SPASHT: An image-enhancement method for sparse-view MPI SPECT

Abhinav K. Jha; Barry A. Siegel; Janice Tania; Nuri Choi; Wenxuan Xue; Zezhang Yang; Zitong Yu

arxiv: 2511.06203 · v2 · submitted 2025-11-09 · 📡 eess.IV

SPASHT: An image-enhancement method for sparse-view MPI SPECT

Zezhang Yang , Zitong Yu , Nuri Choi , Janice Tania , Wenxuan Xue , Barry A. Siegel , Abhinav K. Jha This is my paper

Pith reviewed 2026-05-18 00:29 UTC · model grok-4.3

classification 📡 eess.IV

keywords MPI SPECTsparse-view imagingimage enhancementperfusion defect detectionAUChuman observer studySPASHT

0 comments

The pith

SPASHT improves AUC for perfusion defect detection in sparse-view MPI SPECT

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces SPASHT as an image-enhancement method for MPI SPECT scans that use fewer projection views to shorten total scan time. MPI SPECT diagnoses coronary artery disease by detecting perfusion defects, but fewer views normally increase noise and sampling artifacts that hurt detection. SPASHT is trained specifically to boost performance on this defect-detection task rather than generic image quality. In a retrospective study on patient data with synthetically inserted defects, SPASHT produced significantly higher AUC scores than standard sparse-view reconstruction at one-half, one-third, and one-sixth the usual number of views. A separate human observer study confirmed better defect detection with the enhanced images.

Core claim

SPASHT yields significantly improved AUC compared to the sparse-view protocol for detecting perfusion defects in MPI SPECT for all tested reductions in projection views, including 1/6, 1/3, and 1/2 of typical counts, with supporting evidence from a human observer study showing improved detection performance.

What carries the argument

SPASHT, the sparse-view SPECT image-enhancement method trained to improve performance specifically on defect-detection tasks.

If this is right

Shorter MPI SPECT scan times become feasible while preserving diagnostic accuracy for coronary artery disease.
Reduced patient discomfort and lower chance of motion artifacts during the procedure.
Reliable defect detection performance holds across multiple levels of view reduction from half down to one-sixth.
Evidence supports moving accelerated sparse-view protocols closer to routine clinical use.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Prospective trials with real defects could test whether the reported gains survive without synthetic insertion.
The training approach focused on the detection task might transfer to other sparse-sampling problems in nuclear medicine.
Combining SPASHT with dose-reduction methods could further lower patient radiation exposure.

Load-bearing premise

Synthetically inserted perfusion defects in retrospective patient data are clinically realistic and performance gains will hold in prospective studies with naturally occurring defects.

What would settle it

A prospective clinical study with patients having naturally occurring perfusion defects that finds no statistically significant AUC improvement or human-observer benefit from SPASHT over standard sparse-view reconstruction.

read the original abstract

Single-photon emission computed tomography for myocardial perfusion imaging (MPI SPECT) is a widely used diagnostic tool for coronary artery disease. However, the procedure requires considerable scanning time, leading to patient discomfort and the potential for motion-induced artifacts. Reducing the number of projection views while keeping the time per view unchanged provides a mechanism to shorten the scanning time. However, this approach leads to increased sampling artifacts, higher noise, and hence limited image quality. To address these issues, we propose sparseview SPECT image enhancement (SPASHT), inherently training the algorithm to improve performance on defect-detection tasks. We objectively evaluated SPASHT on the clinical task of detecting perfusion defects in a retrospective clinical study using data from patients who underwent MPI SPECT, where the defects were clinically realistic and synthetically inserted. The study was conducted for different numbers of fewer projection views, including 1/6, 1/3, and 1/2 of the typical projection views for MPI SPECT. Performance on the detection task was quantified using area under the receiver operating characteristic curve (AUC). Images obtained with SPASHT yielded significantly improved AUC compared to those obtained with the sparse-view protocol for all the considered numbers of fewer projection views. To further assess performance, a human observer study on the task of detecting perfusion defects was conducted. Results from the human observer study showed improved detection performance with images reconstructed using SPASHT compared to those from the sparse-view protocol. The results provide evidence of the efficacy of SPASHT in improving the quality of sparse-view MPI SPECT images and motivate further clinical validation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

SPASHT claims task-specific enhancement for sparse-view MPI SPECT with better defect detection AUC, but the abstract gives almost no method details or numbers so the claims are hard to assess.

