Task-Driven CT Image Quality Optimization for Low-Contrast Lesion Detectability with Tunable Neural Networks
Low-contrast lesions are difficult to detect in noisy low-dose CT images. Improving CT image quality for this detection task has the potential to improve diagnostic accuracy and patient outcomes. In this work, we use tunable neural networks for CT image restoration with a hyperparameter to control the variance/bias tradeoff. We use clinical images from a super-high-resolution normal-dose CT scan to synthesize low-contrast low-dose CT images for supervised training of deep learning CT reconstruction models. Those models are trained using with multiple noise realizations so that variance and bias can be penalized separately. We use a training loss function with one hyperparameter called the denoising level, which controls the variance/bias tradeoff. Finally, we evaluate the CT image quality to find the optimal denoising level for low-contrast lesion detectability. We evaluate performance using a shallow neural network model classification model to represent a suboptimal image observer. Our results indicate that the optimal networks for low-contrast lesion detectability are those that prioritize bias reduction rather than mean-squared error, which demonstrates the potential clinical benefit of our proposed tunable neural networks.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2023 |
---|---|
Erschienen: |
2023 |
Enthalten in: |
Zur Gesamtaufnahme - volume:12463 |
---|---|
Enthalten in: |
Proceedings of SPIE--the International Society for Optical Engineering - 12463(2023) vom: 31. Feb. |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Tivnan, Matthew [VerfasserIn] |
---|
Links: |
---|
Themen: |
---|
Anmerkungen: |
Date Revised 09.01.2024 published: Print-Electronic Citation Status PubMed-not-MEDLINE |
---|
doi: |
10.1117/12.2653936 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM366788027 |
---|
LEADER | 01000naa a22002652 4500 | ||
---|---|---|---|
001 | NLM366788027 | ||
003 | DE-627 | ||
005 | 20240114233153.0 | ||
007 | cr uuu---uuuuu | ||
008 | 240114s2023 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1117/12.2653936 |2 doi | |
028 | 5 | 2 | |a pubmed24n1253.xml |
035 | |a (DE-627)NLM366788027 | ||
035 | |a (NLM)38188182 | ||
035 | |a (PII)124631M | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Tivnan, Matthew |e verfasserin |4 aut | |
245 | 1 | 0 | |a Task-Driven CT Image Quality Optimization for Low-Contrast Lesion Detectability with Tunable Neural Networks |
264 | 1 | |c 2023 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ƒaComputermedien |b c |2 rdamedia | ||
338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
500 | |a Date Revised 09.01.2024 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status PubMed-not-MEDLINE | ||
520 | |a Low-contrast lesions are difficult to detect in noisy low-dose CT images. Improving CT image quality for this detection task has the potential to improve diagnostic accuracy and patient outcomes. In this work, we use tunable neural networks for CT image restoration with a hyperparameter to control the variance/bias tradeoff. We use clinical images from a super-high-resolution normal-dose CT scan to synthesize low-contrast low-dose CT images for supervised training of deep learning CT reconstruction models. Those models are trained using with multiple noise realizations so that variance and bias can be penalized separately. We use a training loss function with one hyperparameter called the denoising level, which controls the variance/bias tradeoff. Finally, we evaluate the CT image quality to find the optimal denoising level for low-contrast lesion detectability. We evaluate performance using a shallow neural network model classification model to represent a suboptimal image observer. Our results indicate that the optimal networks for low-contrast lesion detectability are those that prioritize bias reduction rather than mean-squared error, which demonstrates the potential clinical benefit of our proposed tunable neural networks | ||
650 | 4 | |a Journal Article | |
700 | 1 | |a Lee, Tzu-Cheng |e verfasserin |4 aut | |
700 | 1 | |a Zhang, Ruoqiao |e verfasserin |4 aut | |
700 | 1 | |a Boedeker, Kirsten |e verfasserin |4 aut | |
700 | 1 | |a Cai, Liang |e verfasserin |4 aut | |
700 | 1 | |a Sulam, Jeremias |e verfasserin |4 aut | |
700 | 1 | |a Stayman, J Webster |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t Proceedings of SPIE--the International Society for Optical Engineering |d 1984 |g 12463(2023) vom: 31. Feb. |w (DE-627)NLM166567531 |x 0277-786X |7 nnns |
773 | 1 | 8 | |g volume:12463 |g year:2023 |g day:31 |g month:02 |
856 | 4 | 0 | |u http://dx.doi.org/10.1117/12.2653936 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 12463 |j 2023 |b 31 |c 02 |