Ryan T. Barnard

Depth estimation, spatially variant image registration, and super-resolution using a multi-lenslet camera

With a multi-lenslet camera, we can capture multiple low resolution (LR) images of the same scene and use them to reconstruct a high resolution (HR) image. For this purpose, two major computation problems need to be solved, the image registration and the super resolution (SR) reconstruction. For the first, one major hurdle is the spatially variant shifts estimation, because objects in a scene are often at different depths, and due to parallax, shifts between imaged objects often vary on a pixel basis. This poses a great computational challenge as the problem is NP complete. The multi-lenslet camera with a single focal plane provides us a unique opportunity to take advantage of the parallax phenomenon, and to directly relate object depths with their shifts, and thus we essentially reduced the parameter space from a two dimensional (x, y) space to a one dimensional depth space, which would greatly reduce the computational cost. As results, not only we have registered LR images, the estimated depth map can also be valuable for some applications. After registration, LR images along with estimated shifts can be used to reconstruct an HR image. A previously developed algorithm will be employed to efficiently compute for a large HR image in the size of 1024x1024.

Matching highly non-ideal ocular images: An information fusion approach

We consider the problem of matching highly non-ideal ocular images where the iris information cannot be reliably used. Such images are characterized by non-uniform illumination, motion and de-focus blur, off-axis gaze, and non-linear deformations. To handle these variations, a single feature extraction and matching scheme is not sufficient. Therefore, we propose an information fusion framework where three distinct feature extraction and matching schemes are utilized in order to handle the significant variability in the input ocular images. The Gradient Orientation Histogram (GOH) scheme extracts the global information in the image; the modified Scale Invariant Feature Transform (SIFT) extracts local edge anomalies in the image; and a Probabilistic Deformation Model (PDM) handles nonlinear deformations observed in image pairs. The simple sum rule is used to combine the match scores generated by the three schemes. Experiments on the extremely challenging Face and Ocular Challenge Series (FOCS) database and a subset of the Face Recognition Grand Challenge (FRGC) database confirm the efficacy of the proposed approach to perform ocular recognition.

Development and Validation of a Video-Animated Tool for Assessing Mobility

Background. Existing self-report measures of mobility ignore important contextual features of movement and require respondents to make complex judgments about specific tasks. Thus, we describe the development and validation of a short form (sf) video-animated tool for assessing mobility, the Mobility Assessment Tool—MAT-sf. Methods. This study involves cross-sectional and longitudinal analyses examining the measurement properties of the MAT-sf. The MAT-sf consists of 10 animated video clips that assess respondents’ level of proficiency in performing each task. The main outcome measures used for validation included the Pepper Assessment Tool for Disability (PAT-D), the Short Physical Performance Battery (SPPB), and 400-m walk test. Results. Participants (n = 234), 166 women and 68 men, had an average age of 81.9 years and a variety of comorbidities with 65.4% having high blood pressure. An average SPPB score of 8.6 (range 2–12) suggests that the study sample had evidence of compromised physical function but was quite heterogeneous. The MAT-sf had good content validity, excellent test–retest reliability (r = .93), and criterion-related validity with the PAT-D. Moreover, the MAT-sf added considerable variance to the prediction of both SPPB scores and 400-m gait speed over and above the PAT-D mobility subscale. The MAT-sf also discriminated between older adults who completed or failed the 400-m walk test. Conclusion. The MAT-sf is an innovative psychometrically sound measure of mobility. It has utility in epidemiological studies, translational science, and clinical practice.

High-resolution iris image reconstruction from low-resolution imagery

We investigate the use of a novel multi-lens imaging system in the context of biometric identification, and more specifically, for iris recognition. Multi-lenslet cameras offer a number of significant advantages over standard single-lens camera systems, including thin form-factor and wide angle of view. By using appropriate lenslet spacing relative to the detector pixel pitch, the resulting ensemble of images implicitly contains subject information at higher spatial frequencies than those present in a single image. Additionally, a multi-lenslet approach enables the use of observational diversity, including phase, polarization, neutral density, and wavelength diversities. For example, post-processing multiple observations taken with differing neutral density filters yields an image having an extended dynamic range. Our research group has developed several multi-lens camera prototypes for the investigation of such diversities. In this paper, we present techniques for computing a high-resolution reconstructed image from an ensemble of low-resolution images containing sub-pixel level displacements. The quality of a reconstructed image is measured by computing the Hamming distance between the Daugman4 iris code of a conventional reference iris image, and the iris code of a corresponding reconstructed image. We present numerical results concerning the effect of noise and defocus blur in the reconstruction process using simulated data and report preliminary work on the reconstruction of actual iris data obtained with our camera prototypes.