read the letter

The main point is that this paper introduces SPASHT as an enhancement method trained specifically for the perfusion defect detection task in reduced-view MPI SPECT, and the abstract reports improved AUC plus better human observer performance on retrospective data for 1/6, 1/3, and 1/2 views compared to plain sparse-view reconstruction. That focus on the clinical task instead of just PSNR or visual quality is a reasonable choice for this area. Using real patient data and including a human study shows they are thinking about diagnostic utility rather than isolated image metrics. The work targets a practical issue in cardiac SPECT where shorter scans would help with patient comfort and motion. What is new is the named method and the claim of inherent task-specific training, though without the full text it is impossible to see how that training actually differs from standard supervised enhancement. The soft spots are clear from the abstract alone. There is no description of the architecture, loss function, or training procedure, no quantitative AUC values or confidence intervals, and no statistical test details. The evaluation rests on synthetically inserted defects in retrospective scans; the abstract calls them clinically realistic but supplies no protocol or realism check, which leaves open whether the gains would appear with naturally occurring defects in prospective data. Without those pieces the central claim is difficult to verify. This paper is for researchers working on AI reconstruction and enhancement for nuclear cardiology. A reader already familiar with sparse-view SPECT might pick up the task-driven angle, but most would need the full methods and results to decide on usefulness. It deserves peer review so referees can examine the actual implementation and data rather than desk-rejecting on the thin abstract.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces SPASHT, an image-enhancement method for sparse-view MPI SPECT to reduce scan time while improving image quality for perfusion defect detection. It evaluates the approach on retrospective patient data with synthetically inserted clinically realistic defects for 1/6, 1/3, and 1/2 of typical projection views, claiming significantly improved AUC over the sparse-view protocol, supported by a human observer study showing better detection performance.

Significance. If the reported gains are confirmed with quantitative detail and prospective validation, SPASHT could meaningfully shorten MPI SPECT procedures, reducing patient discomfort and motion artifacts while enhancing diagnostic accuracy for coronary artery disease. The emphasis on task-specific training for defect detection represents a constructive design choice.

major comments (3)

[Abstract] Abstract: The claim that 'Images obtained with SPASHT yielded significantly improved AUC' provides no numerical AUC values, confidence intervals, or statistical test results (e.g., p-values) for the 1/6, 1/3, and 1/2 view reductions. This information is required to assess the magnitude and reliability of the central performance claim.
[Abstract] Abstract: No description is given of the SPASHT method, including its architecture, training procedure, or how it is 'inherently training the algorithm to improve performance on defect-detection tasks.' This detail is load-bearing for evaluating the technical novelty and reproducibility.
[Abstract] Abstract: The evaluation rests on synthetically inserted perfusion defects stated as 'clinically realistic,' yet supplies no insertion protocol, realism validation against natural defects, or discussion of limitations for prospective studies. This assumption underpins the clinical relevance of both the AUC and human observer results.

minor comments (1)

[Abstract] The abstract would be strengthened by briefly noting study limitations or the need for prospective validation to frame the findings appropriately.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and will revise the abstract to improve clarity and completeness while preserving its concise nature.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that 'Images obtained with SPASHT yielded significantly improved AUC' provides no numerical AUC values, confidence intervals, or statistical test results (e.g., p-values) for the 1/6, 1/3, and 1/2 view reductions. This information is required to assess the magnitude and reliability of the central performance claim.

Authors: We agree that numerical details would strengthen the abstract. We will revise the abstract to report the specific AUC values, confidence intervals, and p-values for each of the 1/6, 1/3, and 1/2 view reductions. revision: yes
Referee: [Abstract] Abstract: No description is given of the SPASHT method, including its architecture, training procedure, or how it is 'inherently training the algorithm to improve performance on defect-detection tasks.' This detail is load-bearing for evaluating the technical novelty and reproducibility.

Authors: The abstract is space-constrained, but we acknowledge the need for additional context. We will expand the abstract with a brief description of the SPASHT architecture, training procedure, and its task-specific optimization for defect detection, while retaining full technical details in the Methods section. revision: yes
Referee: [Abstract] Abstract: The evaluation rests on synthetically inserted perfusion defects stated as 'clinically realistic,' yet supplies no insertion protocol, realism validation against natural defects, or discussion of limitations for prospective studies. This assumption underpins the clinical relevance of both the AUC and human observer results.

Authors: We will add a concise statement to the abstract summarizing the defect insertion protocol and its basis in clinical realism. Detailed protocol, validation steps, and limitations for prospective studies are covered in the Methods and Discussion sections of the manuscript; we will ensure the abstract references these to better support the evaluation's relevance. revision: partial

Circularity Check

0 steps flagged

No circularity in method proposal or evaluation

full rationale

The available abstract describes SPASHT as a proposed image-enhancement method trained to improve defect-detection performance, then reports empirical results on retrospective patient data using the independent AUC metric for perfusion-defect detection plus a separate human observer study. No equations, derivations, fitted parameters renamed as predictions, or self-citations appear in the text. The central claims rest on external evaluation benchmarks rather than reducing to inputs by construction, satisfying the criteria for a self-contained result with no circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only abstract available, so limited visibility into internals. The approach likely relies on learned parameters from training data and the assumption that synthetic defects match real clinical ones.

axioms (1)

domain assumption Synthetically inserted defects are clinically realistic
The retrospective study used patient data with synthetically inserted defects described as clinically realistic.

pith-pipeline@v0.9.0 · 5579 in / 1292 out tokens · 69875 ms · 2026-05-18T00:29:28.422888+00:00 · methodology

SPASHT: An image-enhancement method for sparse-view MPI SPECT

Core claim

What carries the argument

If this is right

Where Pith is reading between the lines

Load-bearing premise

What would settle it

discussion (0)