2023
DOI: 10.3390/diagnostics13030417
|View full text |Cite
|
Sign up to set email alerts
|

Using Ultrasound Image Augmentation and Ensemble Predictions to Prevent Machine-Learning Model Overfitting

Abstract: Deep learning predictive models have the potential to simplify and automate medical imaging diagnostics by lowering the skill threshold for image interpretation. However, this requires predictive models that are generalized to handle subject variability as seen clinically. Here, we highlight methods to improve test accuracy of an image classifier model for shrapnel identification using tissue phantom image sets. Using a previously developed image classifier neural network—termed ShrapML—blind test accuracy was… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
5
0

Year Published

2023
2023
2024
2024

Publication Types

Select...
9
1

Relationship

2
8

Authors

Journals

citations
Cited by 12 publications
(5 citation statements)
references
References 44 publications
0
5
0
Order By: Relevance
“…This will be critical to evaluate prior to real-time implementation, as subject variability is clinically important, and a deployable model must be generalized to handle this variability. Additional data augmentation or ensemble prediction approaches may be needed to make objection models more generalized going forward [45]. Third, a limited number of object detection models were evaluated in this study, and as a result, a more optimal model may exist.…”
Section: Discussionmentioning
confidence: 99%
“…This will be critical to evaluate prior to real-time implementation, as subject variability is clinically important, and a deployable model must be generalized to handle this variability. Additional data augmentation or ensemble prediction approaches may be needed to make objection models more generalized going forward [45]. Third, a limited number of object detection models were evaluated in this study, and as a result, a more optimal model may exist.…”
Section: Discussionmentioning
confidence: 99%
“…A widely used approach is to augment image inputs so that AI models less easily focus on image artifacts not associated with the injury. For this effort, we used rotation, flip, zoom, translation augmentations similar to approaches successfully used in other AI US imaging efforts ( Hussain et al, 2017 ; Xu et al, 2022 ; Snider et al, 2023 ). Another approach taken was to include validation patience during training so that training ceased if validation loss did not decrease for five training epochs.…”
Section: Discussionmentioning
confidence: 99%
“…Furthermore, the performance of the model can be assessed for each sample. The LOO method is recommended and is quite commonly used in the case of small datasets (from a dozen to several dozen samples) [59][60][61][62][63].…”
Section: Data Analysis Methods and Methodologymentioning
confidence: 99%