BFHD: Bidirectional Feature Harmonization Decomposition for Heterogeneous Clinical Assessments

Yuanhao Zhuo; Zixi Qin; Ling Qin; Wanqing Li

cs.LGOct 14, 2026

BFHD: Bidirectional Feature Harmonization Decomposition for Heterogeneous Clinical Assessments

Authors: Yuanhao Zhuo, Zixi Qin, Ling Qin, Wanqing Li

Organizations: University of Wollongong, Wollongong, NSW 2522, Australia · The People’s Hospital of Guangxi Zhuang Autonomous Region, Nanning, Guangxi 530021, China

Abstract

Clinical assessments are often collected using heterogeneous assessment systems across centers and time, leading to records that mix different but related sets of measurements. This motivates harmonization beyond total-score linking. We formulate clinical harmonization at the measurement level as a bidirectional recoverability problem: given paired observations from two assessment systems, the goal is to identify which measurements can be reliably translated in both directions within an application-defined tolerance, while separating non-translatable components. We propose Bidirectional Feature Harmonization Decomposition (BFHD), a feasibility-driven framework that enforces bidirectionally coupled translation and uses feature-wise output gating to produce an explicit decomposition in the original measurement space. Experiments on synthetic data and real clinical assessment pairs show that BFHD achieves broader feasible harmonization coverage and improved subset stability compared to baselines.

Figures & tables

Model	Acc X	Acc Y	$J_{X}$	$J_{Y}$
Ours	$0.880_{\pm 0.024}$	$0.830_{\pm 0.024}$	0.933	0.924
CMLP	$0.855_{\pm 0.015}$	$0.810_{\pm 0.049}$	0.975	0.789
NICE	$0.735_{\pm 0.039}$	$0.725_{\pm 0.040}$	0.837	0.814

Table 1: Synthetic feasibility recovery under ground-truth

\varepsilon

(

s_{gen}

=5). Acc X and Acc Y denote index-level structure recovery accuracy for systems

X

and

Y

, respectively (mean ±std across 10 seeds).

J_{X}

and

J_{Y}

report Jaccard similarity. For baseline models without gating, recoverability is inferred post hoc using a relaxed

2\varepsilon_{\text{gen}}

feasibility criterion. Additional results are provided in Appendix D.1 .

Data	Method	$\#Rec_{A}$	$J_{A}$	$\#A>0.7~$	$\#A>0.5~$	$\#Rec_{B}$	$J_{B}$	$\#B>0.7~$	$\#B>0.5~$
	IRT	11	0.733	4.5	10.5	7	0.785	4	7
CAS(A)	XGB	9	0.710	6	9	7	0.842	5	7
$\leftrightarrow$	FRF	9	0.683	4.5	9	7.5	0.858	6	7.5
SNAP(B)	CMLP	9	0.640	5	9	7	0.764	6	7
	NICE	11	0.867	4	11	9	0.774	5	9
	Ours	11	0.930	4	11	9	0.817	5	9

Table 2: Real-world harmonization results under

\varepsilon_{\text{cutoff}}=0.6

. For each row, A and B correspond to the two assessment systems.

\#\mathrm{Rec}_{A}

\#\mathrm{Rec}_{B}

are the median numbers of recoverable measurements and

J_{A}

J_{B}

are Jaccard stability across 10 seeds.

\#A{>}0.7

\#A{>}0.5

(and analogously for B) count selected measurements whose CCC exceeds 0.7 or 0.5, used as diagnostic measures of feasibility.

Figure 1: $\varepsilon$ -sweep on EMBIC and NeuroBAT. Median number of recoverable measurements ( $\#\mathrm{rec}$ ) as a function of the feasibility threshold $\varepsilon$ . CAS $\leftrightarrow$ SNAP results are reported in Appendix D.2 .

Method	#Ret A	Ret% A	#Ret B	Ret% B
IRT	3	$42.9_{\pm 0.0}$	0	N/A
XGB	6	$85.7_{\pm 0.0}$	0.5	$62.5_{\pm 48.4}$
FRF	5	$56.0_{\pm 8.0}$	0	$0.0_{\pm 0.0}$
CMLP	6	$71.0_{\pm 7.5}$	1	$90.0_{\pm 10.0}$
NICE	6	$77.4_{\pm 9.9}$	2	$49.2_{\pm 10.5}$
Ours	7	$100.0_{\pm 0.0}$	3	$100.0_{\pm 0.0}$

Table 3: Temporal deployment utility under protocol drift on the EMBIC(A)

\leftrightarrow

NeuroBAT(B) assessment pair.

\varepsilon_{\text{cutoff}}=0.6

. #Ret A /#Ret B denote the numbers of training-period recoverable measurements that remain feasible on the held-out test split. Ret% A /Ret% B report the corresponding retention rates. (mean ±std across 10 seeds) Additional results in Appendix D.3 .

Variant	$\#\mathrm{Rec}_{A}$	$J_{A}$	$\#\mathrm{Rec}_{B}$	$J_{B}$
Ours (full)	7	0.910	2	0.689
- TWMLP	7	0.886	1	0.600
- $L_{0}$ penalty	$0^{\dagger}$	N/A	$0^{\dagger}$	N/A
- No gating	7	0.806	2	0.523
- No hinge	$13^{\dagger}$	N/A	$9^{\dagger}$	N/A

Table 4: Ablation study on EMBIC (A)

\leftrightarrow

NeuroBAT (B) under

\varepsilon_{\text{cutoff}}=0.6

. Notation are consistent with Table 2 . Additional results in Appendix D.4 .

\dagger

indicate trivial gating solutions.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Deploy $X\rightarrow\hat{Y}$ (source-side: $X$ )
Method	Acc (Obs)	Acc (Trans)
X-only	$0.909\pm 0.047$	-
CMLP	$0.919\pm 0.024$	$0.881\pm 0.047$
NICE	$0.916\pm 0.023$	$0.897\pm 0.033$
Ours	$0.913\pm 0.023$	$0.901\pm 0.037$
Deploy $Y\rightarrow\hat{X}$ (source-side: $Y$ )

Appendix

Table 5: Downstream prediction under size-matched harmonized subsets. Results are reported as mean

\pm

standard deviation over ten random seeds. Obs denotes using observed measurements in the prediction space, while Trans denotes using translated measurements obtained from the opposite view. Single-view reference (X-only / Y-only) serve as reference result.

$p_{\text{miss}}$	Method	Missing in $X$ ( $X_{\text{miss}}\rightarrow\hat{Y}$ )			Missing in $Y$ ( $Y_{\text{miss}}\rightarrow\hat{X}$ )
		#Ret	Retain	Std	#Ret	Retain	Std
0.1	Ours	$9$	$0.860$	$0.097$	$7$	$0.813$	$0.140$
	NICE	$8.5$	$0.788$	$0.152$	$7$	$0.877$	$0.121$
	CMLP	$8$	$0.815$	$0.196$	$6$	$0.725$	$0.293$
0.2	Ours	$6.5$	$0.513$	$0.158$	$5.5$	$0.634$	$0.112$
	NICE	$6.5$	$0.524$	$0.203$	$6$	$0.632$	$0.168$

Appendix

Table 6: Robustness under MCAR missingness. For each method, #Ret denotes the median number of measurements that remain feasible under missingness. Retain denotes the retention rate relative to the reference feasible set defined at

p=0

(Experiment 2). Results are reported as mean

\pm

standard deviation across random seeds.

Data	Method	Acc X	Acc Y	$J_{X}$	$J_{Y}$	$J_{\mathrm{overall}}$	CCC X	CCC Y
	Ours	$0.915\pm 0.023$	$0.900\pm 0.000$	0.948	1.000	0.974	$0.738\pm 0.014$	$0.754\pm 0.063$
$s_{\mathrm{gen}}$ = 1	CMLP	$0.890\pm 0.037$	$0.900\pm 0.000$	0.896	1.000	0.948	$0.768\pm 0.019$	$0.758\pm 0.014$
	NICE	$0.850\pm 0.000$	$0.805\pm 0.047$	1.000	0.824	0.912	$0.722\pm 0.013$	$0.745\pm 0.013$
	Ours	$0.830\pm 0.056$	$0.755\pm 0.042$	0.783	0.812	0.797	$0.574\pm 0.033$	$0.619\pm 0.049$
$s_{\mathrm{gen}}$ = 2	CMLP	$0.810\pm 0.037$	$0.710\pm 0.030$	0.848	0.722	0.785	$0.612\pm 0.023$	$0.678\pm 0.024$
	NICE	$0.550\pm 0.032$	$0.605\pm 0.035$	0.400	0.544	0.472	$0.668\pm 0.043$	$0.648\pm 0.016$

