BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160559Z
LOCATION:Track 11
DTSTART;TZID=America/New_York:20201113T121500
DTEND;TZID=America/New_York:20201113T124000
UID:submissions.supercomputing.org_SC20_sess229_ws_ai4s104@linklings.com
SUMMARY:Deep Learning-Based Low-Dose Tomography Reconstruction with Hybrid
 -Dose Measurements
DESCRIPTION:Workshop\n\nDeep Learning-Based Low-Dose Tomography Reconstruc
 tion with Hybrid-Dose Measurements\n\nWu, Bicer, Liu, Andrade, Zhu...\n\nS
 ynchrotron-based X-ray computed tomography is widely used for investigatin
 g inner structures of specimens at high spatial resolutions. However, pote
 ntial beam damage to samples often limits the X-ray exposure during tomogr
 aphy experiments. Proposed strategies for eliminating beam damage also dec
 rease reconstruction quality. Here we present a deep learning-based method
  to enhance low-dose tomography reconstruction via a hybrid-dose acquisiti
 on strategy composed of extremely sparse-view normal-dose projections and 
 full-view low-dose projections. Corresponding image pairs are extracted fr
 om low-/normal-dose projections to train a deep convolutional neural netwo
 rk, which is then applied to enhance  full-view noisy low-dose projections
 .\n\nEvaluation on two experimental datasets under different hybrid-dose a
 cquisition conditions show significantly improved structural details and r
 educed noise levels compared to uniformly distributed  acquisitions with t
 he same number of total dosage. The resulting reconstructions also preserv
 e more structural information than reconstructions processed with traditio
 nal analytical and regularization-based iterative reconstruction methods f
 rom uniform acquisitions. Our performance comparisons show that our implem
 entation, HDrec, can perform de-noising of a real-world experimental data 
 410x faster than the state-of-the-art Xlearn method while providing better
  quality. This framework can be applied to other tomographic or scanning b
 ased X-ray imaging techniques for enhanced analysis of  dose-sensitive sam
 ples and has great potential for studying fast dynamic processes.\n\nRegis
 tration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR