QICOM: Open-Source Software for Reproducible Image Quality Evaluation, Blinded Assessment, and Radiomic Comparison

Fabio Mattiussi; Switinder Ghotra

eess.IVOct 22, 2026

QICOM: Open-Source Software for Reproducible Image Quality Evaluation, Blinded Assessment, and Radiomic Comparison

Authors: Fabio Mattiussi, Switinder Ghotra

Organizations: Integrated Diagnostics Institute of Southern Switzerland (IDISI), Ente Ospedaliero Cantonale (EOC), 6900 Lugano, Switzerland · School of Health Sciences (HESAV), University of Applied Sciences and Arts Western Switzerland (HES-SO), 1011 Lausanne, Switzerland · Department of Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland

Abstract

QICOM (Quality Image COMparator) is open-source software for quantitative image comparison, paired radiomics, and blinded assessment. It saves processing settings, image identities, measurements, and response mappings so a comparison can be reopened and recalculated. We evaluate reference agreement, responses to controlled perturbations, and recovery of saved analyses. Synthetic experiments cover 37 quantitative descriptors using independentnumerical references and separate upstream integration comparisons. The QICOM/MIRP backend meets the reference tolerances for all 482 numeric IBSI 1 digital-phantom entries and all 1,622 entries across configurations A-E of the public CT reference case; these comprise 1,804 feature/aggregation comparisons and 300 preprocessing diagnostics. Restricting masked SSIM and HFEN to complete kernel support removes the observed dependence on excluded pixels. In a synthetic filtering example, stronger smoothing increases SSIM while worsening MSE and HFEN relative to milder smoothing and changing GLCM contrast more strongly. Reopened radiomics projects reproduce all 43 feature pairs per comparison, and two blinded deliveries return four correctly mapped scripted responses. These experiments document the calculations and data transfers linking the three workflows.

Figures & tables

Figure 1: QICOM’s three workbenches, their inputs and outputs, and the numerical and simulated-response experiments used in this study.

Figure 2: The quantitative workbench captured from the running application with the central-plane previews of the synthetic filtering inputs. The synchronized image views show the reference and filtered image. Additional workbench captures appear in Supplementary Section A.5 .

IBSI 1 configuration	Features	Diagnostics	Passed / total
Digital phantom	482	0	482 / 482
CT A	346	60	406 / 406
CT B	346	60	406 / 406
CT C	210	60	270 / 270
CT D	210	60	270 / 270
CT E	210	60	270 / 270

Table 1: Reference agreement through the QICOM/MIRP backend. The five CT configurations use one public case; entries count feature/aggregation combinations and preprocessing diagnostics.

Figure 3: Raw absolute relative errors against IBSI reference values. Each point represents one worksheet row with a nonzero reference; zero residuals are placed at the left plotting floor. Acceptance uses the row-specific absolute tolerance and truncation rule in the Methods.

Figure 4: Public IBSI CT reference case in NIfTI format. Axial plane through the largest mask cross-section, displayed at $[-1000,400]$ HU; the white contour is the supplied mask. The inset enlarges the same ROI. Display adapted from the CT image and mask supplied by Lambin, CancerData (2016) [ 18 , 17 ] . The CT-derived image and contour retain the source license, CC BY-NC 3.0 .

Figure 5: MSE and SSIM responses to noise, Gaussian blur, and intensity offsets on the synthetic phantom. Points are computed values; lines connect successive levels.

Spectrum / task	$d^{\prime}_{PW}$	$d^{\prime}_{NPW}$	$d^{\prime}_{NPWE}$
flat	1.500000	1.500000	1.397820
colored flat task	1.517647	1.279204	1.142984
colored gaussian task	0.987565	0.767542	0.507999

Table 2: Observer outputs for synthetic spectrum and task fixtures.

Missing probability	Coverage	MC standard error	Mean CI width
0.00	0.95	0.0218	0.943
0.15	0.95	0.0218	1.016

Table 3: Crossed reader–case simulations: 100 generated datasets per setting. The 95% Wilson interval for coverage is approximately 0.888–0.978 in both settings.

Condition	MSE $\downarrow$	SSIM $\uparrow$	HFEN $\downarrow$	GLCM contrast
Noisy	63.001	0.6198	0.5932	3.4814
$\sigma=0.65$	15.849	0.8724	0.5054	1.3488
$\sigma=1.5$	18.421	0.9119	0.6998	0.6356

Table 4: Quantitative and radiomic results for the filtering example. Clean-reference GLCM contrast is 1.4411. Arrows indicate decreasing error or increasing similarity. MSE is in squared arbitrary intensity units; the other displayed measures are dimensionless.

Figure 6: Central plane of the clean, noisy, and filtered volumes. The common display range is 0–255; the white contour marks the radiomic ROI.

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Class	Absolute $a$	Relative $r$
float32	1e-06	1e-05
float64	1e-10	1e-10
hfen	1e-05	1e-04
integration	2e-06	2e-05
nss	1e-07	1e-07
ssim	2e-05	2e-05

Appendix

Table 5: Numerical acceptance tolerances for the convention-matched quantitative references.

Figure 7: Representative synthetic inputs generated by code, with known intensity and spatial transformations. Display normalization in this montage is only for visualization; computations use the saved numerical arrays.

CT	Resegmentation (HU)	Texture bins	IVH bins
A	$[-500,400]$	Width 25 HU	None
B	$[-500,400]$	Count 32	None
C	$[-1000,400]$	Width 25 HU	Width 2.5 HU
D	$\pm 3$ SD	Count 32	None
E	$[-1000,400]$ and $\pm 3$ SD	Count 32	Count 1000

Appendix

Table 6: IBSI CT preprocessing configurations. IVH: intensity–volume histogram. A–B use the specified planar aggregations; C–E use volumetric aggregations.

Computation package	Installed version
lpips	0.1.4
nibabel	5.4.2
numpy	2.5.2
piq	0.8.0
scikit-image	0.26.0
scipy	1.18.1

Appendix

Table 7: Principal computation dependencies. The complete environment manifests include additional packages and learned-weight identities.

Package	Role in QICOM
Qt for Python / PySide6 [ 70 ]	Desktop widgets, rendering, and worker orchestration
PyQtGraph [ 71 ]	Interactive image views, plots, and ROI interaction
Pillow [ 72 ]	Raster-image reading, writing, and blinded-delivery metadata handling
dicom-anonymizer [ 73 ]	DICOM de-identification support within QICOM’s additional checks
openpyxl [ 74 ]	Spreadsheet report generation
PyYAML [ 75 ]	Metric descriptors and configuration serialization

Appendix

Table 15

Medical Image Quality Assessment

See all 23 papers

RIQE: a NIQE-style reference model for Computed Tomography
Sep 30, 2026Fabio MattiussiImage Quality Assessment Computed Tomography
EOC
RadMatch: Auditable Radiology Report Evaluation via Finding-Level Matching
Sep 1, 2026Charles Corbière, Léo Machado, Aubin Charley +3LLM-as-a-Judge Medical Image Benchmarks
Raidium, Paris, France · Department of Radiology, Fondation Ophtalmologique Adolphe de Rothschild, Paris, France · Service de Radiologie Interventionnelle, Département d’Anesthésie, Chirurgie et Imagerie Interventionnelle (DACI), Gustave Roussy, Villejuif, France
CBCT-IQ: A Publicly Available Annotated Cone-Beam CT Dataset for Image Quality Assessment and Benchmarking
Jul 31, 2026Sepideh Hatamikia, Anna Breger, Clemens Karner +14Cone-Beam CT Image Quality Assessment
Austrian Center for Medical Innovation and Technology (ACMIT), Wiener Neustadt, Austria · Department of Medicine, Faculty of Medicine and Dentistry, Danube Private University, Krems, Austria · Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, Vienna, Austria +8

Sep 30, 2026J/K move · Enter open · S save

RIQE: a NIQE-style reference model for Computed Tomography

Fabio Mattiussi

EOC

Image Quality Assessment Computed Tomography LDCT Denoising Medical Image Quality Assessment

Open paper AskPDF arXiv

The Natural Image Quality Evaluator (NIQE) scores an image by its statistical distance from a model fitted on pristine images, and its distributed model is fitted on photographs. We release the Radiology Image Quality Evaluator (RIQE), a NIQE-style model fitted on 3,792 full-dose slices from 158 patients of the public LDCT-and-Projection-data collection, with a declared intensity mapping, a manifest of every slice and a script that reproduces the fit. On 40 held-out patients, RIQE ranks reduced-dose reconstructions, simulated by projection-domain noise insertion, worse than the full-dose reconstruction of the same slice in 240 of 240 chest and 230 of 240 abdominal pairs, and ranks images with 20% more noise worse than their source in 97.5-100% of cases. Its preferences among filtered images, however, do not follow lesion signal. With a 4 mm, +10 HU lesion inserted in noisy abdominal slices, RIQE prefers bilateral filtering to the unfiltered image in every image up to a 32 HU residual, at which 29% of the lesion's matched-filter signal remains and its detectability index falls from 0.51 to 0.33; it never prefers Gaussian smoothing, which at the same 32 HU residual leaves 70% of the signal and a detectability index of 0.48. Fitted on photographs with the parameters published for NIQE, the same code ranks every simulated reduced-dose abdominal image better than its full-dose counterpart. RIQE is suited to ranking a degraded image against its source; under the conditions tested it should not be the sole criterion for selecting, comparing or tuning denoisers.

Abstract

Medical Image Quality Assessment

Explore similar work

RIQE: a NIQE-style reference model for Computed Tomography

RadMatch: Auditable Radiology Report Evaluation via Finding-Level Matching

CBCT-IQ: A Publicly Available Annotated Cone-Beam CT Dataset for Image Quality Assessment and Benchmarking