We are excited to share the MAISON-LLF Data Challenge (Version 4) as part of the ARIAL workshop. The MAISON-LLF dataset was collected from 18 older adult participants living alone in the community following lower limb fractures. The data was collected using the MAISON, a cloud based multimodal sensor network, which included a smart watch, a smart phone, a sleep mat and an indoor motion sensor aimed at collecting their movement, motion, sleep and physiological data. Each participant contributed data for 8 weeks each, beginning from their first-week post-discharge. This resulted in a total of 1008 days of continuous multimodal sensor data, complemented by biweekly clinical questionnaire data. Missing values were already imputed in the dataset at a weekly window. The anonymized dataset contains 84 features/columns, out of which:
4 features are participant ID and timestamps, e.g. participant', 'timestamp', and 'clinical-timestamp'.
46 numeric features
35 clinical scores
The details of the study protocol and features are presented in the dataset aper.
Different types of machine/deep learning problems can be formulated from this novel dataset, not limited to, classification, clustering, regression, temporal prediction, explainability, visualization, large language models, hypothesis generation etc.
We invite novel contributions that use MAISON-LLF dataset except for those published in the original dataset paper. We encourage researchers to push the enevlope in the context of developing new machine/deep learning algorithms, data processing pipelines and clinical relevance. The submitted papers (short or long) will be peer reviewed as regular submissions and will be part of the workshop proceedings. Ideally, the authors should also release their code as the data is already made publicly available to help advance further research in this area.
All the enqueries related to MAISON-LLF submission can be directed to Dr. Ali Abedi.