Towards More Efficient Data Valuation in Healthcare Federated Learning using Ensembling.

Federated Learning (FL) wherein multiple institutions collaboratively train a machine learning model without sharing data is becoming popular. Participating institutions might not contribute equally - some contribute more data, some better quality data or some more diverse data. To fairly rank the contribution of different institutions, Shapley value (SV) has emerged as the method of choice. Exact SV computation is impossibly expensive, especially when there are hundreds of contributors. Existing SV computation techniques use approximations. However, in healthcare where the number of contributing institutions are likely not of a colossal scale, computing exact SVs is still exorbitantly expensive, but not impossible. For such settings, we propose an efficient SV computation technique called SaFE (Shapley Value for Federated Learning using Ensembling). We empirically show that SaFE computes values that are close to exact SVs, and that it performs better than current SV approximations. This is particularly relevant in medical imaging setting where widespread heterogeneity across institutions is rampant and fast accurate data valuation is required to determine the contribution of each participant in multi-institutional collaborative learning.

Heart rate variability can be used to estimate sleepiness-related decrements in psychomotor vigilance during total sleep deprivation.

STUDY OBJECTIVES: To assess whether changes in psychomotor vigilance during sleep deprivation can be estimated using heart rate variability (HRV). DESIGN: HRV, ocular, and electroencephalogram (EEG) measures were compared for their ability to predict lapses on the Psychomotor Vigilance Task (PVT). SETTING: Chronobiology and Sleep Laboratory, Duke-NUS Graduate Medical School Singapore. PARTICIPANTS: Twenty-four healthy Chinese men (mean age ± SD = 25.9 ± 2.8 years). INTERVENTIONS: Subjects were kept awake continuously for 40 hours under constant environmental conditions. Every 2 hours, subjects completed a 10-minute PVT to assess their ability to sustain visual attention. MEASUREMENTS AND RESULTS: During each PVT, we examined the electrocardiogram (ECG), EEG, and percentage of time that the eyes were closed (PERCLOS). Similar to EEG power density and PERCLOS measures, the time course of ECG RR-interval power density in the 0.02-0.08-Hz range correlated with the 40-hour profile of PVT lapses. Based on receiver operating characteristic curves, RR-interval power density performed as well as EEG power density at identifying a sleepiness-related increase in PVT lapses above threshold. RR-interval power density (0.02-0.08 Hz) also classified subject performance with sensitivity and specificity similar to that of PERCLOS. CONCLUSIONS: The ECG carries information about a person's vigilance state. Hence, HRV measures could potentially be used to predict when an individual is at increased risk of attentional failure. Our results suggest that HRV monitoring, either alone or in combination with other physiologic measures, could be incorporated into safety devices to warn drowsy operators when their performance is impaired.