DARWIN: Deformable Patient Avatar Representation With Deep Image Network

Singh, V. P.; Ma, Kai; Tamersoy, Birgi; Chang, Yao‐Jen; Wimmer, Andreas; O’Donnell, Thomas; Chen, Terrence

doi:10.1007/978-3-319-66185-8_56

Cited by 27 publications

(22 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each pixel in the depth image describes the distance from the camera to the closest object surface. First, the algorithm detects the pose of the patient and body regions using the depth measurements and the known table position and shape [ 12 ]. After selection of the scan range, the ideal table height for the individual patient and the scheduled examination is proposed by the 3D camera such that the isocenter of the selected body region and the scanner isocenter align.…”

Section: Methodsmentioning

confidence: 99%

Automated patient positioning in CT using a 3D camera for body contour detection: accuracy in pediatric patients

et al. 2020

Self Cite

View full text Add to dashboard Cite

Objective To assess the accuracy of a 3D camera for body contour detection in pediatric patient positioning in CT compared with routine manual positioning by radiographers. Methods and materials One hundred and ninety-one patients, with and without fixation aid, which underwent CT of the head, thorax, and/or abdomen on a scanner with manual table height selection and with table height suggestion by a 3D camera were retrospectively included. The ideal table height was defined as the position at which the scanner isocenter coincides with the patient's isocenter. Table heights suggested by the camera and selected by the radiographer were compared with the ideal height. Results For pediatric patients without fixation aid like a baby cradle or vacuum cushion and positioned by radiographers, the median (interquartile range) absolute table height deviation in mm was 10.2 (16.8) for abdomen, 16.4 (16.6) for head, 4.1 (5.1) for thorax-abdomen, and 9.7 (9.7) for thorax CT scans. The deviation was less for the 3D camera: 3.1 (4.7) for abdomen, 3.9 (6.3) for head, 2.2 (4.3) for thorax-abdomen, and 4.8 (6.7) for thorax CT scans (p < 0.05 for all body parts combined). Conclusion A 3D camera for body contour detection allows for automated and more accurate pediatric patient positioning than manual positioning done by radiographers, resulting in overall significantly smaller deviations from the ideal table height. The 3D camera may be also useful in the positioning of patients with fixation aid; however, evaluation of possible improvements in positioning accuracy was limited by the small sample size. Key Points • A 3D camera for body contour detection allows for automated and accurate pediatric patient positioning in CT. • A 3D camera outperformed radiographers in positioning pediatric patients without a fixation aid in CT. • Positioning of pediatric patients with fixation aid was feasible using the 3D camera, but no definite conclusions were drawn regarding the positioning accuracy due to the small sample size.

show abstract

Section: Methodsmentioning

confidence: 99%

Automated patient positioning in CT using a 3D camera for body contour detection: accuracy in pediatric patients

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…As discussed in Section I, this approach is substantially different than existing state-of-the-art mesh estimation methods such as HMR [12]. While HMR also regresses mesh parameters from feature representations, it shares the same limitation as Singh et al [11], i.e., it can be trained only for one modality. Specifically, HMR is a one-branch architecture that estimates mesh parameters given data from a single modality.…”

Section: B Dynamic Multi-modal Mesh Inferencementioning

confidence: 99%

“…With our design, the system (with the model trained with both RGB and thermal data) can be used in all the hospitals above without requiring any retraining in each individual hospital. Note that this proposed approach is substantially different than existing state-of-the-art 3D mesh modeling methods such as HMR [12], which shares the same multi-modal limitations as Singh et al [11], i.e., it can be trained only for one modality. A useful byproduct of our algorithm design is built-in redundancy to ensure system robustness.…”

mentioning

confidence: 94%

“…However, in many relevant use-cases such as automated thickness estimation [10] and radiation dose optimization, one needs a full 3D mesh and not just the keypoint locations. While there is some recent progress [11] in addressing this problem, this method is limited to CTspecific poses and requires depth data. If we change either the application (e.g., X-ray poses and protocols) or even the particular type of the camera (e.g., some applications may be limited to RGB-only camera), this method will need (a) fresh collection and annotation of data, and (b) retraining the model with this new data, both of which may be prohibitively expensive to do repeatedly for each application separately.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Towards Contactless Patient Positioning

Karanam¹,

Ren²,

Yang³

et al. 2020

IEEE Trans. Med. Imaging

Self Cite

View full text Add to dashboard Cite

The ongoing COVID-19 pandemic, caused by the highly contagious SARS-CoV-2 virus, has overwhelmed healthcare systems worldwide, putting medical professionals at a high risk of getting infected themselves due to a global shortage of personal protective equipment. This has in-turn led to understaffed hospitals unable to handle new patient influx. To help alleviate these problems, we design and develop a contactless patient positioning system that can enable scanning patients in a completely remote and contactless fashion. Our key design objective is to reduce the physical contact time with a patient as much as possible, which we achieve with our contactless workflow. Our system comprises automated calibration, positioning, and multi-view synthesis components that enable patient scan without physical proximity. Our calibration routine ensures system calibration at all times and can be executed without any manual intervention. Our patient positioning routine comprises a novel robust dynamic fusion (RDF) algorithm for accurate 3D patient body modeling. With its multi-modal inference capability, RDF can be trained once and used across different applications (without retraining) having various sensor choices, a key feature to enable system deployment at scale. Our multi-view synthesizer ensures multi-view positioning visualization for the technician to verify positioning accuracy prior to initiating the patient scan. We conduct extensive experiments with publicly available and proprietary datasets to demonstrate efficacy. Our system has already been used, and had a positive impact on, hospitals and technicians on the front lines of the COVID-19 pandemic, and we expect to see its use increase substantially globally.

show abstract

“…An algorithm analyzes the camera images. 18 It detects anatomical landmarks such as the head top, chin, shoulders, hip, and so on, and infers the pose of the patient (eg, feet-first supine). In addition, a virtual patient Avatar is fitted to the depth data.…”

Section: D Cameramentioning

confidence: 99%