Dataset Information

Egocentric 3D Skeleton Learning in a Deep Neural Network Encodes Obese-like Motion Representations.

ABSTRACT: Obesity is a growing health concern, mainly caused by poor dietary habits. Yet, accurately tracking the diet and food intake of individuals with obesity is challenging. Although 3D motion capture technology is becoming increasingly important in healthcare, its potential for detecting early signs of obesity has not been fully explored. In this research, we used a deep LSTM network trained with individual identity (identity-trained deep LSTM network) to analyze 3D time-series skeleton data from mouse models with diet-induced obesity. First, we analyzed the data from two different viewpoints: allocentric and egocentric. Second, we trained various deep recurrent networks (e.g., RNN, GRU, LSTM) to predict the identity. Lastly, we tested whether these models effectively encode obese-like motion representations by training a support vector classifier with the latent features from the last layer. Our experimental results indicate that the optimal performance is achieved when utilizing an identity-trained deep LSTM network in conjunction with an egocentric viewpoint. This approach suggests a new way to use deep learning to spot health risks in mouse models of obesity and should be useful for detecting early signs of obesity in humans.

SUBMITTER: Kwon J

PROVIDER: S-EPMC11247279 | biostudies-literature | 2024 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Egocentric 3D Skeleton Learning in a Deep Neural Network Encodes Obese-like Motion Representations.

Kwon Jea J Sa Moonsun M Kim Hyewon H Seong Yejin Y Lee C Justin CJ

Experimental neurobiology 20240601 3

Obesity is a growing health concern, mainly caused by poor dietary habits. Yet, accurately tracking the diet and food intake of individuals with obesity is challenging. Although 3D motion capture technology is becoming increasingly important in healthcare, its potential for detecting early signs of obesity has not been fully explored. In this research, we used a deep LSTM network trained with individual identity (identity-trained deep LSTM network) to analyze 3D time-series skeleton data from mo ...[more]

PMID: 38993079

Similar Datasets

Project description:Background: Performance of wrist actigraphy in assessing sleep not only depends on the sensor technology of the actigraph hardware but also on the attributes of the interpretative algorithm (IA). The objective of our research was to improve assessment of sleep quality, relative to existing IAs, through development of a novel IA using deep learning methods, utilizing as input activity count and heart rate variability (HRV) metrics of different window length (number of epochs of data). Methods: Simultaneously recorded polysomnography (PSG) and wrist actigraphy data of 222 participants were utilized. Classic deep learning models were applied to: (a) activity count alone (without HRV), (b) activity count + HRV (30-s window), (c) activity count + HRV (3-min window), and (d) activity count + HRV (5-min window) to ascertain the best set of inputs. A novel deep learning model (Haghayegh Algorithm, HA), founded on best set of inputs, was developed, and its sleep scoring performance was then compared with the most popular University of California San Diego (UCSD) and Actiwatch proprietary IAs. Results: Activity count combined with HRV metrics calculated per 5-min window produced highest agreement with PSG. HA showed 84.5% accuracy (5.3-6.2% higher than comparator IAs), 89.5% sensitivity (6.2% higher than UCSD IA and 6% lower than Actiwatch proprietary IA), 70.0% specificity (8.2-34.3% higher than comparator IAs), and 58.7% Kappa agreement (16-23% higher than comparator IAs) in detecting sleep epochs. HA did not differ significantly from PSG in deriving sleep parameters-sleep efficiency, total sleep time, sleep onset latency, and wake after sleep onset; moreover, bias and mean absolute error of the HA model in estimating them was less than the comparator IAs. HA showed, respectively, 40.9% and 54.0% Kappa agreement with PSG in detecting rapid and non-rapid eye movement (REM and NREM) epochs. Conclusions: The HA model simultaneously incorporating activity count and HRV metrics calculated per 5-min window demonstrates significantly better sleep scoring performance than existing popular IAs.

Project description:In this study, we propose a novel markerless motion capture system (MCS) for monkeys, in which 3D surface images of monkeys were reconstructed by integrating data from four depth cameras, and a skeleton model of the monkey was fitted onto 3D images of monkeys in each frame of the video. To validate the MCS, first, estimated 3D positions of body parts were compared between the 3D MCS-assisted estimation and manual estimation based on visual inspection when a monkey performed a shuttling behavior in which it had to avoid obstacles in various positions. The mean estimation error of the positions of body parts (3-14 cm) and of head rotation (35-43°) between the 3D MCS-assisted and manual estimation were comparable to the errors between two different experimenters performing manual estimation. Furthermore, the MCS could identify specific monkey actions, and there was no false positive nor false negative detection of actions compared with those in manual estimation. Second, to check the reproducibility of MCS-assisted estimation, the same analyses of the above experiments were repeated by a different user. The estimation errors of positions of most body parts between the two experimenters were significantly smaller in the MCS-assisted estimation than in the manual estimation. Third, effects of methamphetamine (MAP) administration on the spontaneous behaviors of four monkeys were analyzed using the MCS. MAP significantly increased head movements, tended to decrease locomotion speed, and had no significant effect on total path length. The results were comparable to previous human clinical data. Furthermore, estimated data following MAP injection (total path length, walking speed, and speed of head rotation) correlated significantly between the two experimenters in the MCS-assisted estimation (r = 0.863 to 0.999). The results suggest that the presented MCS in monkeys is useful in investigating neural mechanisms underlying various psychiatric disorders and developing pharmacological interventions.

Project description:Intracranial hemorrhage (ICH) occurs when a blood vessel ruptures in the brain. This leads to significant morbidity and mortality, the likelihood of which is predicated on the size of the bleeding event. X-ray computed tomography (CT) scans allow clinicians and researchers to qualitatively and quantitatively diagnose hemorrhagic stroke, guide interventions and determine inclusion criteria of patients in clinical trials. There is no currently available open source, validated tool to quickly segment hemorrhage. Using an automated pipeline and 2D and 3D deep neural networks, we show that we can quickly and accurately estimate ICH volume with high agreement with time-consuming manual segmentation. The training and validation datasets include significant heterogeneity in terms of pathology, such as the presence of intraventricular (IVH) or subdural hemorrhages (SDH) as well as variable image acquisition parameters. We show that deep neural networks trained with an appropriate anatomic context in the network receptive field, can effectively perform ICH segmentation, but those without enough context will overestimate hemorrhage along the skull and around calcifications in the ventricular system. We trained with all data from a multi-center phase II study (n = 112) achieving a best mean and median Dice coefficient of 0.914 and 0.919, a volume correlation of 0.979 and an average volume difference of 1.7 ml and root mean squared error of 4.7 ml in 500 out-of-sample scans from the corresponding multi-center phase III study. 3D networks with appropriate anatomic context outperformed both 2D and random forest models. Our results suggest that deep neural network models, when carefully developed can be incorporated into the workflow of an ICH clinical trial series to quickly and accurately segment ICH, estimate total hemorrhage volume and minimize segmentation failures. The model, weights and scripts for deployment are located at https://github.com/msharrock/deepbleed . This is the first publicly available neural network model for segmentation of ICH, the only model evaluated with the presence of both IVH and SDH and the only model validated in the workflow of a series of clinical trials.

Dataset Information

Egocentric 3D Skeleton Learning in a Deep Neural Network Encodes Obese-like Motion Representations.

Publications

Egocentric 3D Skeleton Learning in a Deep Neural Network Encodes Obese-like Motion Representations.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets