category patterns and endoscopists adopt a pattern-based philosophy for GLD: they identify and summarize typical sub-category patterns with specific medical meanings and conduct GLD based on these patterns. The current gastroscopic lesion detectors follow the classical data-driven deep-learning-based training paradigm, which differs from the endoscopists’ diagnosis process and leads to low interpretability, limiting their performance and potential for daily clinical practice and patient care. Directly integrating the philosophy into GLD tasks requires re-annotating the whole GLD datasets with sub-category labels, which is prohibitively expensive. An affordable solution is to annotate limited sub-category labels and train the detectors in a few-shot manner. To address the challenge, we propose a Pattern-Anchored Adaptive Prototype Learning (PAAPL) for Gastroscopic Lesion Detection. Specifically, to integrate the pattern-based philosophy, PAAPL introduces a Prototype-based Gastroscopic Lesion Detector (PGLD) for prototype-based sub-category pattern detection and a Pattern-Anchored Adaptive Learning (PAAL) method to learn prototypes from GLD datasets with limited sub-category pattern annotations. The PAAL can soft-anchor prototypes to limited-annotated patterns with specific medical meanings and adaptively learn pattern characteristics from unannotated data samples in GLD datasets. We evaluate PAAPL on the LGLDD and Endo21 datasets, demonstrating its ability to learn and detect sub-category patterns trained with limited annotated examples. By doing this, PAAPL enhances detector interpretability and yields significant performance improvement (+3.7AP on LGLDD/+5.4AP on Endo21).
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/1441_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{ZhaXua_PatternAnchored_MICCAI2025,
author = { Zhang, Xuanye and Hu, Xiaoqing and Li, Guanbin and Liu, Si-Qi and Wan, Xiang and Xiong, Yuanhuan},
title = { { Pattern-Anchored Adaptive Prototype Learning for Gastroscopic Lesion Detection and Beyond } },
booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
year = {2025},
publisher = {Springer Nature Switzerland},
volume = {LNCS 15968},
month = {September},
page = {437 -- 446}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper addresses a relevant challenge in the field of gastroscopic lesion detection, specifically targeting the performance limitations of current state-of-the-art methods. The authors propose a novel approach that is structured into two main components, designed to align with the general clinical workflow of gastroscopic lesion detection and to mitigate the common issue of limited annotated data. The proposed method is evaluated using two datasets, and the results are presented through both qualitative and quantitative analyses to demonstrating the potential effectiveness of the approach.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
The paper introduces a new approach that is built upon existing state-of-the-art methods for gastroscopic lesion detection. The proposed solution is well-validated, with evaluation performed on two datasets, supported by both qualitative and quantitative results.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
The paper is somewhat difficult to follow, which affects the overall readability. Additionally, the authors do not provide any discussion on how the proposed approach could be integrated into real clinical workflows or how it could assist medical professionals during gastroscopic procedures, which limits the practical relevance and potential impact of the work.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not provide sufficient information for reproducibility.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(3) Weak Reject — could be rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The paper proposes a new approach with promising results, but it lacks essential details for assessing its clinical relevance. The integration into real clinical workflows is not addressed, and the size and complexity of the proposed network are not discussed, which raises questions about its feasibility. Moreover, the potential benefits for clinicians compared to existing state-of-the-art methods are not clearly explained.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
-
The manuscript provides an anchor-based prototypical approach for gastroscopic lesion detection (GLD) that closely follows the endoscopists’ diagnostic process that relies heavily on the subcategory patterns.
-
The dataset augmented with 84 annotated examples of the pattern, which is used as anchors for the learnable prototypes. These prototypes further improve the reasoning of the lesion proposals coming from a Region Proposal Network.
-
The reported results show improvement on two datasets and further shows adaptability to variations in lesion patterns.
-
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
-
The paper is well organized and easy to follow.
-
The idea of utilizing the patterns most commonly used by endoscopists as an anchor is a well thought approach to augment and improve detection performance.
-
Results on t-SNE shows separability of patterns discovered in the data and that contributes to model performance improvement on the datasets. Per-class performance is higher in the cases (ulcer and cancer) where the dataset reports more variations in sub-category patterns.
-
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
-
Based on the results in Table 1., the model seems to perform well on challenging cases of “uls” and “can” with varied patterns, but not on the patterns “pol” or “smt”. The reasoning provided in Page 7 is not sufficient to explain why it underperforms on these patterns even though the annotations of these patterns are provided. Is it because the annotations for these patterns are not optimal?
- How do the quality and quantity of annotated examples affect detection performance? For instance, varying the proportion of the 84 examples—or including additional ones—would have provided a more complete analysis of the point at which diminishing returns begin to set in.
- How does the value of N affect the performance? When N is low or high, does Situation 1 or Situation 2 happen more frequently? The authors should clarify more on this. Also, section 2.3 can be improved and it is sometimes unclear in few places. For example, “identify the less similar between them as the more unique example”, this can be clarified in a better way.
- Typos in the manuscript such as “brunch” in most of the places can be fixed.
- The values of loss weights are not mentioned in the manuscript.
-
- Please rate the clarity and organization of this paper
Good
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — could be accepted, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The manuscript sets up the need of gastroscopic lesion detectors with particular emphasis on the attributes that a general endoscopists look for. With few annotated examples for prototype development, the proposed method improves over different styles of detectors particularly on cancer and ulcer classes that have more morphological variations. However, there are small details about the ablations, choice of N, typos that needs to be addressed.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
This paper proposes Pattern-Anchored Adaptive Prototype Learning (PAAPL) for gastroscopic lesion detection (GLD), aiming to integrate the pattern-based diagnostic philosophy of endoscopists into a deep learning framework with minimal sub-category pattern annotations. The key motivation is to enhance interpretability and performance while reducing annotation costs. PAAPL consists of two main components: Prototype-based Gastroscopic Lesion Detector (PGLD): Introduces a prototype branch after the Region Proposal Network (RPN), enabling sub-category lesion detection based on similarity to learned pattern prototypes. These prototypes are obtained using contrastive clustering. Pattern-Anchored Adaptive Learning (PAAL): Anchors prototypes to limited-annotated patterns with medical relevance and adaptively refines them using unannotated data. It features a novel vector-wise prototype formulation and an update mechanism based on similarity thresholds to ensure stability and mitigate forgetting. The paper concludes with a comprehensive ablation study that demonstrates the superior performance of the proposed approach.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
Strengths The major strength of the paper lies in its effective translation and integration of clinical pattern-based diagnostic philosophy into a structured deep learning framework. Notably, this is achieved without the need for extensive manual annotation, which significantly enhances the practicality of the approach. The proposed method is flexible, rich in medically relevant information, and reproducible. The use of softer annotations combined with contrastive and adaptive learning forms an interesting methodology.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
The main weakness lies in the claim of enhanced interpretability, which is not sufficiently supported by concrete evidence or quantitative analysis. Additionally, while the annotation requirements are reduced, the method still relies on human-provided, and potentially subjective, pattern annotations—introducing a dependency that may affect scalability and generalizability
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2025/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The idea of integrating clinical diagnostic philosophy into a more rigid deep learning framework is compelling. The methodology proposed to achieve this integration is well-structured and technically sound
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
N/A
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.
N/A