Correlation of Patient-Reported Symptoms With Rhinogram Features Beyond Simple Airway Resistance

Abstract

Introduction:

Rhinomanometry, a reference measure for the nasal airway, is often considered a research tool with only weak-to-moderate correlations with patient symptoms. However, like lung spirometry curves offer information beyond forced expiratory volume (FEV), rhinomanometry curves (rhinograms) have characteristics beyond simple nasal resistance at 150 Pascals. This study explored the correlation between rhinogram curve features and patient-reported outcomes (PROMs), when compared with nasal airway resistance.

Methods:

A diagnostic cross-sectional study was conducted on patients from a rhinology clinic. PROMs collected included ordinal nasal obstruction and visual analogue scale (VAS) of the more obstructed side. Rhinomanometry curves underwent mathematical polynomial fitting to extract 835 features. The primary outcome was correlation using Spearman’s rho (ρ) comparing curve-derived features with nasal airway resistance at 150 Pascals. Machine learning was applied to the top 8 correlated features to generate an AI predictive model.

Results:

About 601 patients (mean age 45 ± 16 years, 45% female) were analysed. Curve-derived features (ρ = 0.305) correlated more than total NAR at 150 Pa (ρ = 0.222) with VAS. Similarly with ordinal nasal obstruction, curve-derived features correlated more (ρ = 0.230) than total NAR at 150 Pa (ρ = 0.112). The best performing AI prediction models achieved correlations of 0.133 (VAS) and 0.117 (nasal obstruction).

Conclusion:

This study offers a novel method for rhinogram analysis with curve-derived features for correlation and predictive modelling. Whilst correlation scores remain weak-moderate with PROMs, they outperform nasal airway resistance. Therefore, rhinograms produced from rhinomanometry may offer more clinical information than a simplistic numerical resistance testing.

Keywords

rhinomanometry patient reported outcome measures mathematical computing airway resistance image interpretation

Introduction

In rhinology, there is a need for reliable objective methods to assess nasal patency and obstruction.^1,2 Four-phase active anterior rhinomanometry, forwarded by Vogt et al, is an objective tool used to assess nasal airflow and resistance quantitatively.² Its common applications include the assessment of nasal obstruction, preoperative evaluation in rhinoplasty and evaluation of response to treatment.¹ In current practice, rhinomanometry remains largely a research based tool due to technical expertise in setup, human resource in application and the variable correlation with patient reported outcomes (PROMs) such as the sinonasal outcome test (SNOT22), nasal obstruction and septoplasty effectiveness score (NOSE) and visual analogue score (VAS).^1,2 Correlation scores between PROMs and rhinomanometry derived nasal airway resistance (NAR) have ranged from weak to high (correlation r score −.3 to .9), with stronger relationships for individual nasal passages and in patients with allergic rhinitis or septal deviations.^3-5 Variable qualities of studies and assessment metrics (such as Pearson’s correlation vs Spearman’s rho) have also made interpretation of such correlations difficult.^3-5 To date the majority of research has utilised only the basic measurement of airway resistance at 150 Pascals whilst leaving the graphical rhinomanometry curve (or rhinogram) for subjective interpretation by individual practitioners.^3,5 This renders simplistic interpretation, reducing complex airflow curves (rhinograms) to simple numeric values of resistance.^3-5 Intuitively, the rhinogram may harbour further rich clinical information about the patient experience, much like 4 phase spirometry offers more than only Forced Expiratory Volume (FEV) in respiratory practice.¹

The key component of the rhinogram is the pressure-flow curve formed during the inspiration and expiration of a patient. Whilst rhinomanometry machines generate this curve as an image, there are no raw coordinates data or underlying curve equations available. This limits the application of these curves to qualitative and often anecdotal analysis. To address similar gaps, mathematical techniques such quadratic and polynomial fitting used to derive curve equations from image data have been used in medical applications such as electrocardiogram (ECG), spirometry and optical coherence tomography (OCT) interpretation.^6-8 These approaches allow for higher-order and non-linear quantitative feature extraction from visual data such as gradients, curvatures, asymmetry and areas. In addition, artificial intelligence-based techniques such as machine learning can be applied in the analysis large numbers of image-based features. Thus creating interpretable information to support clinician understanding and deeper analysis of physiological attributes beyond single-point numerical metrics.

