RECORD
Optimizing the Procedure of CT Segmentation Labeling
Computer Vision and Pattern Recognition · Submitted 24 Mar 2023
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim We compare the effects of those virtues on the model performance using open medical CT datasets and conclude, that quality is more important than diversity early during labeling; the diversity, in turn, is more important than completeness.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
In Computed Tomography, machine learning is often used for automated data processing. However, increasing model complexity is accompanied by increasingly large volume datasets, which in turn increases the cost of model training. Unlike most work that mitigates this by advancing model architectures and training algorithms, we consider the annotation procedure and its effect on the model performance. We assume three main virtues of a good dataset collected for a model training to be label quality, diversity, and completeness. We compare the effects of those virtues on the model performance using open medical CT datasets and conclude, that quality is more important than diversity early during labeling; the diversity, in turn, is more important than completeness. Based on this conclusion and additional experiments, we propose a labeling procedure for the segmentation of tomographic images to minimize efforts spent on labeling while maximizing the model performance.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | In Computed Tomography, machine learning is often used for automated data processing. | background 1.00 | 0.04 | 0.01 | asserted 0.92 |
| s1 | However, increasing model complexity is accompanied by increasingly large volume datasets, which in turn increases the cost of model training. | background 0.98 | 0.07 | 0.02 | asserted 1.00 |
| s2 | Unlike most work that mitigates this by advancing model architectures and training algorithms, we consider the annotation procedure and its effect on the model performance. | contribution 0.75 | 0.22 | 0.03 | asserted 0.90 |
| s3 | We assume three main virtues of a good dataset collected for a model training to be label quality, diversity, and completeness. | method 0.54 | 0.24 | 0.03 | asserted 0.86 |
| s4 | We compare the effects of those virtues on the model performance using open medical CT datasets and conclude, that quality is more important than diversity early during labeling; the diversity, in turn, is more important than completeness. | result 0.99 | 0.96 | 0.88 | asserted 1.00 |
| s5 | Based on this conclusion and additional experiments, we propose a labeling procedure for the segmentation of tomographic images to minimize efforts spent on labeling while maximizing the model performance. | contribution 0.90 | 0.80 | 0.24 | asserted 0.98 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
In Computed Tomography, machine learning is often used for automated data processing. However, increasing model complexity is accompanied by increasingly large volume datasets, which in turn increases the cost of model training. Unlike most work that mitigates this by advancing model architectures and training algorithms, we consider the annotation procedure and its effect on the model performance. We assume three main virtues of a good dataset collected for a model training to be label quality, diversity, and completeness. We compare the effects of those virtues on the model performance using open medical CT datasets and conclude, that quality is more important than diversity early during labeling; the diversity, in turn, is more important than completeness. Based on this conclusion and additional experiments, we propose a labeling procedure for the segmentation of tomographic images to minimize efforts spent on labeling while maximizing the model performance.
Human review
How the literature uses it
What this paper does with the arXiv papers it cites
What it cites · 3 sentences without a confident label
Uses the method or tool of 2 papers
-
Therefore, we always train the same model (UNet [ref] with ResNet-18 [ref] as the backbone), with the same optimizer (Adam [ref] with $3e-4$ learning rate), for the same amount of epochs (100 epochs and 10 epochs long cooldown of the early stopping).
U-Net: Convolutional Networks for Biomedical Image Segmentation · Datasets Preparation & Model Training · 0.98
-
Therefore, we always train the same model (UNet [ref] with ResNet-18 [ref] as the backbone), with the same optimizer (Adam [ref] with $3e-4$ learning rate), for the same amount of epochs (100 epochs and 10 epochs long cooldown of the early stopping).
Deep Residual Learning for Image Recognition · Datasets Preparation & Model Training · 0.98
Cites as background 3 papers
-
Multiple vectors of research are focused on the question, of how to minimize the amount of labeled data required for training: transfer learning [ref] , self-supervised learning [ref] , active learning [ref] , etc.
Self-supervised Visual Feature Learning with Deep Neural Networks: A Survey · Introduction · 0.99
-
Multiple vectors of research are focused on the question, of how to minimize the amount of labeled data required for training: transfer learning [ref] , self-supervised learning [ref] , active learning [ref] , etc.
A Comprehensive Survey on Transfer Learning · Introduction · 0.99
-
Multiple vectors of research are focused on the question, of how to minimize the amount of labeled data required for training: transfer learning [ref] , self-supervised learning [ref] , active learning [ref] , etc.
A Survey of Deep Active Learning · Introduction · 0.99
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.