Does Reducing Videofluoroscopy Frame Rate Affect DIGEST grades in Modified Barium Swallow Studies?

Abstract

Understanding the effects of reduced frame rates on the reliability of Modified Barium Swallow (MBS) ratings for swallowing safety and efficiency is essential for clinical practice. While previous research has examined frame rate (simulated pulse rates) implications concerning penetration, aspiration, and residue ratings, the impact on summary grading systems like the Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) remains unclear. This study analyzed a total of 315 full MBS videos collected from 278 patients, and 76.25% of patients were diagnosed with head and neck cancer (mean age 62.51 years, SD 8.53; 12.23% female). We performed pairwise comparisons of DIGEST grades and DIGEST grade components (Safety and Efficiency) derived independently across studies with 30, 15, and 7.5 frames per second (FPS). Weighted Cohen’s kappa values consistently exceeded 0.84 across all assessments, indicating “almost perfect” agreement among the different simulated pulse rates. Exact agreement for all comparisons surpassed 85%. These findings suggest that the DIGEST grading system is robust to variations in frame rate, allowing for reliable assessments even under reduced pulse rate conditions.

Introduction:

Dysphagia, or difficulty swallowing, is a prevalent and debilitating complication commonly experienced by patients with head and neck cancer (HNC). This condition not only impairs quality of life but also poses significant risks such as malnutrition, dehydration, and aspiration pneumonia, which can complicate cancer control and survivorship. Accurate grading and monitoring of swallowing function are critical components of the comprehensive care plan for patients with cancer.

The Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) is a pivotal tool in this process, providing a reproducible approach to grade swallowing function on videofluoroscopy in alignment with cancer toxicity conventions(1, 2). This evidence-based grading system comprises two key components: the Safety grade and the Efficiency grade. The Safety grade assesses patterns of aspiration and penetration (per penetration-aspiration scale [PAS] scores (3)), critical factors that can lead to severe respiratory complications if not properly managed. Meanwhile, the Efficiency grade evaluates the residue left in the pharynx after a swallow, an indicator of how fully a patient can clear their bolus, which impacts nutritional intake and the risk of post-swallow aspiration. DIGEST was initially developed and implemented in videofluoroscopy (or the modified barium swallow [MBS] study). MBS studies are a workhorse in evaluating swallowing disorders by providing detailed, real-time visualization of the swallowing mechanics across different phases. While indispensable, MBS studies carry the inherent risk of ionizing radiation exposure, which while limited must be managed per the “as low as reasonably achievable” (ALARA) principle.

In MBS studies, the de facto favored pulse rate of 30 pulse per second (PPS) has traditionally been considered to strike a balance between capturing detailed imaging of swallowing mechanics and managing radiation exposure(4). However, limited evidence is available to guide the optimal choice under the ALARA principle, which mandates minimizing radiation exposure to the lowest levels possible, while maintaining diagnostic accuracy. Each frame in fluoroscopic imaging corresponds to a beam pulse of X-ray radiation; therefore, reducing the pulse rate directly decreases the number of X-ray pulses emitted. Lower pulse rate s could potentially reduce radiation exposure significantly (5–7) and continue to be implemented clinically in many environments. In Europe, Australia, Japan, and South America, standard video frame rates are slightly lower at 25 (frame per second) FPS due to differences in image capture standards, with pulse rates remaining unchanged, resulting in fewer images for clinical evaluation(8, 9). In the review paper (10), the authors suggest a frame rate of over 15 FPS is sufficient for pediatric MBS, however, robust data are not yet available to support this assertion. Also, it remains uncertain if these practices strike the right compromise between valid diagnostic results and patient safety.

While reducing the pulse rate in MBS studies is a common method to decrease radiation exposure, this approach inherently involves dropping information from the video—50% of frames are lost when reducing from 30 to 15 PPS, and 75% when reducing further to 7.5 PPS. This significant reduction in frames may result in the loss of critical moments in the swallowing process that are essential for accurate rating. Swallowing involves rapid, complex movements that can vary significantly within milliseconds. Key events such as the brief airway entry or the quick passage of a bolus through the pharynx might occur between the frames captured at lower rates, potentially leading to underestimations of airway invasion or swallowing safety. Reducing the number of frames captured not only limits the temporal resolution but may also obscure subtle abnormalities that are vital for an accurate assessment. Previous studies have begun to explore the impact of reducing fluoroscopy pulse rate in MBS studies on clinical ratings by adjusting the video’s frame rate to simulate this reduction. Mulheren et al. (11). compared the PAS and Modified Barium Swallow Impairment Profile (MBSImP) scores between 30 and 15 FPS with 20 patients and found no significant differences in PAS ratings, though MBSImP scores did vary, suggesting that only some diagnostic elements may be frame-rate sensitive. Similarly, Bonilha et al. (5) assessed PAS rating agreement across multiple frame rates—30, 15, 7.5, and 4 FPS—with 5 patients, finding the highest agreement between 30 and 15 FPS (24% disagreement rate), with lower agreement at the reduced rates of 7.5 and 4 FPS(31% disagreement rate). They later conducted another study with larger samples (patients = 200) examining the impact of reduced fluoroscopy pulse rates on diagnostic accuracy in adult MBSs, finding that lower pulse/frame rates compromised the detection of key swallowing events, such as PAS and MBSImP, thereby reducing diagnostic precision(12). Layly et al. (13) reported an ‘almost perfect’ agreement (Cohen’s Kappa of 0.95) between 30 and 15 FPS in a pediatric cohort involving 32 consecutive patients for rating PAS. These studies indicate that while some diagnostic metrics like PAS may not significantly suffer at modestly reduced frame rates in small samples, the broader implications of substantial frame reduction (e.g., to 7.5 FPS) are evident. Furthermore, the effects of frame rate reduction on DIGEST grading remain to date unexplored. This underscores the need for more research to determine the extent to which frame rate reductions can be implemented without compromising the clinical utility of MBS studies.

Our research question was: how does the pulse/frame rate of videofluoroscopic imaging affect Safety, Efficiency, and DIGEST grades in MBS studies? We compared MBS videos at the standard 30 PPS/FPS to those simulated at reduced rates of 15 and 7.5 PPS/FPS. We anticipated that results would add to the evidence base for selecting an optimal pulse rate that balances the need for effective diagnostic imagery with the minimization of patient radiation exposure.

2. Method

2.1. Data and Patients

This research was conducted as a retrospective analysis of Modified Barium Swallow (MBS) studies collected from patients examined at the University of Texas MD Anderson Cancer Center from 2016 to 2021. The protocol for this study received approval from the institutional review board (IRB number PA19–0261). All included patients underwent MBS examinations as part of their standard healthcare procedures. The MBS exams used in this study were originally recorded at 30 FPS for clinical rating. The fluoroscopy videos were generated by Artis Zee 90016 (Siemens™, Munich, German). The videos were captured and encoded by TIMS MVP (TIMS Medical™, Chelmsford, MA, USA), For each MBS exam, the standard bolus protocol included 11 individual bolus trials; 2 trials each of 5-mL, 10-mL, and self-administered cup sip volumes of thin liquid barium (Varibar^®, Bracco Diagnostics Inc., Monroe Township, New Jersey, USA), barium pudding (Varibar^®, Bracco Diagnostics Inc., Monroe Township, New Jersey, USA), and cracker coated in barium paste in the lateral plane plus AP thin liquid trial. Each MBS exam was assessed by trained, certified Speech-Language Pathologists (SLPs), who graded swallowing function across three parameters and documented the results in the electronic health record (EHR) using structured, custom-built flowsheets in Epic (Epic Systems Corporation, Verona, WI, USA). Flowsheet data were extracted and transferred to an electronic data capture platform (REDCap, Vanderbilt University(14)): Safety, Efficiency, and DIGEST, each graded on a 5-point scale from 0 to 4.