Appendix

Table 7: Experiment 1 (Synthetic): feasibility structure recovery and stability across different data-generation seeds (

s_{\mathrm{gen}}\in\{1,\dots,5\}

). Acc is index-level recoverability label accuracy (mean

\pm

std over training seeds).

J_{X}

and

J_{Y}

are Jaccard similarities of selected recoverable subsets across training seeds;

J_{\mathrm{overall}}

is the average of

J_{X}

and

J_{Y}

. CCC statistics summarize agreement strength among selected measurement indices (mean ± std).

Figure 2: $\varepsilon$ -sweep on CAS $\leftrightarrow$ SNAP. Median number of recoverable measurements (#Rec) as a function of the feasibility threshold $\varepsilon$ , shown separately for CAS (A) and SNAP (B). Results are aggregated over random seeds.

EMBIC $\rightarrow$ NeuroBAT (target: NeuroBAT)
Direction	Method	#Ret	Ret% (mean $\pm$ std)	CCC $>$ 0.5	CCC $>$ 0.6	CCC $>$ 0.7
FRF		0	$0.0\pm 0.0$	0	0	0
IRT		0	-	0	0	0
XGB		0.5	$62.5\pm 48.4$	0.5	0	0
CMLP		1	$90.0\pm 10.0$	1	1	0
NICE		2	$49.2\pm 10.5$	2	1	0

Appendix

Table 8: Extended results for experiment 3. For each deployment direction (source

\rightarrow

target), #Ret is the median number of retained recoverable measurements in the target space, and Ret% is the corresponding retention rate (mean

\pm

std across random seeds), and is reported in percentage units. CCC

>

0.5/0.6/0.7 report the median number of retained measurements whose concordance correlation coefficient on the test period exceeds the corresponding threshold. “-” denotes not applicable (no recoverable measurement is identified on the training period, hence retention is undefined). All results use

\varepsilon_{\text{cutoff}}=0.6

Variant	$\#\mathrm{Rec}_{A}$	$J_{A}$	$\#\mathrm{Rec}_{B}$	$J_{B}$
Ours (full)	11	0.930	9	0.817
TWMLP backbone	11	0.929	9	0.947
$L_{0}$ penalty	$0^{\dagger}$	-	$0^{\dagger}$	-
No gating	11	0.806	9	0.774
No hinge	$27^{\dagger}$	-	$18^{\dagger}$	-

Appendix

Table 9: Ablation results on CAS (A)

\leftrightarrow

SNAP (B) under

\varepsilon_{\text{cutoff}}=0.6

. #Rec A /#Rec B denote median numbers of recoverable measurements.

J_{A}

J_{B}

denote Jaccard similarity across random seeds. Degenerate variants are marked with “-”.

Feb 25, 2026

IHF-Harmony: Multi-Modality Magnetic Resonance Images Harmonization using Invertible Hierarchy Flow Model

Retrospective MRI harmonization is limited by poor scalability across modalities and reliance on traveling subject datasets. To address these challenges, we introduce IHF-Harmony, a unified invertible hierarchy flow framework for multi-modality harmonization using unpaired data. By decomposing the translation process into reversible feature transformations, IHF-Harmony guarantees bijective mapping and lossless reconstruction to prevent anatomical distortion. Specifically, an invertible hierarchy flow (IHF) performs hierarchical subtractive coupling to progressively remove artefact-related features, while an artefact-aware normalization (AAN) employs anatomy-fixed feature modulation to accurately transfer target characteristics. Combined with anatomy and artefact consistency loss objectives, IHF-Harmony achieves high-fidelity harmonization that retains source anatomy. Experiments across multiple MRI modalities demonstrate that IHF-Harmony outperforms existing methods in both anatomical fidelity and downstream task performance, facilitating robust harmonization for large-scale multi-site imaging studies. Code is available at https://github.com/Idea89560041/IHF-Harmony.

