Individual coactivation patterns improve the subject identification and their behavior association

Brain states can be characterized by recurring coactivation patterns (CAPs). Traditional CAP analysis is performed at the group-level, while the human brain is individualized and the functional connectome has shown the uniqueness as fingerprint. Whether stable individual CAPs could be obtained from a single fMRI scan and could individual CAPs improve the identification is unclear. An open dataset, the midnight scan club was used in this study to answer these questions. Four CAP states were identified at three distinct levels (group-, subject- and scan-level) separately, and the CAPs were then reconstructed for each scan. Identification rate and differential identifiability were used to evaluate the identification performance. Our results demonstrated that the individual CAPs were unstable when using a single scan. By maintaining high intra-subject similarity and inter-subject differences, subject-level CAPs achieved the best identification performance. Brain regions that contributed to the identifiability were mainly located in higher-order networks (e.g., frontal-parietal network). Besides, head motion reduced the intra-subject similarity, while its impact on identification rate was non-significant. Finally, a pipeline was developed to depict brain-behavior associations in dataset with few samples but dense sampling, and individualized CAP dynamics showed an above-chance level correlation with IQ.

Reliability and Generalization of Gait Biometrics Using 3D Inertial Sensor Data and 3D Optical System Trajectories

Particularities in the individuals’ style of walking have been explored for at least three decades as a biometric trait, fueling the automatic gait recognition field. Whereas, gait recognition works usually focus on improving end-to-end performance measures, and this work aims at understanding which individuals’ traces are more relevant to improve subjects’ separability. For such, a manifold projection technique and a multi-sensor gait dataset were adopted to investigate the impact of each data source characteristics on this separability. The assessments have shown it is hard to distinguish individuals based only on their walking patterns in a subject identification scenario. In this scenario, the subjects’ separability is more related to their physical characteristics than their movements related to gait cycles and biomechanical events. However, this study’s results also points to the feasibility of learning identity characteristics from individuals’ walking patterns learned from similarities and differences between subjects in a verification setup. The explorations concluded that periodic components occurring in frequencies between 6Hz and 10Hz are more significant for learning these patterns than events and other biomechanical movements related to the gait cycle, as usually explored in the literature.

Using Electronic Medical Records to Identify Potentially Eligible Study Subjects for Lung Cancer Screening with Biomarkers

Lung cancer screening trials using low-dose computed tomography (LDCT) show reduced late-stage diagnosis and mortality rates. These trials have identified high-risk groups that would benefit from screening. However, these sub-populations can be difficult to access and retain in trials. Implementation of national screening programmes further suggests that there is poor uptake in eligible populations. A new approach to participant selection may be more effective. Electronic medical records (EMRs) are a viable alternative to population-based or health registries, as they contain detailed clinical and demographic information. Trials have identified that e-screening using EMRs has improved trial retention and eligible subject identification. As such, this paper argues for greater use of EMRs in trial recruitment and screening programmes. Moreover, this opinion paper explores the current issues in and approaches to lung cancer screening, whether records can be used to identify eligible subjects for screening and the challenges that researchers face when using EMR data.

An Attention Based Hierarchical LSTM Architecture for ECG Biometric System

The electrocardiogram (ECG) based biometric sys-<br>tem has recently gained popularity. Easy signal acquisition and<br>robustness against falsification are the major advantages of the<br>ECG based biometric system. This biometric system can help<br>automate the subject identification and authentication aspect of<br>personalised healthcare services. In this paper, we have designed<br>a novel attention based hierarchical long short-term memory<br>(LSTM) model to learn the biometric representation correspond-<br>ing to a person. The hierarchical LSTM model proposed in this<br>paper can learn the temporal variation of the ECG signal in<br>different abstractions. This addresses the long term dependency<br>issue of the LSTM network in our application. The attention<br>mechanism of the model learns to capture the ECG complexes<br>that have more biometric information corresponding to each<br>person. These ECG complexes are given more weight to learn<br>better biometric representation. The proposed system is less<br>complex and more efficient as it does not require the detection<br>of any fiducial points. We have evaluated the proposed model for<br>both the person verification and identification problems using<br>two on-the-person ECG databases and two off-the-person ECG<br>databases. The proposed framework is found to perform better<br>than the existing fiducial and non-fiducial point based methods.<br>

An Attention Based Hierarchical LSTM Architecture for ECG Biometric System

The electrocardiogram (ECG) based biometric sys-<br>tem has recently gained popularity. Easy signal acquisition and<br>robustness against falsification are the major advantages of the<br>ECG based biometric system. This biometric system can help<br>automate the subject identification and authentication aspect of<br>personalised healthcare services. In this paper, we have designed<br>a novel attention based hierarchical long short-term memory<br>(LSTM) model to learn the biometric representation correspond-<br>ing to a person. The hierarchical LSTM model proposed in this<br>paper can learn the temporal variation of the ECG signal in<br>different abstractions. This addresses the long term dependency<br>issue of the LSTM network in our application. The attention<br>mechanism of the model learns to capture the ECG complexes<br>that have more biometric information corresponding to each<br>person. These ECG complexes are given more weight to learn<br>better biometric representation. The proposed system is less<br>complex and more efficient as it does not require the detection<br>of any fiducial points. We have evaluated the proposed model for<br>both the person verification and identification problems using<br>two on-the-person ECG databases and two off-the-person ECG<br>databases. The proposed framework is found to perform better<br>than the existing fiducial and non-fiducial point based methods.<br>

Subject matter identification by means of community outreach

The purpose of this practical thesis project was to create a guidebook for collecting subject-based information gathered through community participation and collaboration. Specifically, this involved collecting subject descriptions for a photograph collection based on an organized and planned meeting with local residents familiar with the contents of the images. All fieldwork was completed over a nine-week period, from June 5 through August 2, 2006, at the Bruce County Museum and Cultural Centre, Southampton, Ontario with the John H. Scougall Collection. 105 images were selected for discussion and subject identification by community members. What follows is a guidebook to inform others how to carry out such a project. Each section begins with general comments and ideas, followed by specific examples of what took place with the Scougall Collection and the participating residents of Kincardine.