Embedded Dictation Detection for Physician Encounter Audio
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately distinguish between dictation and non-dictation portions in audio recordings of physician-patient encounters, requiring physician input and leading to inefficiencies in scribing processes.
Innovation Solution
A system that utilizes ambient listening devices and machine learning algorithms to analyze speech patterns, proximity, and linguistic cues to automatically identify dictation segments, allowing for efficient navigation and transcription.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing systems rely on physician self-identification of dictation portions, then the system can capture all potential dictation segments, but the scribing efficiency decreases and physician workload increases
Solution Approach 1:
The system enables self-service by allowing the audio recording to automatically identify and segment dictation portions through analysis of acoustic features, speech patterns, and contextual cues, eliminating the need for physician intervention in the identification process
Solution Approach 2:
The patent replaces the mechanical process of manual physician identification with an automated computational system that uses machine learning models and audio analysis algorithms to detect and classify dictation segments automatically
2Measurement precision
If the system analyzes only limited dictation portions, then processing speed increases, but measurement precision of dictation boundaries deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the audio recording into distinct segments based on detected dictation boundaries, allowing the system to process and analyze each segment independently with appropriate precision while managing overall system complexity through modular analysis
Solution Approach 2:
The system changes parameters by dynamically adjusting analysis thresholds, window sizes, and feature extraction parameters based on the detected speech characteristics and contextual information, enabling adaptive precision without fixed complex configurations
3Productivity
If the scribe must listen to entire audio recordings linearly, then no information is missed, but the time required for transcription increases significantly
Solution Approach 1:
The system performs preliminary action by pre-identifying and marking all dictation segments before the scribe begins transcription, allowing the scribe to navigate directly to relevant portions rather than listening linearly through entire recordings
Solution Approach 2:
The patent introduces a temporal dimension by creating a non-linear navigation structure with time stamps and segment markers, allowing the scribe to jump between different portions of the recording based on the identified dictation segments rather than following linear playback
Data Source
AI summary
Systems and techniques for identifying dictation within an audio stream are disclosed including training a first neural network model stored in a computer memory that is configured to process audio data to identify one or more audio segments that are indicative of dictation without using automatic speech recognition, training a second neural network model stored in the computer memory that is configured to process audio data using automatic speech recognition, analyzing the stored audio data using the first neural network to identify one or more segments in the audio data that are indicative of dictation, and analyzing the one or more identified segments that are indicative of dictation to extract one or more features from the one or more identified segments.