Our original dataset comprised 13,056 MBS exams conducted between 2011 and 2016 as routine healthcare procedures retrospectively reviewed under the same Institutional Review Board (IRB) protocol (PA19–0261). Upon cross-referencing the clinical datasets with IRB-approved research protocol databases, we identified 1,741 MBS exams that had been rated for DIGEST by both clinicians and laboratory raters at 30 FPS. Among these, 1,188 exams achieved exact concordance in Safety, Efficiency, and DIGEST ratings at 30 FPS. From these 1,188 MBS, a random sample was generated using Python’s random function to form an analytic dataset. To ensure a diverse representation of swallowing functions according to the DIGEST, MBS with DIGEST grades of 0, 1, 2, and 3 were proportionally sampled, while all MBS with a DIGEST grade of 4 were all included due to the rarity and clinical significance of severe swallowing dysfunction. This sampling schema yielded an analysis dataset of 315 MBS exams from 278 patients. The details of the patient demography and MBS are outlined in Table 1. A subset of patients has multiple MBSS records included in the dataset. Specifically, 244 patients have a single MBS, 31 patients have two MBSs, and 3 patients have three MBSs. For those patients with multiple exams, the shortest time interval between two MBSSs was 35 days.

Table 1.

Demographic and Clinical Characteristics of the Study Population and Distribution of DIGEST Grades Among MBS exams (n=315 MBS examinations from 278 patients)

Age	Mean (Standard Deviation)	62.51 (8.53)
Gender	Female	34 (12.23%)
	Male	244 (87.77%)
Primary diagnosis category (Patient numbers)	HNC	212 (76.25%)
	Multiple cancers, including HNC	18 (6.47%)
	Blood cancer	3 (1.08%)
	Metastatic cancer	4 (1.44%)
	Thoracic cancer	1 (0.03%)
	Multiple cancers, excluding HNC	2 (0.07%)
	Other solid tumors	7 (2.52%)
	other non-cancerous diseases	31 (11.15%)
DIGEST (MBS numbers)	0	80 (25.40%)
	1	75 (23.81%)
	2	74 (23.49%)
	3	79 (25.08%)
	4	7 (2.22%)

2.2. Video simulation for frame reduction

To investigate the impact of reduced pulse rates on diagnostic accuracy, the original MBS videos recorded at 30 FPS were processed to generate simulated videos at 15 and 7.5 FPS. Every second frame was removed for the 15 FPS simulations, and the first three frames of every four were removed for the 7.5 FPS simulations. After frame removal, the videos were re-encoded at the new frame rates to ensure they retained the same playback duration as the original recordings. This frame reduction and video re-encoding was implemented using Python scripts (version 3.9.16) with OpenCV library (version 4.8.0). Each simulated video was manually verified to prevent potential errors or bugs from automated processing, and ensure no disruptive artifacts were introduced.

2.3. DIGEST rating for simulated MBS.

A total of 630 reduced-frame-rate MBS videos were created, comprising 315 MBS videos at 15 FPS and 315 MBS at 7.5 FPS. All simulated videos were shuffled and anonymized to prepare for blind review. Three independent raters, all experienced in interpreting MBS studies and rating DIGEST, were enlisted to review the anonymized videos. Each rater was tasked with independently assessing and grading the MBS for Safety, Efficiency, and DIGEST according to standard laboratory procedures, blinded to research question and examination details.

2.4. Statistical analysis

To evaluate the effect of different simulated pulse rate s on the grading of DIGEST, we first summarized the exact agreement percent for pairwise comparisons between each frame rate pair (30 FPS vs. 15 FPS, 30 FPS vs. 7.5 FPS, and 15 FPS vs. 7.5 FPS). We also calculated weighted Cohen’s kappa coefficient with radical weights (k_w) and corresponding 95% confidence intervals (95% CI). The strength of agreement was interpreted using the following levels: zero as ‘poor’, 0.0–0.20 as ‘slight’, 0.21–0.40 as ‘fair’, 0.41–60 as ‘moderate’, 0.61–0.80 as ‘substantial’, and 0.81–1.0 as ‘almost perfect’(15). Because contingency tables had imbalanced marginals, we also computed a prevalence- and bias-adjusted kappa (PABAK)(16) and Gwet’s first-order agreement coefficient (AC1) (17), and 95% CI for each comparison. Our statistical analysis was performed using SAS version 9.4 (SAS Institute Inc®, Cary, North Carolina, USA).

Subsequent to the full dataset analysis, we also performed a subgroup assessment by excluding patients with a DIGEST score of 0 at 30 FPS (315–80 = 235 MBSs). We considered that patients with a DIGEST score of 1–4, indicating swallowing impairments, may provide more meaningful insights when frame rates are reduced, since the MBS with a DIGEST of 0 by definition do not have higher grade PAS or residue events that would go undetected with down-sampling the temporal resolution. As such, we then analyzed the same metrics, including Agreement Percentage (%), Weighted Cohen’s Kappa, PABAK, and AC1, across different frame rates for this subset. Additionally, we performed a sensitivity analysis to evaluate the potential impact of including only one MBSS per patient, and the details and results were summarized in the Supplementary Information document.

3. Results

Exact agreement of ratings for Safety, Efficiency, and DIGEST at different frame rates(simulated pulse rates) exceeded 85% across all iterations (30 FPS, 15 FPS, and 7.5 FPS) as shown in Figure 1 and Table 2. Each sub-figure of Figure 1 represents the comparison between two frame rates: 30 FPS vs. 15 FPS, 30 FPS vs. 7.5 FPS, and 15 FPS vs. 7.5 FPS.

Fig. 1.

Contingency matrixes of agreement for Safety, Efficiency, and DIGEST across different frame rates.

Table 2.

Agreement analysis of Safety, Efficiency, and DIGEST grades at different frame rates

Grade	FPS	No. of disagreements out of 315 pairs	Over estimation (%)	Under estimation (%)	Agreement percentage (%)	Simple Cohen’s Kappa (95% CI)	Weighted Cohen’s Kappa 95% CI)	PABAK (95% CI)	AC1 (95% CI)
Safety	30 vs. 15	40	12 (3.81)	28 (8.89)	87.30	0.819 (0.767, 0.870)	0.860 (0.788, 0.932)	0.841 (0.795, 0.887)	0.846 (0.801, 0.891)
	30 vs. 7.5	38	13 (4.13)	25 (7.94)	87.94	0.827 (0.777, 0.878)	0.870 (0.798, 0.942)	0.849 (0.804, 0.894)	0.854 (0.810, 0.898)
	15 vs. 7.5	39	22 (6.98)	17 (5.40)	87.62	0.821 (0.769, 0.873)	0.858 (0.785, 0.931)	0.845 (0.800, 0.891)	0.850 (0.806, 0.895)
Efficiency	30 vs. 15	40	25 (7.94)	15 (4.76)	87.30	0.823 (0.773, 0.873)	0.859 (0.785, 0.933)	0.841 (0.795, 0.887)	0.845 (0.800, 0.890)
	30 vs. 7.5	44	32 (10.16)	12 (3.81)	86.03	0.805 (0.753, 0.858)	0.850 (0.776, 0.923)	0.825 (0.778, 0.873)	0.830 (0.783, 0.877)
	15 vs. 7.5	43	26 (8.25)	17 (5.40)	86.35	0.811 (0.759, 0.862)	0.850 (0.777, 0.923)	0.829 (0.782, 0.877)	0.830 (0.783, 0.877)
DIGEST	30 vs. 15	47	21 (6.67)	26 (8.25)	85.08	0.804 (0.752, 0.855)	0.844 (0.777, 0.910)	0.814 (0.764, 0.863)	0.816 (0.767, 0.864)
	30 vs. 7.5	42	25 (7.94)	17 (5.40)	86.67	0.825 (0.775, 0.874)	0.865 (0.797, 0.930)	0.833 (0.786, 0.880)	0.835 (0.789, 0.882)
	15 vs. 7.5	41	26 (8.25)	15 (4.76)	86.98	0.829 (0.780, 0.877)	0.863 (0.796, 0.930)	0.837 (0.791, 0.884)	0.839 (0.794, 0.885)

The values in the cells indicate the number of cases rated with the corresponding grade. There were 42 disagreements for both the 30 FPS vs. 7.5 FPS comparison, and 41 disagreements for the 15 FPS vs. 7.5 FPS comparison, indicating only a minor disagreement difference among the comparisons. The Safety grades and the Efficiency grades showed similar results with the DIGEST comparisons, and there was a high level of agreement ranging from 85.08% to 87.94%.

The Weighted Cohen’s Kappa values show a high level of agreement across all comparisons, ranging from 0.850 to 0.864, and all the lower limits of the confidence intervals are above 0.77, indicating that agreement was in the ‘almost perfect’ range for each DIGEST grade (S, E, D) over all measure FPS conditions. The reliability of ratings remains consistently high across different frame rates, with all values falling within the ‘almost perfect’ range (0.81–1.0). The Kappa values varied slightly across the three comparisons: for the Safety grade the Kappa values varied by 0.012 (0.870 – 0.858); for the Efficiency grade, the Kappa values varied by 0.009 (0.859 – 0.850); for the DIGEST grade, the Kappa values showed a variation of 0.020 (0.864 – 0.844). The PABAK and AC1 also demonstrated high levels of agreement across all grades and frame rate comparisons. The PABAK values ranged from 0.8135 to 0.8492, and the AC1 values ranged from 0.8158 to 0.8539.

For Safety scores, the underestimation modestly exceeds overestimation when reducing from 30 FPS to 15 FPS (8.89% vs. 3.81%) and 7.5 FPS (7.94% vs. 4.13%), though this trend diminishes from 15 FPS to 7.5 FPS where overestimation is more equivalent (5.40% vs. 6.98%); in Efficiency scores, the trend is more pronounced, with overestimation consistently higher than underestimation across all FPS reductions (e.g., 7.94% vs. 4.76%, 10.16% vs. 3.81%, and 8.25% vs. 5.40%). In overall DIGEST scores, the error patterns displayed less distinct trends.

The results of the analysis excluding patients with a DIGEST score of 0 at 30 FPS are shown in Table 3. The agreement percentages of the Safety grades, Efficiency grades, and the DIGEST are all above 80% (ranging from 82.13% to 85.11%). The Weighted Cohen’s Kappa values range from 0.778 to 0.837, and almost all DIGEST grades are clustered around the boundary between the ‘substantial’ range (0.61–0.80) and the ‘almost perfect’ range (0.81–1.0).

Table 3.

Agreement analysis of Safety, Efficiency, and DIGEST grades at different frame rates (excluding MBSs with DIGEST = 0 at 30 FPS)

Grade	FPS	No. of disagreements	Agreement percentage (%)	Simple Cohen’s Kappa (95% CI)	Weighted Cohen’s Kappa (95% CI)	PABAK (95% CI)	AC1 (95% CI)
Safety	30 vs. 15	39	83.40	0.780 (0.715, 0.837)	0.828 (0.777, 0.878)	0.793 (0.734, 0.851)	0.795 (0.736, 0.854)
	30 vs. 7.5	37	84.26	0.792 (0.730, 0.852)	0.837 (0.788, 0.886)	0.803 (0.745, 0.856)	0.806 (0.748, 0.864)
	15 vs. 7.5	37	84.26	0.791 (0.729, 0.848)	0.831 (0.779, 0.882)	0.803 (0.745, 0.856)	0.806 (0.748, 0.864)
Efficiency	30 vs. 15	35	85.11	0.779 (0.709, 0.843)	0.821 (0.766, 0.876)	0.814 (0.755, 0.867)	0.821 (0.765, 0.876)
	30 vs. 7.5	40	82.98	0.745 (0.676, 0.807)	0.796 (0.738, 0.853)	0.787 (0.729, 0.846)	0.796 (0.737, 0.854)
	15 vs. 7.5	39	83.40	0.752 (0.677, 0.811)	0.799 (0.741, 0.857)	0.793 (0.729, 0.851)	0.801 (0.743, 0.859)
DIGEST	30 vs. 15	42	82.13	0.741 (0.669, 0.808)	0.778 (0.717, 0.838)	0.777 (0.713, 0.835)	0.784 (0.724, 0.844)
	30 vs. 7.5	37	84.26	0.771 (0.701, 0.832)	0.805 (0.747, 0.863)	0.803 (0.745, 0.862)	0.81 (0.753, 0.867)
	15 vs. 7.5	36	84.68	0.779 (0.714, 0.841)	0.810 (0.752, 0.867)	0.809 (0.750, 0.862)	0.815 (0.759, 0.871)

Compared with the analysis that included all the MBSs (Table 4), both the agreement percentages and the Weighted Cohen’s Kappa values show a certain degree of decline. For the agreement percentage, the decrease in subset grading across different frame rates is no more than 5% (ranging from 2.51% to 4.47%). The decrease in the Weighted Cohen’s Kappa values is slightly larger, with the maximum decline being only 7.82% (DIGEST at 30 FPS vs. 15 FPS).

Table 4.

Comparison of agreement analysis between All MBSs and MBS Excluding DIGEST = 0 @30 FPS

Grade	FPS	Agreement percentage (%)			Weighted Cohen’s Kappa
		All MBSs	MBSs with DIGEST 1–4 @30 FPS	Difference (%)	All MBSs	MBSs with DIGEST 1–4 @30 FPS	Difference (%)
Safety	30 vs. 15	87.30	83.40	3.90 (4.47)	0.860	0.828	0.032 (3.72)
	30 vs. 7.5	87.94	84.26	3.68 (4.18)	0.870	0.837	0.033 (3.79)
	15 vs. 7.5	87.62	84.26	3.36 (3.83)	0.858	0.831	0.027 (3.15)
Efficiency	30 vs. 15	87.30	85.11	2.19 (2.51)	0.859	0.821	0.038 (4.42)
	30 vs. 7.5	86.03	82.98	3.05 (3.55)	0.850	0.796	0.054 (6.35)
	15 vs. 7.5	86.35	83.40	2.95 (3.42)	0.850	0.799	0.051 (6.00)
DIGEST	30 vs. 15	85.08	82.13	2.95 (3.47)	0.844	0.778	0.066 (7.82)
	30 vs. 7.5	86.67	84.26	2.41 (2.78)	0.865	0.805	0.060 (6.94)
	15 vs. 7.5	86.98	84.68	2.30 (2.64)	0.863	0.810	0.053 (6.14)

4. Discussion

This study evaluated the relationship between reduced frame rates (simulated pulse rate) and the rating of MBS assessments, specifically examining Safety, Efficiency, and DIGEST grades. The findings of this study demonstrate high agreement levels in the Safety, Efficiency, and DIGEST grades across different frame rates (30 FPS, 15 FPS, and 7.5 FPS), indicating the robustness and stability of DIGEST across diverse imaging settings. While we expected that agreement might attenuate with larger degrees of frame rate reduction, the percent agreement and weighted Cohen’s kappa values exhibited only slight variations across the different frame rates. These study results imply that it is possible to apply DIGEST in lower pulse or frame rate MBS collection scenarios. The observed Cohen’s kappa of 0.844 to 0.870 in this study, while indicative of ‘almost perfect’ agreement, reflects a 15–16% discrepancy in DIGEST grading potentially influenced by pulse/frame rate variations. It is important to note that the observed less-than-perfect agreement may not be solely attributable to changes in pulse/frame rate. Our previous research on DIGEST reliability reported an inter-rater weighted kappa of 0.67 and an intra-rater weighted kappa ranging from 0.82 to 0.84, suggesting that some variability in rater agreement is inherent in human raters interpreting medical imaging data(18). While it was not possible to isolate the exact contribution of inter-/intra-rater variability from variability due to pulse/frame rate in this work, the observed agreement in almost perfect range across raters and diverse FPS conditions supports DIGEST robustness across practice variations.

As part of a DIGEST dissemination and implementation (D&I) project (R01CA271223), this study was designed to examine the scalability of the tool to diverse imaging acquisition protocols. Scalability refers to the ability of a tool that has been shown effective in ideal conditions to maintain effectiveness in real-world conditions (19). These results suggest that DIGEST is scalable in diverse imaging acquisition protocols and can be applied reliably in clinical or research settings that do not have access to the field standard of 30 FPS.

DIGEST provides an overall or composite score aggregating events that occur across the entire videofluoroscopic examination. The nature or structure of the DIGEST likely explains our frame rate results. The Safety grade measurement process primarily relies on the PAS, an 8-scale metric, for each swallowing trial. Previous studies have analyzed that reducing the frame rate from 30 FPS to 15 FPS and 7.5 FPS has a limited impact on PAS ratings (5, 11). This may explain the high agreement observed in the Safety grade across different frame rates. Identifying penetration/aspiration relies on accurately determining the frame that marks the initiation of the pharyngeal swallow. Differences in temporal resolution in MBS videos may compromise this process by omitting critical frames. More importantly, the design of the Safety grade enhances its robustness against variations in PAS ratings: the rating process utilizes bins of the PAS, specifically grouping PAS scores into four categories: PAS 1–2, 3–4, 5–6, and 7–8. While a reduced frame rate might cause a subtle difference in PAS, such as misinterpretation of a PAS 1 in 30 FPS to PAS 2 in 15 FPS, but this difference would not affect the final Safety grade, as both scores fall within the same bin. Consequently, the essential visual information required for Safety grade assessments remains intact, ensuring consistent ratings across different frame rates. In cases of disagreement, we observed that reducing the frame rate from 30 FPS to 15 FPS and 7.5 FPS results in a higher underestimation error compared to overestimation, suggesting a greater likelihood of missing events than over-detecting them; however, this trend becomes less pronounced when the frame rate decreases from 15 FPS to 7.5 FPS.

The Efficiency grade primarily relies on the percentage of pharyngeal residue present for each swallowing trial and the summarization of the maximum level of residue. Lower frame rates might miss subtle, brief movements of residue being cleared or accumulating. However, the presence of residue is typically a more sustained event compared to brief penetration or aspiration events. Therefore, even at lower frame rates, the overall amount and location of residue remain detectable. The other factor (modifier) in rating Efficiency is the bolus type (liquid, pudding, and/or cracker), but this information is independent of the frame rate. In cases of disagreement, we unexpectedly observed a higher likelihood of overestimation across all frame rate comparisons. This may be due to the video appearing jumpy or laggy at lower frame rates, causing raters to perceive the swallowing process as less smooth and, consequently, infer greater pharyngeal residue, which led to an overestimation of the Efficiency score.

Our study primarily focused on HNC patients who had undergone radiotherapy, making them particularly vulnerable to radiation-induced dysphagia. These patients often require frequent MBS evaluations due to the risk of aspiration and impaired pharyngeal clearance. Notably, non-diagnostic bolus trials, such as liquid washes or therapeutic swallowing techniques, are commonly required and thus increase radiation exposure, but they do not affect DIGEST grading accuracy. Reducing pulse rate to decrease radiation is a viable solution, though it resulted in a slight decrease in Cohen’s kappa to 0.844 to 0.870. This kappa value range still indicates substantial agreement, which is clinically acceptable. The agreement rates (85.08% to 87.94%) reflect high reliability, suggesting that the trade-off between reduced radiation and diagnostic precision is acceptable.

Physiologically, critical swallowing events such as bolus transit and airway protection mechanisms may not be adequately captured at lower frame rates, leading to potential misinterpretation of the mechanisms underlying swallowing safety and efficiency. Temporal measures, such as the duration of swallowing phases and the timing of events like pharyngeal transit time, pharyngeal delay time, and duration of upper esophageal sphincter opening, can also be distorted when frame rates are decreased(20). This distortion may hinder the accurate assessment of swallowing dynamics, as the precise timing of events is crucial for understanding the swallowing process and characterizing dysphagia(21).

The DIGEST framework is designed to evaluate swallowing function with safety and efficiency, rather than to delineate specific swallowing physiological impairments. In contrast to the other measurement, such as MBSImP, which offers detailed assessments of physiological swallowing components, DIGEST adopts a higher-level summary of safety and efficiency. This study conducted the first investigation into the effects of pulse/frame rate on the grading of DIGEST outcomes. It uniquely incorporates efficiency metric, an aspect that is absent from prior studies examining pulse/frame rate in relation to other swallowing measurement tools. While the PAS determines the safety score of DIGEST, it extends beyond existing research by integrating PAS bins and a frequency modifier. Consequently, there is truly no direct comparison between this study and preceding analyses of pulse/frame rate in alternative swallowing assessment frameworks.

In our subgroup analysis excluding DIGEST = 0 samples at 30 FPS and comparing the grading results for DIGEST = 1–4 across different frame rates, we observed a slight decrease in grading consistency. This is likely because DIGEST = 0 cases have minimal penetration-aspiration and residue, making them less sensitive to frame rate reductions. Even when restricting the analysis to more impaired MBS (DIGEST >0), the decrease in agreement was limited, with the largest drop in Weighted Kappa not exceeding 10% and all agreement levels remained within the “substantial” to “almost perfect” range (0.778–0.837). Thus, our interpretation remains that DIGEST grades maintain acceptable agreement even when taken in patients with impairment (DIGEST >0) in clinical environments where MBS are acquired at <30 fps.

Limitations

Despite the promising results, several limitations must be acknowledged. First, our sample of DIGEST grades is imbalanced; each category of DIGEST 0–3 has 70–80 cases, but DIGEST 4 has only 7 cases (correspondingly, Safety grade 4 has only 4 cases and Efficiency grade 4 has only 5 cases). This is because grade 4 (profound/life threatening) cases were rare in our study sample. As shown in Figure 1, there were proportionately more disagreements in grade 4 examinations. However, because they were under-represented, we cannot conclude whether score 4 was more sensitive to frame rate reduction. Also, our study only investigated DIGEST and its two components. However, a complete MBS exam evaluation of swallowing function requires many other metrics, such as the MBSImP with multiple components. Current studies on the impact of pulse/frame rate reduction on these parameters are still limited. Therefore, although this study provides valuable insights into the impact of frame rate reduction on the reliability of DIGEST in a large imaging dataset, it does not provide sufficient evidence to recommend changing the frame rate for the existing MBS exam. That is, while high levels of agreement were observed across different frame rates for our target metric DIGEST, the study’s findings do not support a clinical guideline for reducing pulse/frame rates to 15 FPS or 7.5 FPS in standard practice.

Another potential issue is that, with practical fluoroscopy systems, radiation dosage (detector entrance air KERMA [Kinetic Energy Released per Unit Mass] per frame) varies with pulse rate, potentially influencing image quality (6). In this study, the video frame rates were reduced retrospectively by removing frames from a 30FPS acquisition, which may not be strictly equivalent to lower pulse rate (e.g., 15 or 7.5 FPS) due to unadjusted radiation parameters. This represents a limitation of our methodology. Future research should also consider actual reductions in the pulse rate of fluoroscopy systems to determine whether image quality change further impact DIGEST grading outcomes. Additionally, although raters were blinded to frame/pulse rates of the videos, the visual presentation of frame speed variation—e.g., a simulated 7.5/15 FPS acquisition played at 30 FPS video setup with 0.133s/0.0667s duration per frame—may differ subtly from a native 30 FPS recording with 0.033s duration per frame. This could potentially compromise the blinding integrity. We notice that this might not substantially affect DIGEST grading, as raters were trained to assess swallowing metrics irrespective of playback characteristics, and not informed of any video modifications prior to the study. Nevertheless, this remains a methodological consideration for future studies.

Conclusion

This study compared the rating results of MBS recorded at 30 FPS with their down-sampled 15 FPS and 7.5 FPS duplicates to assess the impact of reduced frame rates on overall DIGEST grades and grades of the Safety and Efficiency DIGEST components. At varying FPS, results demonstrated high levels of exact agreement (>85%) and kw reached ‘almost perfect’ level. This suggests that reducing the frame rate from 30 FPS to 15 FPS or 7.5 FPS does not significantly degrade the DIGEST scoring process for MBS evaluations and that this metric (DIGEST) may be used in settings where MBS are conducted at lower frame rates without impacting clinical assessment accuracy. However, additional swallowing parameters of importance are susceptible to frame rate reduction and must be considered jointly when selecting optimal frame rate for a given clinical population. Any adjustments should prioritize optimizing other examination parameters while maintaining the highest feasible temporal resolution on a given clinical unit.

Data availability statement:

In accordance with the Final NIH Policy for Data Management and Sharing, NOT-OD-21–013, anonymized tabular analytic data that support the findings of this study are openly available in an NIH-supported generalist scientific data repository (figshare) at https://figshare.com/s/693fa1d301f244812aa7.