Development of a Time Sensitivity Score for Frequently Occurring Motor Vehicle Crash Injuries

BACKGROUND Injury severity alone is a poor indicator of the time sensitivity of injuries. The purpose of the study was to quantify the urgency with which the most frequent motor vehicle crash injuries require treatment, according to expert physicians. STUDY DESIGN The time sensitivity was quantified for the top 95% most frequently occurring Abbreviated Injury Scale (AIS) 2+ injuries in the National Automotive Sampling System-Crashworthiness Data System (NASS-CDS) 2000-2011. A Time Sensitivity Score was developed using expert physician survey data in which physicians were asked to determine whether a particular injury should go to a Level I/II trauma center and the urgency with which that injury required treatment. RESULTS When stratifying by AIS severity, the mean Time Sensitivity Score increased with increasing AIS severity. The mean Time Sensitivity Scores by AIS severity were as follows: 0.50 (AIS 2); 0.78 (AIS 3); 0.92 (AIS 4); 0.97 (AIS 5); and 0.97 (AIS 6). When stratifying by anatomical region, the head, thorax, and abdomen were the most time sensitive. CONCLUSIONS Appropriate triage depends on multiple factors, including the severity of an injury, the urgency with which it requires treatment, and the propensity of a significant injury to be missed. The Time Sensitivity Score did not correlate highly with the widely used AIS severity scores, which highlights the inability of AIS scores to capture all aspects of injury severity. The Time Sensitivity Score can be useful in Advanced Automatic Crash Notification systems for identifying highly time sensitive injuries in motor vehicle crashes requiring prompt treatment at a trauma center.

Predicting patients that require care at a trauma center: Analysis of injuries and other factors

INTRODUCTION The detection of occult or unpredictable injuries in motor vehicle crashes (MVCs) is crucial in correctly triaging patients and thus reducing fatalities. The purpose of the study was to develop a metric that indicates the likelihood that an injury sustained in a MVC would require management at a Level I/II trauma centre (TC) versus a non-trauma centre (non-TC). METHODS Transfer Scores (TSs) were computed for 240 injuries that comprise the top 95% most frequently occurring injuries in the National Automotive Sampling System-Crashworthiness Data System (NASS-CDS) with an Abbreviated Injury Scale (AIS) severity of 2 or greater. A TS for each injury was computed using the proportions of patients involved in a MVC from the National Inpatient Sample (NIS) that were transferred to a TC or managed at a non-TC. Similarly, a TSMAIS that excludes patients with higher severity co-injuries was calculated using the proportion of patients with a maximum AIS (MAIS) equal to the AIS severity of a given injury. RESULTS The results indicated for injuries of a given AIS severity, body region, and injury type, there were large variations in the TSMAIS. Overall results demonstrated higher TSMAIS values when injuries were internal, haemorrhagic, intracranial or of moderate severity (AIS 3-5). Specifically, injuries to the head possessed a TSMAIS that ranged from 0.000 to 0.889, with head injuries of AIS 3-5 severities being the most likely to be transferred. DISCUSSION AND CONCLUSIONS The analysis indicated that the TSMAIS is not solely correlated with AIS severity and therefore it captures other important aspects of injury such as predictability and trauma system capabilities. The TS and TSMAIS can be useful in advanced automatic crash notification (AACN) research for the detection of highly unpredictable injuries in MVCs that require direct transport to a TC.

Development of a robust mapping between AIS 2+ and ICD-9 injury codes

Motor vehicle crashes result in millions of injuries and thousands of deaths each year in the United States. While most crash research datasets use Abbreviated Injury Scale (AIS) codes to identify injuries, most hospital datasets use the International Classification of Diseases, version 9 (ICD-9) codes. The objective of this research was to establish a one-to-one mapping between AIS and ICD-9 codes for use with motor vehicle crash injury research. This paper presents results from investigating different mapping approaches using the most common AIS 2+ injuries from the National Automotive Sampling System-Crashworthiness Data System (NASS-CDS). The mapping approaches were generated from the National Trauma Data Bank (NTDB) (428,637 code pairs), ICDMAP (2500 code pairs), and the Crash Injury Research and Engineering Network (CIREN) (4125 code pairs). Each approach may pair given AIS code with more than one ICD-9 code (mean number of pairs per AIS code: NTDB=211, ICDMAP=7, CIREN=5), and some of the potential pairs are unrelated. The mappings were evaluated using two comparative metrics coupled with qualitative inspection by an expert physician. Based on the number of false mappings and correct pairs, the best mapping was derived from CIREN. AIS and ICD-9 codes in CIREN are both manually coded, leading to more proper mappings between the two. Using the mapping presented herein, data from crash and hospital datasets can be used together to better understand and prevent motor vehicle crash injuries in the future.

Mortality risk in pediatric motor vehicle crash occupants: accounting for developmental stage and challenging Abbreviated Injury Scale metrics

Objective: Survival risk ratios (SRRs) and their probabilistic counterpart, mortality risk ratios (MRRs), have been shown to be at odds with Abbreviated Injury Scale (AIS) severity scores for particular injuries in adults. SRRs have been validated for pediatrics but have not been studied within the context of pediatric age stratifications. We hypothesized that children with similar motor vehicle crash (MVC) injuries may have different mortality risks (MR) based upon developmental stage and that these MRs may not correlate with AIS severity. Methods: The NASS-CDS 2000–2011 was used to define the top 95% most common AIS 2+ injuries among MVC occupants in 4 age groups: 0–4, 5–9, 10–14, and 15–18 years. Next, the National Trauma Databank 2002–2011 was used to calculate the MR (proportion of those dying with an injury to those sustaining the injury) and the co-injury-adjusted MR (MRMAIS) for each injury within 6 age groups: 0–4, 5–9, 10–14, 15–18, 0–18, and 19+ years. MR differences were evaluated between age groups aggregately, between age groups based upon anatomic injury patterns and between age groups on an individual injury level using nonparametric Wilcoxon tests and chi-square or Fishers exact tests as appropriate. Correlation between AIS and MR within each age group was also evaluated. Results: MR and MRMAIS distributions of the most common AIS 2+ injuries were right skewed. Aggregate MR of these most common injuries varied between the age groups, with 5- to 9-year-old and 10- to 14-year-old children having the lowest MRs and 0- to 4-year-old and 15- to 18-year-old children and adults having the highest MRs (all P <.05). Head and thoracic injuries imparted the greatest mortality risk in all age groups with median MRMAIS ranging from 0 to 6% and 0 to 4.5%, respectively. Injuries to particular body regions also varied with respect to MR based upon age. For example, thoracic injuries in adults had significantly higher MRMAIS than such injuries among 5- to 9-year-olds and 10- to 14-year-olds (P =.04; P <.01). Furthermore, though AIS was positively correlated with MR within each age group, less correlation was seen for children than for adults. Large MR variations were seen within each AIS grade, with some lower AIS severity injuries demonstrating greater MRs than higher AIS severity injuries. As an example, MRMAIS in 0- to 18-year-olds was 0.4% for an AIS 3 radius fracture versus 1.4% for an AIS 2 vault fracture. Conclusions: Trauma severity metrics are important for outcome prediction models and can be used in pediatric triage algorithms and other injury research. Trauma severity may vary for similar injuries based upon developmental stage, and this difference should be reflected in severity metrics. The MR-based data-driven determination of injury severity in pediatric occupants of different age cohorts provides a supplement or an alternative to AIS severity classification for pediatric occupants in MVCs.

PERIODIC: Integrated Computational Array Imaging Technology

An array imaging system, dubbed PERIODIC, is presented, capable of exploiting diversities, including subpixel displacement, phase, polarization, and wavelength, to produce superresolution images. The hardware system and software interface described, and sample results are shown.

The Virtual Short Physical Performance Battery

BACKGROUND Performance-based and self-report instruments of physical function are frequently used and provide complementary information. Identifying older adults with a mismatch between actual and perceived function has utility in clinical settings and in the design of interventions. Using novel, video-animated technology, the objective of this study was to develop a self-report measure that parallels the domains of objective physical function assessed by the Short Physical Performance Battery (SPPB)-the virtual SPPB (vSPPB). METHODS The SPPB, vSPPB, the self-report Pepper Assessment Tool for Disability, the Mobility Assessment Tool-short form, and a 400-m walk test were administered to 110 older adults (mean age = 80.6±5.2 years). One-week test-retest reliability of the vSPPB was examined in 30 participants. RESULTS The total SPPB (mean [±SD] = 7.7±2.8) and vSPPB (7.7±3.2) scores were virtually identical, yet moderately correlated (r = .601, p < .05). The component scores of the SPPB and vSPPB were also moderately correlated (all p values <.01). The vSPPB (intraclass correlation = .963, p < .05) was reliable; however, individuals with the lowest function overestimated their overall lower extremity function while participants of all functional levels overestimated their ability on chair stands, but accurately perceived their usual gait speed. CONCLUSION In spite of the similarity between the SPPB and vSPPB, the moderate strength of the association between the two suggests that they offer unique perspectives on an older adults physical function.