As curve-based mathematical analysis allows for the extraction of more nuanced features, its application to rhinogram analysis offers the potential for new clinical insights in understanding the subjective patient experience of nasal obstruction. This study applied mathematical techniques and machine learning modelling to examine the relationship between rhinograms and patient-reported outcomes (PROMs), when compared with standard nasal airway resistance testing at 150 Pascals.

Materials and Methods

Study Design

A cross-sectional study was performed to evaluate the relationship between patient-reported symptom scores and rhinograms with polynomial fitting mathematical techniques. Baseline characteristics, rhinomanometry and patient-reported outcomes (collected as part of routine care) were retrospectively collected from patient clinical files. Ethics approval was obtained from the St Vincent’s Hospital Human Research Ethics Committee (2019/PID13822).

Study Population

The population included adult patients who presented to a tertiary rhinology clinic for any reason and underwent 4-phase active anterior rhinomanometry testing.

Rhinomanometry

Four phase active anterior rhinomanometry was measured with a NR6 Rhinomanometer (GM Instruments, UK) at a reference level of 150 Pa (international standard).⁹ With the patient upright, the contralateral nostril was occluded with a foam plug fitted to the end of a pressure-sensing tube and sealed with micropore. The anaesthetic-style mask was held tightly against the patient’s face, and the patient was asked to breathe normally with their mouth closed (Figure 1). Three readings were obtained, with a minimum of 2 readings within 10% of each other. This was repeated to obtain results for each nostril. The results were collated on the NARIS programme (GM Instruments, UK)¹⁰ with the total, left and right nasal airway resistance (NAR) recorded as outcomes for validation. To represent nasal airway resistance for the more obstructed side, left and right sided NAR measurements were compared and the higher value was selected.

Figure 1.

Image of rhinomanometry testing conducted on a patient with placement of the sealed pressure sensing tube and anaesthetic style mask.

Patient Reported Outcomes

Ordinal Nasal Obstruction Score

Each patient graded the severity of their nasal obstruction with an ordinal 6-point Likert scale prior to rhinomanometry. This produced a response ranging from 0 (no problem) to 5 (problem as bad as it can be). A combination of response ranges without a defined ‘control’ or ‘normal’ group were utilised to allow for a more representative sample with graded relationships.

Visual Analogue Scale (VAS) Obstructed Side

Patients were asked to indicate the severity of their nasal obstruction in each nostril on a scale from 0 (no obstruction) to 100 (total obstruction). The higher score from each patient’s nostrils was taken as the more obstructed side used for analysis.

Image Pre-Processing

All images were exported as portable network graphics (PNG) files and cropped to only capture the graphical component of the rhinomanometry output (ie, rhinogram) without any clinical information or numerical values to avoid erroneous noise. The images were standardised to grayscale, targeted brightness to 225 and contrast to 39 and resized to 640 × 640-pixel resolution. The fixed labels from the images and grid lines were removed to produce the isolated curves. Pre-processing was conducted in Python with libraries including pandas, numpy, matplotlib and seaborn. Other pre-processing techniques including Gaussian Blur and enhancement, binary threshold and morphological cleaning, contour filtering and edge cleanup were utilised. A visual representation of the pre-processing phases is demonstrated in Figure 2.

Figure 2.

Stepwise process of pre-processing for polynomial fitting of rhinogram. (A) Original rhinogram image with labels denoting the pressure-flow curve with inspiration and expiration. Pressure measured in Pascals (Pa) and flow in cubic centimetres per second (cc/s). (B) Rhinogram with labels and markers removed from image using fixed label removal. (C) Rhinogram with grid lines removed and colour processing. (D) Rhinogram with colour inversion, gaussian blur application and contrast enhancement to make curve lines more prominent for analysis. (E) Rhinogram with cleaned image, edge cleanup and skeletonisation using Zhang-Suen thinning algorithm.¹¹ (F) Rhinogram with third degree polynomial fitting used to derive features.

Curve-Derived Feature Extraction

Polynomial fitting analysis was utilised to capture the fundamental mathematical relationships describing rhinogram curvature. third, fifth, seventh and ninth degree polynomials were utilised to generate 835 features for curve analysis. These features comprised of basic geometric features (19), basic morphological features (ie, shape and curve) (159) and advanced geometric features (657 across 4 polynomial degrees). Advanced geometric features were distributed as third degree: 159 features, fifth degree: 166 features, seventh degree: 166 features, ninth degree: 166 features. A description of the features extracted and calculations are presented in Supplemental Material 1 and an example feature is graphically demonstrated in Figure 3.

Figure 3.