Pengli Zhu, Yitao Zhu, Haowen Pang +1

Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong · School of Integrated Circuits and Electronics, Beijing Institute of Technology, China · Mental Health Research Center, The Hong Kong Polytechnic University, Hong Kong +1

Medical Image Harmonization Magnetic Resonance Imaging Unpaired Image-to-Image Translation Normalizing Flows

Ask PDF

Sep 24, 2026

Towards clinical adoption of voice and speech as measures of health: the need for harmonization

Speech and voice are multidimensional signals that capture both communicative intent and underlying physiological processes, providing a unique, non-invasive window into health. Analyzing these signals has the potential to yield digital biomarkers that (i) provide scalable, objective measurement tools for research and clinical care and (ii) reflect the presence or progression of diverse conditions, including neurological, psychiatric, respiratory, and cardiovascular disorders. Realizing this promise, however, requires the field to overcome pervasive reproducibility and generalizability issues due to heterogeneous data collection, processing, and analysis practices. A major source of this heterogeneity is how underlying acoustic measures themselves are defined and computed. In this paper, we outline key considerations across the speech biomarker discovery lifecycle, from data collection through machine learning modeling to clinical interpretation, needed to achieve reliable, reproducible, and clinically translatable results. Chief among these is the need for harmonization efforts to start from common, precisely specified measure definitions. As a first step, we therefore provide definitions, physiological correlates, and computational implementations for a minimal, clinically interpretable set of core speech measures spanning respiration, phonation, articulation, and fluency. We close by discussing ongoing standardization efforts and the open challenges that remain in advancing the adoption of speech- and voice-based digital biomarkers.

Nicholas Cummins, Vikram Ramanarayanan, Daniel Low +9

Institute of Psychiatry, Psychology and Neuroscience, King’s College London, UK · NIHR Biomedical Research Centre: Maudsley, UK · Modality.AI Inc, San Francisco, California, USA +12

Speech Processing Clinical Speech Processing

Ask PDF

Oct 9, 2026

Frozen Site-Probe Accuracy Is Not a Standalone Harmonization Score: An Audit of Evaluation Practice for Multi-Site MRI

Image-level MRI harmonization is often evaluated with a frozen site probe, a site classifier trained on raw scans and applied unchanged to harmonized ones: a drop in its accuracy is read as removal of site effects. We audit this practice on ABIDE I (1,112 T1-weighted scans, 17 sites) with ten output sets from seven harmonization methods and controls for chance, head shape, intensity, preprocessing and target alignment. We find that this verdict depends on how the probe is trained, even among probes that classify raw scans well: across random seeds, its balanced accuracy on the same CycleGAN outputs ranges from 0.13 to 0.59 (chance: 0.06). Low frozen-probe accuracy can also arise without harmonization: a Gaussian blur, which increases the intensity distance to the target site, lowers this accuracy to chance. Moreover, where the frozen probe fails, probes trained on the harmonized scans still recover the site, even from intensity histograms alone: on HACA3 outputs, the frozen probe scores 0.25, a probe trained on those outputs 0.75, and one using only their intensity histograms 0.71. Low frozen-probe accuracy therefore does not show that site information has been removed. We recommend reporting several probes trained to convergence, probes trained on harmonized scans, and target alignment.

Muhammad Muqsit Islam, Gloria García Cuenco, José M. Martínez +1

Universidad Autónoma de Madrid, Spain · Université de Bordeaux, France · Video Processing and Understanding Lab, Universidad Autónoma de Madrid, Spain +1

Medical Image Harmonization Magnetic Resonance Imaging Brain MRI

Ask PDF

Abstract

Explore similar work

IHF-Harmony: Multi-Modality Magnetic Resonance Images Harmonization using Invertible Hierarchy Flow Model

Towards clinical adoption of voice and speech as measures of health: the need for harmonization

Frozen Site-Probe Accuracy Is Not a Standalone Harmonization Score: An Audit of Evaluation Practice for Multi-Site MRI