峰值頻率直方圖指標於非腸胃道手術病患術後腸音評估之可行性探討

Abstract

腸音能夠反映腸道蠕動狀態，具有作為非侵入式生理指標的潛力。然而，傳統聽診法主觀性高且缺乏標準化，限制其臨床應用價值。本研究旨在探討以腸音作為生理指標的可行性，並建立一套客觀、自動化且具臨床實用性的術後腸音分析架構。共納入 14 位接受全身麻醉之非腸胃道手術病人，於圍手術期五個時間點，從四個標準化腹部位置記錄腸音，使用自行組裝之聽診器式麥克風進行收音。於時域方面，提取總腸音時長、事件發生次數與平均持續時間等特徵，並比較人工與自動分段演算法之結果；於頻域方面，則建立「峰值頻率直方圖（Peak Frequency Histogram, PFH）」方法，以視覺化不同時間點之主頻分佈情形。在所有分析指標中，針對 288–384 Hz 頻段所建立之 PFH 指標展現出最一致且具統計顯著性的術後回復趨勢，顯示其作為腸道功能臨床指標之潛力。本研究所提出之分析流程具備可行性、重現性與生理意義，為建立標準化腸音評估作業流程奠定基礎。未來若進一步整合系統，將有望實現即時智慧腸音監測，應用於術後照護、自律神經功能評估及腸胃道疾病處置等臨床場域。

Bowel sounds (BSs) reflect intestinal motility and hold promise as non-invasive physiological indicators. However, traditional auscultation methods are subjective and lack standardization, limiting their clinical utility. This study aimed to explore the feasibility of using BSs as physiological indicators and to develop an objective, automated, and clinically applicable framework for postoperative BS analysis. A total of 14 patients undergoing non-gastrointestinal surgery under general anesthesia were enrolled. BSs were recorded at five perioperative time points from four standardized abdominal sites using a self-built stethoscope microphone. Time-domain features—such as total sound duration, occurrence count, and average duration—were extracted using both manual and automated segmentation algorithms. In parallel, a frequency-domain method was established based on the Peak Frequency Histogram (PFH), which visualized the dominant frequency distribution across time. Among all indicators, a PFH-based index targeting the 288–384 Hz band showed the most consistent and statistically significant recovery trend postoperatively, suggesting its potential as a robust clinical marker of intestinal function. The analysis protocol demonstrated feasibility, reproducibility, and physiological interpretability, paving the way for a standardized operating procedure. Future system integration may enable real-time, intelligent bowel monitoring in postoperative care, autonomic function assessment, and gastrointestinal disorder management.