Example of a Rhinogram curve-derived feature demonstrating resistance_grad_max_asymmetry. This feature corresponds to the difference between the two maximal gradients of the left and right side, as marked. d/dx, derivative or gradient at the marked point.

AI Prediction Model

Following the correlation analysis, the top 8 geometric features were selected for each outcome. These features were evaluated with supervised machine learning models to create a prediction model using linear models (ie, Linear regression, Ridge regression, LASSO regression, ElasticNet), Support Vector Machines (Linear, RBF and Polynomial SVM) and Ensemble Methods (Random Forest and XGBoost). The dataset was split into 70% training and 30% validation sets with 5-fold cross validation.

Two evaluation metrics were utilised in analysis:

Correlation coefficient (R²)

The correlation coefficient R-squared (R²) is a common metric used to assess the amount of variance between variables in a predictive model.¹² Scores typically range from 0 (poor correlation between variables by the model) to 1 (perfect correlation between variables by the model).

2. Root mean squared error (RMSE)

Root mean squared error is another metric utilised to evaluate predictions by square rooting the mean squared differences between true and predicted values.¹² It is an easily interpretable measure of average error size however is overly sensitive to outliers and is better in conjunction with other metrics.¹²

Statistical Analysis

Baseline population characteristics were calculated using SPSS version 29 with continuous normally distributed outcomes presented as mean ± standard deviation (SD), ordinal or continuous, non-normally distributed data presented as median [interquartile range] and categorical results were presented as percentages. Correlation was calculated using the Spearman’s rho (ρ) for non-parametric data with 95% confidence intervals. All P values were 2-tailed, and a value of P < .05 was considered statistically significant. For the predictive modelling, statistical analysis to calculate R² and RMSE was conducted using Python version 3.11 packages for statistics with pandas and scipy libraries.

Results

Population Characteristics

601 patients were analysed in this study 45 ± 16 years and 45% were female. 14% of the population were active smokers and 4% reported current drug use. 56% of patients experienced nasal obstruction to a ‘moderate’ or greater degree and VAS obstructed side of 60 [22-80] also representative of a ‘moderate’ concern. The median total nasal airway resistance was 0.12 [0.10-0.15] Pa/cm³/s below previously identified levels for congested nasal mucosa of 0.25 Pa/cm³/s.¹³ Further details are as per Supplemental Material 3.

Correlation With Patient-Reported Outcomes

Visual Analogue Score (VAS) Obstructed Side

The conventional and established total NAR at 150 Pascals had a correlation with the VAS of .209 (P < .001). When comparing curve-derived features with VAS obstructed side, the highest correlation demonstrated was .305 (P < .001) as shown in Table 1. 241 curve-derived features demonstrated stronger correlation with VAS obstructed side than total nasal airway resistance (NAR) measured at 150 Pa. The top 10 curve-derived features were advanced asymmetry features for third degree polynomials with correlations ranging from 0.292 to 0.305. Further results are detailed in Table 1 and Supplemental Material 4.

Table 1.

Patient Characteristics for Highest Curve-Derived Features.

	Curve characteristics			Total NAR at 150 Pa (Pa/cm³/s)
n	601			601
	Features	Correlation	P value	Correlation	P value
More obstructed side (VAS 0-100)	Highest curve derived feature 1	.305	<.001	.222	<.001
Nasal obstruction (ordinal score 0-5)	Highest curve derived feature 2	.230	<.001	.112	.006

Note. Highest correlated curve-derived features for visual analogue score more obstructed side and nasal obstruction ordinal score in population of 634 with a comparison of total nasal airway resistance (NAR) at 150 Pa. Correlation reported as Spearman’s rho with its corresponding P value. Please refer to Supplement 1 for further explanations of feature names.

Abbreviations: VAS, Visual Analogue Scale; Pa, Pascals; NAR, nasal airway resistance; Pa/cm3/s, Pascals per cubic centimetre per second.

Ordinal Nasal Obstruction Score

The conventional and established total NAR at 150 Pascals had a correlation with ordinal nasal obstruction of .107 (P < .001). When comparing curve-derived features with ordinal nasal obstruction, the highest correlation demonstrated was .230 (P < .001) as shown in Table 2. 377 curve-derived features demonstrated stronger correlation with ordinal nasal obstruction score than total nasal airway resistance (NAR) measured at 150 Pa. In addition, 240 curve-derived features demonstrated higher correlations than total NAR. The top 10 curve-derived features were also advanced asymmetry features for third degree polynomials with correlations ranging from 0.214 to 0.230. Further results are detailed in Table 1 and Supplemental Material 4.

Table 2.

Airflow Analysis for Highest Curve Features.

Resistance value (at 150 Pa)	Curve characteristics
Resistance value (at 150 Pa)	Curve feature	Correlation	95% CI	P value
Left (Pa/cm³/s)	adv3_area_L	.974	0.970-0.978	<.001
Right (Pa/cm³/s)	adv3_area_R	.968	0.962-0.973	<.001
Obstructed side (Pa/cm³/s)	adv3_area_total	.952	0.943-0.959	<.001
Total (Pa/cm³/s)	adv5_area_total	.941	0.931-0.950	<.001

Note. Highest correlated curve-derived features for nasal airway resistance values at 150 Pascals including left, right, obstructed side and total values. Correlation reported as Spearman’s rho with 95% confidence intervals and corresponding P value. Please refer to supplement 1 for further explanations of feature names.

Abbreviations: 95% CI, 95% confidence interval; Pa/cm³/s, Pascals per cubic centimetre per second.

Correlation With Airflow Analysis Outcomes

Nasal Airway Resistance (NAR)

As expected, nasal airway resistance values were strongly correlated with curve-derived features with highest values ranging from 0.941 to 0.974 (P < .001). NAR left and right were more strongly correlated (0.974/0.968 respectively) with a curve-derived feature than NAR obstructed side and total (0.952/0.941 respectively). Correlation scores were greater than 0.90 (with positive or negative correlation) in 25 features (left NAR), 20 features (right NAR), 13 features (NAR obstructed side) and 12 features (NAR total). Results are presented in Table 2.

AI Predictive Model

The best performance model for more obstructed side VAS was ElasticNet achieving R2 of .133 and RMSE of 27.81. The best performance model for ordinal nasal obstruction was RandomForest achieving R² of .117 and RMSE of 1.59. Results are as depicted in Table 3.

Table 3.

Best Performance Models for Predictive Ability With Patient-Reported Outcomes.

	Model	Number of features	Correlation R² (test)	RMSE
More obstructed side (VAS 0-100)	ElasticNet	8	.133	27.81
Nasal obstruction (ordinal score 0-5)	RandomForest	8	.117	1.59

Note. Summary of the best performing models utilising the top 8 curve-derived features to predict patient-reported outcomes including visual analogue score of more obstructed side and nasal obstruction. Evaluation metrics included correlation using R² of the test dataset and prediction error using root mean squared error.

Abbreviations: VAS, visual analogue score; RMSE, root mean squared error.

Discussion

In contrast to rhinomanometry, active 4-phase spirometry is a commonly used tool in respiratory medicine, where various components of the spirogram are clinically interpreted. The inspiratory and expiratory phases are reflective of lung volumes which contribute to total lung capacity and curve features such as concavity and height suggest obstructive or restrictive patterns of disease that is, Chronic obstructive pulmonary disease (COPD) versus interstitial lung disease.^14,15 These features aid with diagnosis, monitoring of disease progression and assessment of therapeutic responses in respiratory disease.^14,15 Advances in artificial intelligence have created unique avenues for image analysis of the spirogram. Largely these applications have focused on deep learning classification of obstructive disease such as asthma and COPD, and restrictive disease such as interstitial lung disease with high accuracies >80%.^16,17 Even recently, SpiroLLM was proposed as a large language model to extract morphological features from raw time series spirograms in the diagnosis of COPD with Area under the Receiver Operating Characteristic Curve (AUC-ROC) close to 90%.¹⁸ Image based analysis offers the ability to detect subtle and nuanced curve features which may outperform traditional metrics and enhance interpretability.

In rhinology, active 4-phase rhinomanometry offers similar data, but despite the image curve, the numeric nasal airway resistance measurement at 150 Pascals is usually all that is considered. To date, literature has only analysed the relationship between numeric values of nasal airway resistance and patient reported symptoms scores.^3-5 The correlation values in literature are largely weak to moderate (−.33 to .261) for VAS and an obstruction score with some stronger correlations seen for relating the Italian NOSE score (.602) and SNOT22 (.524).^4,19-25 A summary of literature relating this is shown in Table 4.^4,19-25

Table 4.

Summary of Literature Correlating Nasal Airway Resistance With Patient-Reported Outcomes.

