Authors

Beniamin Di Veroli, Richard Lederman, Jacob Sosna, Leo Joskowicz

Abstract

Radiological follow-up of oncological patients requires the analysis and comparison of multiple unregistered scans acquired every few months. This process is currently often partial, time-consuming and subject to variability. We present a new, generic, graph-based method for tracking individual lesion changes and detecting patterns in the evolution of lesions over time. The tasks are formalized as graph-theoretic problems in which lesions are vertices and edges are lesion pairings computed by overlap-based lesion matching. We define seven individual lesion change classes and five lesion change patterns that fully summarize the evolution of lesions over time. They are directly computed from the graph properties and its connected components with graph-based methods. Experimental results on lung (83 CTs from 19 patients) and liver (77 CECTs from 18 patients) datasets with more than two scans per patient yielded an individual lesion change class accuracy of 98% and 85%, and identification of patterns of lesion change with an accuracy of 96% and 76%, respectively. Highlighting unusual lesion labels and lesion change patterns in the graph helps radiologists identify overlooked or faintly visible lesions. Automatic lesion change classification and pattern detection in longitudinal studies may improve the accuracy and efficiency of radiological interpretation and disease status evaluation.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43904-9_11

SharedIt: https://rdcu.be/dnwGO

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #2

Please describe the contribution of the paper

The paper presents a follow-up lesion ontology for oncological lesions in CT scans. The ontology serves as base for case-specific automatically constructed knowledge graphs. The possible graph characteristics in terms of connected components and node degrees give rise to change categories of individual lesions and lesion change patterns over time. The method is evaluated on two in-house datasets.
Please list the main strengths of the paper; you should write about 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 structured very well and contains a detailed and sound mathematical formalism.
- The presented lesions over time modeling scheme includes aspects and features that are beyond current clinical practice that rather concentrates on lesion volumetry.
- The method isn’t a black box approach and focuses on transparency and traceability of the automatically computed assessment.
- There is conceptual and clinically relevant novelty in the method.
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
- No evaluation on any publicly available benchmark data set.
Please rate the clarity and organization of this paper

Excellent
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance
- Re-implementation could become difficult as algorithms to generate the case-specific graph-structure are complex, too.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html
- Page 2, last paragraph: which of the five listed novelties “this task” refers to?
- Page 2, between “2)” and “3)”: the term “lesion changes” appears too often. This sentence should be rephrased.
- The listed five novelties appear overambitious. Could the authors be more concise and focus on the most relevant ones? E.g., 5) (= experiments with good results) can be found – in one way or another – in most other scientific publications. Experiments per se aren’t a contribution but will support the major claims.
- Is there some special handling for FP and FN lesions as mentioned in the caption of Figure 1?
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

7
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

In between all the deep learning it is good to again see a classical and transparent method. Capturing longitudinal semantics of cancerous lesions with the supposed graph-based method makes sense to me. Assessing and documenting the additionally identified lesion development use cases may even have a clinical impact as different change patterns may ask for different further patient management.
Reviewer confidence

Very confident
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

Review #1

Please describe the contribution of the paper

The authors propose a graph-based technique for tracking of lesions in longitudinal CT studies. This method was evaluated on lung and liver lesion datasets.
Please list the main strengths of the paper; you should write about 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.

This is a clinically feasible technique. The use of graph optimization adds to the paper’s originality and the reported results are good.
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

One perceived weakness is that this technique requires lesion segmentation as input by the user. This is often times a difficult and time-consuming task.
Please rate the clarity and organization of this paper

Good
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

The method is described in sufficient detail to support reproducibility.
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html

The lesion tracking technique is quite interesting and seems that it could be integrated into a clinical workflow. I wonder if it would be useful to employ a probabilistic graphical model, either via neural networks or not, to predict changes of lesion states.
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

5
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

Nice contribution with some limitations mainly regarding the other pre-requisite data for its application.
Reviewer confidence

Confident but not absolutely certain
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

Review #3

Please describe the contribution of the paper

Adapted from MOT, this paper proposes a graph-based approach to model changes in lesion follow-up in CT images. Using manually segmented lesions, authors define 7 changes in individual lesion and then define 5 patterns to reflect all lesions changes.
Please list the main strengths of the paper; you should write about 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 idea is clear and well translated into algorithms with graphs. The approach is tested on two datasets. Results and discussion are convincing on the usage of the approach (about 2000 lesions analyzed, it helps experts to refine their annotation) The fact that segmentation is done manually is clear and allow to focus on the proposed graph approach.
Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

The approach needs a very good pairwise registration. This point is partially addressed in §2.2. The authors use deformable registration which can modify locally the contrast and thus create a deformable field that will drastically modify the lesion segmentation. The two datasets suffer of class imbalance that not facilitate the analysis. The approach sensibility to its 4 parameters is not discussed.
Please rate the clarity and organization of this paper

Very Good
Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

DLIVER and DLUNG availability …
Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html

The registration in fig 1 seems to be not perfect for the 3 times points (ribs, kidney and liver don’t match). It can be interesting to highlight in the legend or in the text that registration is not perfect but the approach works. In §3, authors should add the unity of d and \delta (pixels or mm) as voxels are not isotropic. The approach needs 4 parameters, their tuning/optimization should be discussed or explained. For future works, other pathologies such as Multiple Sclerosis can be considered. Maybe a typo error on the sentence just before ‘Study 2’ in §3.
Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

6
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

This paper is clear and well written. Results are convincing.
Reviewer confidence

Confident but not absolutely certain
[Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

N/A
[Post rebuttal] Please justify your decision

N/A

Primary Meta-Review

Please provide your assessment of this work, taking into account all reviews. Summarize the key strengths and weaknesses of the paper and justify your recommendation. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.

This paper proposed a new Graph-theoretic automatic lesion tracking method and were evaluated using two internal datasets. All three reviews are positive but the evaluation scale in terms of unique patient numbers is considered as rather small.

Author Feedback

N/A

back to top

Graph-theoretic automatic lesion tracking and detection of patterns of lesion changes in longitudinal CT studies