Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising

Yusuf Talha Başak; Mehmet Ozan Unal; Metin Ertas; Isa Yildirim

doi:10.1109/ISBI61048.2026.11515960

Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising

Yusuf Talha Başak
, Mehmet Ozan Unal
, Metin Ertas
, Isa Yildirim

Department of Electronics and Communication Engineering

Istanbul Technical University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

While Total Variation (TV) excels in noise reduction and edge preservation, its reliance on a scalar regularization parameter limits adaptivity. In this study, we present a Learnable Total Variation (LTV) framework coupling an unrolled TV solver with a LambdaNet that predicts a per-pixel regularization map. The proposed framework is trained end-to-end to optimize reconstruction and regularization jointly, yielding spatially adaptive smoothing. Experiments on the DeepLesion dataset, using realistic LoDoPaB-CT simulation, show consistent gains over classical TV and FBP+U-Net, achieving up to +3.7 dB PSNR and 8% relative SSIM improvement. LTV provides an interpretable alternative to black-box CNNs for low-dose CT denoising.

Original language	English
Title of host publication	ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging
Publisher	IEEE Computer Society
ISBN (Electronic)	9798331577636
DOIs	https://doi.org/10.1109/ISBI61048.2026.11515960
Publication status	Published - 2026
Event	23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026 - London, United Kingdom Duration: 8 Apr 2026 → 11 Apr 2026

Publication series

Name	Proceedings - International Symposium on Biomedical Imaging
Volume	2026-April
ISSN (Print)	1945-7928
ISSN (Electronic)	1945-8452

Conference

Conference	23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026
Country/Territory	United Kingdom
City	London
Period	8/04/26 → 11/04/26

Bibliographical note

Publisher Copyright:
© 2026 IEEE.

Keywords

Denoising
Lambda Mapping
Low-Dose CT
Total Variation
Unrolled Optimization

Access to Document

10.1109/ISBI61048.2026.11515960

Cite this

@inproceedings{8ffdebcd1d0a4b0f92c2f6baf992e18a,

title = "Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising",

abstract = "While Total Variation (TV) excels in noise reduction and edge preservation, its reliance on a scalar regularization parameter limits adaptivity. In this study, we present a Learnable Total Variation (LTV) framework coupling an unrolled TV solver with a LambdaNet that predicts a per-pixel regularization map. The proposed framework is trained end-to-end to optimize reconstruction and regularization jointly, yielding spatially adaptive smoothing. Experiments on the DeepLesion dataset, using realistic LoDoPaB-CT simulation, show consistent gains over classical TV and FBP+U-Net, achieving up to +3.7 dB PSNR and 8\% relative SSIM improvement. LTV provides an interpretable alternative to black-box CNNs for low-dose CT denoising.",

keywords = "Denoising, Lambda Mapping, Low-Dose CT, Total Variation, Unrolled Optimization",

author = "Ba{\c s}ak, \{Yusuf Talha\} and Unal, \{Mehmet Ozan\} and Metin Ertas and Isa Yildirim",

note = "Publisher Copyright: {\textcopyright} 2026 IEEE.; 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026 ; Conference date: 08-04-2026 Through 11-04-2026",

year = "2026",

doi = "10.1109/ISBI61048.2026.11515960",

language = "English",

series = "Proceedings - International Symposium on Biomedical Imaging",

publisher = "IEEE Computer Society",

booktitle = "ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging",

address = "United States",

}

Başak, YT, Unal, MO, Ertas, M & Yildirim, I 2026, Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising. in ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging. Proceedings - International Symposium on Biomedical Imaging, vol. 2026-April, IEEE Computer Society, 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026, London, United Kingdom, 8/04/26. https://doi.org/10.1109/ISBI61048.2026.11515960

Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising. / Başak, Yusuf Talha; Unal, Mehmet Ozan; Ertas, Metin et al.
ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging. IEEE Computer Society, 2026. (Proceedings - International Symposium on Biomedical Imaging; Vol. 2026-April).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising

AU - Başak, Yusuf Talha

AU - Unal, Mehmet Ozan

AU - Ertas, Metin

AU - Yildirim, Isa

PY - 2026

Y1 - 2026

N2 - While Total Variation (TV) excels in noise reduction and edge preservation, its reliance on a scalar regularization parameter limits adaptivity. In this study, we present a Learnable Total Variation (LTV) framework coupling an unrolled TV solver with a LambdaNet that predicts a per-pixel regularization map. The proposed framework is trained end-to-end to optimize reconstruction and regularization jointly, yielding spatially adaptive smoothing. Experiments on the DeepLesion dataset, using realistic LoDoPaB-CT simulation, show consistent gains over classical TV and FBP+U-Net, achieving up to +3.7 dB PSNR and 8% relative SSIM improvement. LTV provides an interpretable alternative to black-box CNNs for low-dose CT denoising.

AB - While Total Variation (TV) excels in noise reduction and edge preservation, its reliance on a scalar regularization parameter limits adaptivity. In this study, we present a Learnable Total Variation (LTV) framework coupling an unrolled TV solver with a LambdaNet that predicts a per-pixel regularization map. The proposed framework is trained end-to-end to optimize reconstruction and regularization jointly, yielding spatially adaptive smoothing. Experiments on the DeepLesion dataset, using realistic LoDoPaB-CT simulation, show consistent gains over classical TV and FBP+U-Net, achieving up to +3.7 dB PSNR and 8% relative SSIM improvement. LTV provides an interpretable alternative to black-box CNNs for low-dose CT denoising.

KW - Denoising

KW - Lambda Mapping

KW - Low-Dose CT

KW - Total Variation

KW - Unrolled Optimization

UR - https://www.scopus.com/pages/publications/105041609106

U2 - 10.1109/ISBI61048.2026.11515960

DO - 10.1109/ISBI61048.2026.11515960

M3 - Conference contribution

AN - SCOPUS:105041609106

T3 - Proceedings - International Symposium on Biomedical Imaging

BT - ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging

PB - IEEE Computer Society

T2 - 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026

Y2 - 8 April 2026 through 11 April 2026

ER -

Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this