Study	Year	N	Diagnosis	Index	Reference test	Outcome
Ciprandi et al¹⁹	2008	103	Allergic rhinitis	VAS	Rhinomanometry – Total NAR	ρ = −0.33
Ciprandi et al²⁰	2009	50	Allergic rhinitis	VAS	Rhinomanometry – Total NAR	ρ = 0.879
Hsu et al⁴	2017	50	Nasal obstruction	VAS, NOSE	Rhinomanometry – Total and Obstructed side NAR	VAS: ρ = 0.214 (total), ρ = 0.279 (obstructed side) NOSE: ρ = 0.249 (total), ρ = 0.188 (obstructed side)
Mendes et al²¹	2012	30	Allergic rhinitis	Obstruction score	Rhinomanometry - Total and Obstructed side NAR	Total: ρ = 0.147, Obstructed side - ρ = 0.261
Mozzanica et al²²	2013	60	Nasal obstruction	I-NOSE	Rhinomanometry - Total and Obstructed side NAR	Total: ρ = 0.602, Obstructed side: ρ = 0.639
Ottaviano et al²³	2022	101	Any reason presenting to ENT clinic	VAS, SNOT22	Rhinomanometry – Total	VAS: ρ = 0.013, SNOT22: ρ = 0.524
Prus-Ostaszewska et al²⁴	2017	70	Nasal obstruction	SNOT20	Rhinomanometry - Total	Post surgery Kendall’s Tau 0.12
Tompos et al²⁵	2010	86	Septal deviation	VAS	Post-op rhinomanometry – Total NAR	Obstructed side ρ = −0.24

Note. Literature review of studies which correlated nasal airway resistance measured with 4-phase active rhinomanometry and patient-reported outcomes scores, specifically including visual analogue score (VAS) or a nasal obstruction score.

Abbreviations: ρ, Spearman’s rho for correlation; VAS, Visual analogue scale; NAR, nasal airway resistance; NOSE, nasal obstruction symptom evaluation; SNOT22, Sino-Nasal Outcome Test 22; SNOT20, Sino-Nasal Outcome Test 20; I-NOSE, Italian nasal obstruction symptom evaluation.

This study aligns with the current literature regarding simple correlation scores of airway resistance at 150 Pascals with VAS (.209) and ordinal nasal obstruction score (.107). With the addition of mathematical models to capture nonlinear and higher order features from curve morphology for analysis, this study was able to show higher correlation than conventional nasal airway resistance at 150 Pascals, albeit still weak to moderate. Rhinogram morphology might lead to greater correlation with the subjective patient experience when compared with simple nasal airway resistance at 150 Pa.^3-5 Accordingly, there is clinical value in interpreting both NAR and Rhinograms features to contextualise nasal airflow symptoms.

Although correlations between objective airflow measurements and patient-reported symptoms remain modest, this does not necessarily limit the clinical value of objective nasal airflow assessment. Patient perception of nasal obstruction is influenced by multiple physiological and neuro-sensory factors, and therefore a perfect correlation between airflow measurements and symptoms is unlikely. Whilst rhinomanometry may be objective, measurement technique needs to be precise and variables such as the nasal cycle and sinonasal disease can influence airflow or nasal airway resistance recorded.¹ Subjective patient-reported scores also may not fully isolate patients’ nasal obstruction experience, with responses influenced by comprehension of nasal congestion versus nasal obstruction and complex factors beyond nasal patency including psychological factors, comorbidities and patient interpretation of questions.⁵

Objective tests such as rhinomanometry instead provide complementary information regarding nasal airway physiology, structural airflow limitation and response to intervention. In this context, enhanced analysis of rhinogram curves may improve the physiological interpretation of rhinomanometry by capturing airflow dynamics that are not reflected in conventional single-point resistance measurements. Such information may ultimately assist in better characterising patterns of nasal airflow limitation and refining objective assessment in both clinical and research settings.

It is also possible that airflow characteristics might align better with the heatflux and radiant cooling effects of nasal airflow, which affect perception of nasal obstruction. In addition, curve analysis could provide insights into the assessment of dynamic nasal valve function, with curve features relating with other airflow characteristics and ultimately human perception of nasal obstruction. Further studies are required to fully understand whether rhinogram morphology should be utilised independently or in conjunction with NAR to optimise correlation with patient symptom scores.

