Scholarlink

<p>The Mutilead Electrocardiogram signal (MECG) provides a comprehensive clinical assessment of cardiac function, necessitating substantial data storage. This study introduces an on-device method for real-time compression of MECG data, utilizing optimal quantization through particle swarm optimization (PSO) to leverage multiple optimality criteria. The initial phase of compression involved the application of nonlinear principal component analysis to achieve dimensionality reduction to eliminate interlead recurrence. The subsequent application of the optimal decomposition level of the tunable Q-wavelet transform (TQWT) yields the minimal error while achieving the shortest length of the end-coefficient. The optimal quantization of end-coefficients was facilitated by a trained multilayer perceptron neural network (MLPNN), which had been previously trained using end-coefficients’ features and optimized outputs from PSO, introducing dynamic cost-function algorithm, incorporating a new quality index, named as ‘Compression_Score’. The analysis of results was conducted on 546 MECG records sourced from the ptbdb database, yielding a compression ratio (CR) of 80.67 and a percentage root mean squared difference (PRD) of 5.93 for beat-by-beat compression. When compressing sequences of beats consecutively, overall CR and PRD values obtained were 80.92 and 6.06 for 5 beats, 84.81 and 6.19 for 10 beats, and 90.66 and 6.45 for 15 beats, respectively. The data processing was executed on Raspberry-Pi utilizing the python programming platform, thereby implementing a fully hardware-based methodology within the field of MECG compression research. The enhanced performance and hardware integration facilitate a broad spectrum of applications within the medical field, demonstrating improved efficacy relative to earlier studies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On-device compression of multilead electrocardiogram using tunable-Q wavelet transform and MLPNN trained using multi-optima optimization based PSO

  • 期刊论文

  • 发表日期：

    2025年5月28日

查看全文
  • Niyasha Patra,
  • Soumyendu Banerjee,
  • Sanjay Bhadra

摘要

The Mutilead Electrocardiogram signal (MECG) provides a comprehensive clinical assessment of cardiac function, necessitating substantial data storage. This study introduces an on-device method for real-time compression of MECG data, utilizing optimal quantization through particle swarm optimization (PSO) to leverage multiple optimality criteria. The initial phase of compression involved the application of nonlinear principal component analysis to achieve dimensionality reduction to eliminate interlead recurrence. The subsequent application of the optimal decomposition level of the tunable Q-wavelet transform (TQWT) yields the minimal error while achieving the shortest length of the end-coefficient. The optimal quantization of end-coefficients was facilitated by a trained multilayer perceptron neural network (MLPNN), which had been previously trained using end-coefficients’ features and optimized outputs from PSO, introducing dynamic cost-function algorithm, incorporating a new quality index, named as ‘Compression_Score’. The analysis of results was conducted on 546 MECG records sourced from the ptbdb database, yielding a compression ratio (CR) of 80.67 and a percentage root mean squared difference (PRD) of 5.93 for beat-by-beat compression. When compressing sequences of beats consecutively, overall CR and PRD values obtained were 80.92 and 6.06 for 5 beats, 84.81 and 6.19 for 10 beats, and 90.66 and 6.45 for 15 beats, respectively. The data processing was executed on Raspberry-Pi utilizing the python programming platform, thereby implementing a fully hardware-based methodology within the field of MECG compression research. The enhanced performance and hardware integration facilitate a broad spectrum of applications within the medical field, demonstrating improved efficacy relative to earlier studies.