Promoting fairness in activity recognition algorithms for patient’s monitoring and evaluation systems in healthcare

• Published in: Computers in biology and medicine Vol. 179; p. 108826
• Main Authors: Mennella, Ciro, Esposito, Massimo, De Pietro, Giuseppe, Maniscalco, Umberto
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.09.2024; Elsevier Limited
• Subjects: Accuracy; Adult; Aged; Algorithms; Artificial intelligence; Artificial neural networks; Bias; Data analysis; Deep learning; Female; Health care; Heterogeneity; Human Activities; Human activity recognition; Human influences; Human motion; Humans; Impact analysis; Internal Medicine; Machine learning; Male; Middle Aged; Motion analysis; Neural networks; Neural Networks, Computer; Other; Pattern recognition; Performance evaluation; Physical characteristics; Physical properties; Population studies; Populations; Rehabilitation; Robustness; State-of-the-art reviews; Time series; Uncertainty analysis; Deep learning; Time-series; Rehabilitation; Artificial intelligence; Motion analysis; Bias
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2024.108826

Researchers face the challenge of defining subject selection criteria when training algorithms for human activity recognition tasks. The ongoing uncertainty revolves around which characteristics should be considered to ensure algorithmic robustness across diverse populations. This study aims to address this challenge by conducting an analysis of heterogeneity in the training data to assess the impact of physical characteristics and soft-biometric attributes on activity recognition performance. The performance of various state-of-the-art deep neural network architectures (tCNN, hybrid-LSTM, Transformer model) processing time-series data using the IntelliRehab (IRDS) dataset was evaluated. By intentionally introducing bias into the training data based on human characteristics, the objective is to identify the characteristics that influence algorithms in motion analysis. Experimental findings reveal that the CNN-LSTM model achieved the highest accuracy, reaching 88%. Moreover, models trained on heterogeneous distributions of disability attributes exhibited notably higher accuracy, reaching 51%, compared to those not considering such factors, which scored an average of 33%. These evaluations underscore the significant influence of subjects’ characteristics on activity recognition performance, providing valuable insights into the algorithm’s robustness across diverse populations. This study represents a significant step forward in promoting fairness and trustworthiness in artificial intelligence by quantifying representation bias in multi-channel time-series activity recognition data within the healthcare domain. •Systems have emerged that provide real-time feedback on patient performance.•Bias in human activity models for healthcare applications must be investigated.•Equity and fairness must be ensured in AI models that impact human safety and health.