Rhinograms can be applied clinically as an objective measure of nasal obstruction. This may aid in differentiating patient meaning when describing nasal congestion – a term which can reflect obstructive, pressure-related or even mucous-related changes.²⁶ Rhinograms can be useful in determining whether nasal surgery might be appropriate for patients with sleep dysfunction. With objective data, it could also help guide clinicians to whether patients are truly referring to nasal airway obstruction or a ‘pressure’ sensation when referring to ‘nasal congestion’. Furthermore, it may help avoid surgical interventions for patients with nasal related functional disorders such as hyperventilation or functional nasal disorders.^27,28

Limitations in this study include the single centre retrospective study, reliance on fitted curves and noise/quality of images affecting results. Noise in image-based analysis is a well-known issue, whilst standardisation and image pre-processing was completed, there may still be small differences in image quality which can affect curve features extracted and skew correlation outcomes.^29,30 In predictive modelling, patient experience may also not be reliably predicted as curve features may only capture part of the complexity of patient perception and multimodal clinical variables may be necessary to improve predictions.

Practical Implementation

The approach described in this study does not require additional testing beyond standard rhinomanometry. Rhinogram pressure–flow curves are routinely generated during rhinomanometry and are already available in digital form in most modern systems. Curve-based feature extraction represents an analytical extension of existing data rather than a new diagnostic procedure. With appropriate software integration, automated analysis of rhinogram curves could generate additional parameters alongside conventional nasal airway resistance measurements without altering clinical workflow. Future work may focus on identifying a smaller number of clinically interpretable curve-derived metrics and integrating automated rhinogram analysis into rhinomanometry platforms for routine clinical use.

Implications and Future Directions

Rhinogram images can be leveraged for AI analysis. Multiple metrics based on curve-derived features may be able to produce reliable and reproducible interpretation of rhinomanometry beyond single-value resistance. This study provides foundational understanding about the importance of side dominant and asymmetry-based feature extraction in nasal airflow and supports the integration of curve analytics into translational research and AI workflows. Future applications of these curve-derived features may be in AI based classification of disease severity, clustering of curve features to identify airflow phenotypes and prediction of post-surgical outcomes to guide surgical planning and patient expectations. Providing an objective basis for nasal obstruction assessment could be beneficial in monitoring medical and surgical response, support medico-legal assessments and developing standardised guidelines for clinician interpretation of rhinomanometry. In addition, future studies may also explore linking objective nasendoscopy or rhinoscopy findings such as nasal valve collapse with objective curve-based characteristics to elucidate more information about fixed and dynamic nasal obstruction.³¹

Conclusion

This study demonstrates that polynomial-derived curve features from rhinograms offer more clinically relevant insights, surpassing simple numeric NAR values at 150 Pascals. These features have higher correlations with the patient experience, suggesting the value of including interpretation of the rhinogram curve morphology in nasal obstruction assessment. Future research should focus on refining feature extraction, expanding patient cohorts and developing AI based assessment tools that bridge the gap between objective measurement and patient experience.

Supplemental Material

sj-docx-1-aor-10.1177_00034894261454654 – Supplemental material for Correlation of Patient-Reported Symptoms With Rhinogram Features Beyond Simple Airway Resistance

Supplemental material, sj-docx-1-aor-10.1177_00034894261454654 for Correlation of Patient-Reported Symptoms With Rhinogram Features Beyond Simple Airway Resistance by Rhea Darbari Kaul, Raquel Alvarado, Christine Choy, Haiyang Sun, Sidong Liu, Elizabeth Hua, Liam Grouse, Masoud Haghighi, Kate Liang, Emma Zou, Aari Desai, Nicholas J Campion, Cedric Thiel, Ghasem Azemi, Raymond Sacks, Raewyn G. Campbell, Larry Kalish, Antonio Di Ieva and Richard J Harvey in Annals of Otology, Rhinology & Laryngology

Footnotes

Acknowledgements

We would like to thank Dr Peta-Lee Sacks for her contributions to patient recruitment and for her valuable assistance in this project.

ORCID iDs

Rhea Darbari Kaul

Raquel Alvarado

Elizabeth Hua

Liam Grouse

Masoud Haghighi

Kate Liang

Emma Zou

Aari Desai

Nicholas J Campion

Cedric Thiel

Ghasem Azemi

Raewyn G. Campbell

Antonio Di Ieva

Richard J Harvey

Ethical Considerations

Ethics approval was obtained from the St Vincent’s Hospital Human Research Ethics Committee (2019/PID13822).

