BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230124T171520Z
LOCATION:D222
DTSTART;TZID=America/Chicago:20221113T094500
DTEND;TZID=America/Chicago:20221113T100000
UID:submissions.supercomputing.org_SC22_sess432_ws_cafcw101@linklings.com
SUMMARY:A Generalized Tumor Segmentation Algorithm for Varying Breast Canc
 er Subtypes
DESCRIPTION:Workshop\n\nA Generalized Tumor Segmentation Algorithm for Var
 ying Breast Cancer Subtypes\n\nBanerjee\n\nBackground. Automated breast tu
 mor segmentation for dynamic contrast-enhanced magnetic resonance (DCE-MR)
  is a crucial step to advance and help with the implementation of radiomic
 s for image-based, quantitative assessment of breast tumors and cancer phe
 notyping. Current studies focus on developing tumor segmentation, which of
 ten requires initial seed points from expert radiologists or atlas-based s
 egmentation methods. We develop a robust, fully automated end-to-end segme
 ntation pipeline for breast cancers on bilateral breast MR studies.\n\nMet
 hods. On IRB-approved diverse breast cancer MR cases, a deep learning segm
 entation algorithm was created and trained. The model’s backbone is UNet++
 , which consists of U-Nets of varying depths whose decoders are densely co
 nnected at the same resolution via the skip connections and all the consti
 tuent UNets are trained simultaneously to learn a shared image representat
 ion. This design not only improves the overall segmentation performance, b
 ut also enables model pruning during the inference time. The model was tra
 ined on the breast tumors located independently by a radiologist with cons
 ensus review by a second radiologist with at least five years of experienc
 e. MRI was performed using a 3.0-T imaging system in the prone position wi
 th a dedicated 16-channel breast coil and T1 weighted DEC-MR images were a
 nalyzed for the study. We used 80:20 random split for training and validat
 ion of the model.\n\nResults. A total of 124 breast cancer patients had pr
 e-treatment MR imaging before the start of NST - the cohort comprised 49 H
 R+HER2-, 37 HR+HER2+, 11 HR-HER2+, and 27 TNBC cases (mean tumor 2.3 cm (+
 /- 3.1mm).) The model was tested on 2571 individual images. Overall, the m
 odel scored 0.85 [0.84 – 0.86, 95% CI] dice score and 0.8[0.79-0.81, 95% C
 I] IoU score. TNBC tumors scored dice [0.88 – 0.89, 95% CI], HER2 neg and 
 ER/PR positive dice [0.84-0.85, 95% CI] and HER2 positive dice [0.84-0.85,
  95% CI]. We observed that model performed equally for the solid tumors an
 d irregular shapes and didn’t observe any difference in the segmentation p
 erformance between residual and non-residual tumors types - dice score [0.
 85 – 0.86, 95% CI] and [0.83 – 0.84, 95% CI] respectively. \n\nConclusion.
  The proposed segmentation model can perform equally well on various clini
 cal breast cancer subtypes. The model has high false positive rate towards
  biopsy clip and high background enhancement which we plan to solve by add
 ing annotation of the clip and high non-cancer enhancement in future train
 ing data. We will release the trained model with open-source license to in
 crease the scalability of the radiomics studies with fully automated segme
 ntation. Given the importance of breast cancer subtypes as prognostic fact
 ors in women with operable breast cancer, automated segmentation of varyin
 g breast tumor subtypes will help to analyze imaging biomarkers embedded w
 ithin the standard of care imaging studies in a larger scale study which w
 ill ¬potentially help radiologists, pathologists, surgeons, and clinicians
  understand features driving breast cancer phenotypes and pave the way for
  developing digital twin for breast cancer patients.\n\nSession Format: Re
 corded\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR