Respiratory motion arises from an integrated neurophysiological loop involving brainstem rhythm-generating circuits [18] and cortical networks that mediate volitional modulation of breathing [19]. Sensory feedback from pulmonary and thoracic mechanoreceptors is rapidly processed at the cortical level, as evidenced by respiratory-related evoked potentials detected with high-density EEG [20]. Furthermore, respiration interacts dynamically with large-scale cortical oscillatory networks in a phase-dependent manner [21], and interoceptive processing shapes individual variability in respiratory control [22]. Within this neurophysiological framework, visual biofeedback has been shown to enhance breath-hold consistency during radiotherapy [23]. Visual cues typically provide continuous, actively monitored information that facilitates moment-to-moment self-adjustment, whereas auditory guidance tends to be intermittent and passively received, potentially engaging a distinct breath-hold control mechanism [24]. Given the potential variability in individual patients’ aptitude for different feedback modalities, a respiratory indicator system that incorporates both approaches in a complementary manner is likely to offer greater clinical utility. In this context, the Abches system is a uniquely capable device, as it already integrates both functions within a single platform. In our previous study, we reported that there was no correlation between auditory and visual guidance in terms of breath-hold reproducibility on a per-patient basis, suggesting that the optimal guidance method should be determined individually [15]. However, a major limitation of that study was the fixed order in which the guidance methods were applied. In the present study, we addressed this limitation by including a new cohort in which the order of guidance was reversed, thereby enabling a more balanced comparison. The results were largely consistent with those of the previous study, further reinforcing the reliability of our earlier findings.

The utility of audio-visual biofeedback in respiratory-gated radiotherapy has been explored in several previous studies [23, 25]. George et al. investigated the effects of combined audio-visual feedback on respiratory-gated treatment and demonstrated that it effectively reduced intra-gating respiratory motion, as measured by the standard deviation of respiratory signals within the gating window [23]. Venkat et al. compared two types of visual feedback displays—a bar model and a wave model—and found that the wave display achieved over 50% and 70% reductions in displacement and period variability, respectively, compared with free breathing [25]. Similarly, Nakajima et al. assessed visual and audio-visual feedback using the Abches system in healthy volunteers and reported that both feedback modes improved respiratory stability compared with no feedback; however, direct comparisons between audio and visual guidance were not performed [26]. While these studies demonstrated the value of biofeedback for respiratory control, most focused on free-breathing or respiratory-gated treatments rather than voluntary breath-hold. Differences in study design, feedback methods, and monitoring systems also make direct comparison difficult. The present study extends previous work by directly comparing auditory and visual guidance during breath-hold in a clinical setting.

Our current study revealed no significant differences of breath-hold reproducibility between preferences of guidance methods. In this study, 55% of patients (22/40) preferred visual guidance, 30% (12/40) preferred auditory guidance, and 15% (6/40) reported no preference. As there was no significant difference in breath-hold reproducibility between the visual and auditory methods, visual coaching appears to be the most appropriate option for approximately 70% of patients. However, considering that patients with visual impairments and roughly 30% of all participants preferred auditory feedback, allowing patient-specific selection may facilitate better continuity and adherence to breath-hold radiotherapy.

Although the geometric differences observed in this study are small and, based on prior SBRT simulation and phantom studies, are expected to translate into only modest changes—typically ≤ 1% per millimeter in volume-based target dose metrics—across most clinical scenarios [27,28,29]. However, regions with very steep dose gradients, such as PTV–OAR interfaces, can still exhibit voxel-level dose differences on the order of 1–2 Gy per millimeter of displacement. Accordingly, even submillimeter to millimeter-scale uncertainties should be managed carefully when high-gradient regions are adjacent to critical organs.

Recent systematic reviews have also highlighted the considerable inter-patient variation in breath-hold stability and reproducibility, particularly in upper abdominal radiotherapy. Farrugia et al. summarized data from 41 studies and reported a median cranio-caudal inter-fraction reproducibility of 0.6 mm in exhale breath-hold and greater variability in DIBH [30]. Our findings were largely consistent with these results. However, they noted that no single technique emerged as clearly superior, as the reported confidence intervals overlapped and the study methodologies differed in breath-hold definitions, imaging modalities, surrogate markers, and analysis criteria [30].

Although audio guidance was more frequently preferred by patients, preference did not consistently correlate with improved breath-hold reproducibility. Guidance selection based on post-CT image analysis yielded slightly better reproducibility than patient preference (mean difference < 0.5 mm per axis). While the magnitude of this difference is small and unlikely to have a major clinical impact in most conventional settings, it highlights the potential importance of objective evaluation using CT images over subjective preference when selecting guidance modes. As emphasized in AAPM TG-76, patient engagement and training are essential for effective respiratory management [31]. Accordingly, we advocate for a patient-centered approach, wherein the guidance modality is tailored based on both imaging feedback and individual patient characteristics.

There are several limitations to the present study. First, there were notable differences between the 2009 and 2022 cohorts, with the latter demonstrating improved accuracy (Table 3). Several clearly identifiable technological advancements likely contributed to this improvement. One factor was the enhancement of CT image quality: the Hi Speed DX/i used in 2009 employed a single–detector-row system, whereas the Canon Aquilion LB used in 2022 had a 16-detector-row configuration, enabling more precise respiratory positional measurements. In addition, improvements in the operability of the Abches control software, along with nearly a decade of accumulated staff experience with Abches-based breath-hold techniques, may have further contributed to the superior performance observed in the 2022 cohort.

Second, the study included a mix of inhalation and exhalation breath-hold techniques, with inhalation cases being predominant. For example, Kimura et al. reported that the end-expiration phase was more stable than the end-inspiration phase with respect to breath-hold reproducibility [32]. Therefore, further investigation focusing on exhalation breath-holds with a larger sample size is warranted. Third, regarding the auditory–visual system evaluated in this study, the optimal approach must be tailored to each patient, and therefore it does not inherently improve efficiency. Nonetheless, our findings show that both techniques can achieve acceptable breath-hold accuracy. Accordingly, for patients who cannot use one modality (e.g., those who are totally blind or have hearing impairment), the other can serve as a practical alternative. Devices that incorporate both modalities remain limited, and we believe that this dual functionality offers meaningful clinical benefit.