Author Contributions

RDK: conceptualisation, investigation, methodology, software, formal analysis, writing of manuscript, editing, accepting final version of manuscript. RA, CC: investigation, methodology, project administration, editing, accepting final version of manuscript. HS, SL, GA: software, formal analysis, editing, accepting final version of manuscript. EH, LG, MH, KL, EZ, AD, NJC, CT: investigation, resources, editing, accepting final version of manuscript. RS, RGC, LK, ADI: conceptualisation, supervision, editing, accepting final version of manuscript. RJH: conceptualisation, methodology, investigation, supervision, formal analysis, editing, accepting final version of manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Richard J Harvey is a consultant/advisory board with Medtronic, Novartis, GSK and Meda pharmaceuticals. Research grant funding received from Glaxo-Smith-Kline. He has been on the speakers’ bureau for Glaxo-Smith-Kline, Astra-Zeneca, Meda Pharmaceuticals and Seqirus. Larry Kalish is on the speakers’ bureau for Care Pharmaceuticals, Mylan and Seqirus Pharmaceuticals. Raymond Sacks is a consultant for Medtronic. Antonio Di Ieva has received research funding from B Braun, Abbott, Servier, NEVRO and BrainLab. All other authors have no financial disclosures or conflicts of interest.

Data Availability Statement

The majority of data supporting the findings of this study are available within the paper and its Supplemental Information. Any other data generated during this study are available from the corresponding author on reasonable request.*

Supplemental Material

Supplemental material for this article is available online.

References

Patil

Jain

Rhinomanometry: a comprehensive review of its applications and advancements in rhinology practice. Cureus. 2024;16(5):e61370. doi:10.7759/cureus.61370

Vogt

Jalowayski

Althaus

, et al. 4-phase-rhinomanometry (4PR)–basics and practice 2010. Rhinol Suppl. 2010;21:1-50.

Mozzanica

Gera

Bulgheroni

Ambrogi

Schindler

Ottaviani

Correlation between objective and subjective assessment of nasal patency. Iran J Otorhinolaryngol. 2016;28(88):313-319.

Hsu

Tan

Chang

, et al. Evaluation of nasal patency by visual analogue scale/nasal obstruction symptom evaluation questionnaires and anterior active rhinomanometry after septoplasty: a retrospective one-year follow-up cohort study. Clin Otolaryngol. 2017;42(1):53-59. doi:10.1111/coa.12662

Gao

Philpott

A systematic review to examine the relationship between objective and patient-reported outcome measures in sinonasal disorders: recommendations for use in research and clinical practice. Int Forum Allergy Rhinol. 2021;11(5):910-923. doi:10.1002/alr.22744

Jothiramalingam

Anitha

Hemanth

. Diagnosis of coronary artery occlusion by fitting polynomial curve with the ECG signal based on optimization techniques. Netw Model Anal Health Inform Bioinform. 2022/03/26;11(1):16. doi:10.1007/s13721-022-00354-6

Maldonado-Franco

Giraldo-Cadavid

Tuta-Quintero

Cagy

Bastidas Goyes

Botero-Rosas

DA.

Curve-modelling and machine learning for a better COPD diagnosis. Int J Chron Obstruct Pulmon Dis. 2024;19:1333-1343. doi:10.2147/COPD.S456390

Lake

Bottema

Williams

Reynolds

The correlation between optical coherence tomography retinal shape irregularity and axial length. PLoS One. 2019;14(12):e0227207. doi:10.1371/journal.pone.0227207

Clement

PA.

Committee report on standardization of rhinomanometry. Rhinology. 1984;22(3):151-155.

10.

Liang

Wijermars

Darbari Kaul

, et al. Nasal valve tensioning versus crural strut grafts in the management of external nasal valve dysfunction. Plast Reconstr Surg. 2025:1–17.

11.

Chen

Sui

Lang

. Improved Zhang-Suen thinning algorithm in binary line drawing applications. 2012 International Conference on Systems and Informatics, ICSAI 2012. May 1, 2012. doi:10.1109/ICSAI.2012.6223430

12.

Kocak

Klontzas

Stanzione

, et al. Evaluation metrics in medical imaging AI: fundamentals, pitfalls, misapplications, and recommendations. Eur J Radiol Artif Intell. 2025;3:100030. doi:10.1016/j.ejrai.2025.100030

13.

Merkle

Kohlhas

Zadoyan

Mösges

Hellmich

Rhinomanometric reference intervals for normal total nasal airflow resistance. Rhinology. 2014;52(4):292-299. doi:10.4193/Rhino13.220

14.

Bhakta

McGowan

Ramsey

, et al. European Respiratory Society/American Thoracic Society technical statement: standardisation of the measurement of lung volumes, 2023 update. Eur Respir J. 2023;62(4):2201519. doi:10.1183/13993003.01519-2022

15.

Graham

Steenbruggen

Miller

, et al. Standardization of spirometry 2019 update. An official American Thoracic Society and European Respiratory Society technical statement. Am J Respir Crit Care Med. 2019;200(8):e70-e88. doi:10.1164/rccm.201908-1590ST

16.

Hussain

Marlowe

Ali

, et al. A systematic review of artificial intelligence applications in the management of lung disorders. Cureus. 2024;16(1):e51581. doi:10.7759/cureus.51581

17.

Sunjaya

Edwards

Harvey

, et al. Validation of artificial intelligence spirometry diagnostic support software in primary care: a blinded diagnostic accuracy study. ERJ Open Res. 2025;11:00116-02025. doi:10.1183/23120541.00116-2025

18.

Mei

Long

Cao

, et al. SpiroLLM: finetuning pretrained LLMs to understand spirogram time series with clinical validation in COPD reporting. arXiv preprint arXiv:250716145. 2025.

19.

Ciprandi

Klersy

Ameli

Cirillo

Clinical assessment of a nasal decongestion test by visual analog scale in allergic rhinitis. Am J Rhinol. 2008;22(5):502-505. doi:10.2500/ajr.2008.22.3214

20.

Ciprandi

Mora

Cassano

Gallina

Mora

Visual analog scale (VAS) and nasal obstruction in persistent allergic rhinitis. Otolaryngol Head Neck Surg. 2009;141(4):527-529. doi:10.1016/j.otohns.2009.06.083

21.

Mendes

AIS

Wandalsen

Solé

. Objective and subjective assessments of nasal obstruction in children and adolescents with allergic rhinitis. J Pediatr. 2012;88(5):389-395. doi:10.2223/jped.2213

22.

Mozzanica

Urbani

Atac

, et al. Reliability and validity of the Italian nose obstruction symptom evaluation (I-NOSE) scale. Eur Arch Otorhinolaryngol. 2013;270(12):3087-3094. doi:10.1007/s00405-013-2426-z

23.

Ottaviano

Pendolino

Scarpa

, et al. Correlations between peak nasal inspiratory flow, acoustic rhinometry, 4-phase rhinomanometry and reported nasal symptoms. J Pers Med. 2022;12(9):1–9. doi:10.3390/jpm12091513

24.

Prus-Ostaszewska

Wysocki

Niemczyk

Balcerzak

The correlation of the results of the survey SNOT-20 of objective studies of nasal obstruction and the geometry of the nasal cavities. Otolaryngol Pol. 2017;71(2):1-7. doi:10.5604/01.3001.0009.8408

25.

Tompos

Garai

Zemplén

Gerlinger

Sensation of nasal patency compared to rhinomanometric results after septoplasty. Eur Arch Otorhinolaryngol. 2010;267(12):1887-1891. doi:10.1007/s00405-010-1278-z

26.

McCoul

Mohammed

Debbaneh

Carratola

Patel

AS.

Differences in the intended meaning of congestion between patients and clinicians. JAMA Otolaryngol Head Neck Surg. 2019;145(7):634-640. doi:10.1001/jamaoto.2019.1023

27.

Png

Kalish

Sacks

Empty nose syndrome: the case for “functional nasal obstruction” as a predisposing risk prior to nasal surgery. Curr Otorhinolaryngol Rep. 2023;11(4):422-429. doi:10.1007/s40136-023-00487-w

28.

Smith

“Functional disorders”: one of medicine’s biggest failures. BMJ. 2023;380:221. doi:10.1136/bmj.p221

29.

Hussain

Hyeon Gu

. Exploring the impact of noise and image quality on deep learning performance in DXA images. Diagnostics. 2024;14(13):1–22. doi:10.3390/diagnostics14131328

30.

Mudhafar

RAA

Abbadi

NKE

. Noise in Digital Image Processing: A Review Study. Institute of Electrical and Electronics Engineers. 2022:79-84.

31.

Gagnieur

Fieux

Louis

Béquignon

Bartier

Vertu-Ciolino

Objective diagnosis of internal nasal valve collapse by four-phase rhinomanometry. Laryngoscope Investig Otolaryngol. 2022;7(2):388-394. doi:10.1002/lio2.784

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.03 MB

0.00 MB