BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260605T154542Z
LOCATION:Bldg. 8 - Room B 101
DTSTART;TZID=Europe/Stockholm:20260629T140000
DTEND;TZID=Europe/Stockholm:20260629T143000
UID:submissions.pasc-conference.org_PASC26_sess110_msa265@linklings.com
SUMMARY:Carbon-Aware Compression Evaluation for Sustainable Medical Image 
 Classification
DESCRIPTION:Carlos Barrios Hernandez (Universidad Industrial de Santander,
  LIG/INRIA - CITI Laboratory)\n\nDeep learning models for medical imaging 
 often require substantial computational resources, resulting in high energ
 y consumption and carbon emissions that limit deployment in resource-const
 rained clinical environments. We propose a carbon-aware evaluation framewo
 rk for assessing deep learning compression methods during training and inf
 erence. The framework quantifies the trade-off between diagnostic performa
 nce, computational efficiency, and environmental impact. As a case study, 
 knowledge distillation was applied to the CheXpert chest X-ray dataset usi
 ng DenseNet-121 as the teacher model and MobileNetV2 as the student model.
  The distilled model achieved diagnostic performance comparable to the tea
 cher model while substantially reducing energy consumption and CO₂ emissio
 ns. Carbon impact was evaluated under realistic inference workloads throug
 h post-training measurements of energy usage and equivalent emissions. Res
 ults demonstrate that model compression can significantly improve energy e
 fficiency without compromising clinical relevance. The proposed framework 
 provides a reproducible methodology for advancing sustainable and environm
 entally responsible AI in medical imaging, particularly for low-resource a
 nd clinical deployment settings.\n\nDomain: Engineering, Life Sciences, Co
 mputational Methods and Applied Mathematics\n\nSession Chairs: John Anders
 on Garcia Henao (University of Bern, ARTORG Center for Biomedical Engineer
 ing Research) and Carlos Barrios Hernandez (Universidad Industrial de Sant
 ander, LIG/INRIA - CITI Laboratory)\n\n
END:VEVENT
END:VCALENDAR