What does this research mean for the field?

The Hybrid Shape-Aware Network (Hy-SAN) effectively removes artifacts and preserves structural details in sparse-view X-ray Computed Tomography of Through-Silicon Vias, achieving state-of-the-art reconstruction performance and faster inference than existing deep learning methods. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To develop an efficient deep learning refinement framework that removes artifacts and preserves structural details during sparse-view X-ray computed tomography inspection of through-silicon vias.

September 8, 2026Nondestructive Testing And Evaluation

Hy-SAN: a hybrid shape-aware network for artifact-suppressed sparse-view computed tomography of through-silicon vias

View Full Paper

Ask AI

Bookmark

Authors

SCShanglei ChaiShenzhen University ZZZhiyuan ZhangSingapore Management University DLDajiang LuShenzhen University

Discussion

Member takes

Implication

Validation study demonstrates artifact-suppressed sparse-view computed tomography reconstruction in through-silicon vias, enabling rapid and reliable semiconductor defect detection.

Key Points

To develop an efficient deep learning refinement framework that removes artifacts and preserves structural details during sparse-view X-ray computed tomography inspection of through-silicon vias.
Constructed a hybrid refinement framework combining an initial analytical reconstruction with the Hybrid Shape-Aware Network (Hy-SAN).
Integrated a Shape-Aware Attention module, a Hybrid Lightweight Block, and a Multi-scale Prior Injection Module to incorporate morphological shape priors into the refinement process.
Trained and evaluated the compact model against domain-agnostic deep learning reconstruction methods on a through-silicon via X-ray computed tomography dataset.
Hy-SAN achieved state-of-the-art reconstruction fidelity, yielding an average peak signal-to-noise ratio (PSNR) of 39.18 dB on the through-silicon via dataset.
The network maintained a compact footprint of 6.10 million parameters and delivered the fastest inference speed among all evaluated deep learning models.

Cite This Study

Chai et al. (2026) studied this question.

synapsesocial.com/papers/6a9fd74958e84d0ff5b45d49 https://doi.org/10.1080/10589759.2026.2728079

View Full Paper

Ask AI

Bookmark

September 8, 2026Nondestructive Testing And Evaluation

Hy-SAN: a hybrid shape-aware network for artifact-suppressed sparse-view computed tomography of through-silicon vias

View Full Paper

Ask AI

Bookmark

Authors

SCShanglei ChaiShenzhen University ZZZhiyuan ZhangSingapore Management University DLDajiang LuShenzhen University

Discussion

Member takes

Implication

Validation study demonstrates artifact-suppressed sparse-view computed tomography reconstruction in through-silicon vias, enabling rapid and reliable semiconductor defect detection.

Key Points

To develop an efficient deep learning refinement framework that removes artifacts and preserves structural details during sparse-view X-ray computed tomography inspection of through-silicon vias.
Constructed a hybrid refinement framework combining an initial analytical reconstruction with the Hybrid Shape-Aware Network (Hy-SAN).
Integrated a Shape-Aware Attention module, a Hybrid Lightweight Block, and a Multi-scale Prior Injection Module to incorporate morphological shape priors into the refinement process.
Trained and evaluated the compact model against domain-agnostic deep learning reconstruction methods on a through-silicon via X-ray computed tomography dataset.
Hy-SAN achieved state-of-the-art reconstruction fidelity, yielding an average peak signal-to-noise ratio (PSNR) of 39.18 dB on the through-silicon via dataset.
The network maintained a compact footprint of 6.10 million parameters and delivered the fastest inference speed among all evaluated deep learning models.

Cite This Study

Chai et al. (2026) studied this question.

synapsesocial.com/papers/6a9fd74958e84d0ff5b45d49 https://doi.org/10.1080/10589759.2026.2728079

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark