Assessing nocturnal scratch with actigraphy in atopic dermatitis patients

Abstract

Nocturnal scratch is one major factor leading to impaired quality of life in atopic dermatitis (AD) patients. Therefore, objectively quantifying nocturnal scratch events aids in assessing the disease state, treatment effect, and AD patients’ quality of life. In this paper, we describe the use of actigraphy, highly predictive topological features, and a model-ensembling approach to develop an assessment of nocturnal scratch events by measuring scratch duration and intensity. Our assessment is tested in a clinical setting against the ground truth obtained from video recordings. The new approach addresses unmet challenges in existing studies, such as the lack of generalizability to real-world applications, the failure to capture finger scratches, and the limitations in the evaluation due to imbalanced data in the current literature. Furthermore, the performance evaluation shows agreement between derived digital endpoints and the video annotation ground truth, as well as patient-reported outcomes, which demonstrated the validity of the new assessment of nocturnal scratch.

Quantifying Nocturnal Scratch in Atopic Dermatitis: A Machine Learning Approach Using Digital Wrist Actigraphy

“…Furthermore, upon comparing the data from accelerometers and gyroscopes, it becomes apparent that accelerometer data deliver slightly better performance. This finding differs from a previously reported study [ 13 ], indicating the need for further exploration in the future. In addition, an accelerometer generally consumes less energy compared to a gyroscope, thereby conserving battery life when it is the only sensor used.…”

“…The results indicate that the feature engineering model exhibits superior performance in terms of the F1 score metric compared to the CNN model. Furthermore, our model achieves a comparable F1 score to the published work [13], but with less than half the number of subjects. In addition to the accelerometer data, the gyroscope data are utilized independently to train the feature engineering model.…”

“…In addressing the challenge of an unbalanced dataset in our study, we implemented a two-pronged strategy to enhance the robustness and accuracy of our machine learning models. Firstly, we redefined the classification problem from distinguishing between “scratching vs. non-scratching” actions to differentiating between “scratching movement vs. non-scratching movement” [ 13 , 15 ]. This approach significantly increased the ratio of positive samples, thereby enriching our dataset with more relevant positive samples for model training.…”

“…For instance, Ji et al employed a combination of commonly used time and frequency domain features, topological features, and deep learning extracted features. They trained a LightGBM machine to predict nocturnal scratches [ 13 ]. Similarly, Moreau et al utilized raw actigraphy data directly to train recurrent neural networks (RNN) and long short-term memory (LSTM) models for the same purpose [ 14 ].…”