BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000712Z
LOCATION:506
DTSTART;TZID=America/Denver:20231112T160000
DTEND;TZID=America/Denver:20231112T161500
UID:submissions.supercomputing.org_SC23_sess430_ws_cafcw121@linklings.com
SUMMARY:Automated Whole-Body Tumor Segmentation and Prognosis of Cancer on
  PET/CT
DESCRIPTION:Kevin H. Leung (Johns Hopkins University)\n\nBackground: Cance
 r is the second leading cause of death in the United States [1]. Automatic
  characterization of malignant disease is an important clinical need to fa
 cilitate early detection and treatment of cancer [2]. Advances in machine 
 learning (ML) and deep learning (DL) have shown significant promise for ra
 diological and oncological applications [3]. Radiomic analysis extracts qu
 antitative features from radiologic data about a cancerous tumor [4]. DL m
 ethods require large training datasets with sufficiently annotated images,
  which are difficult to obtain for radiological applications. The objectiv
 e of this study was to develop a deep semi-supervised transfer learning ap
 proach for automated whole-body tumor segmentation and prognosis on positr
 on emission tomography (PET)/computed tomography (CT) scans using limited 
 annotations (Fig. 1a). \n\nMethods: Five datasets consisting of 1,019 pros
 tate, lung, melanoma, lymphoma, head and neck, and breast cancer patients 
 with prostate-specific membrane antigen (PSMA) and fluorodeoxyglucose (FDG
 ) PET/CT scans were used in this study (Table 1). A nnUnet backbone was cr
 oss-validated on the tumor segmentation task via a 5-fold cross-validation
 . Predicted segmentations were iteratively improved using radiomic analysi
 s. Transfer learning generalized the segmentation task across PSMA and FDG
  PET/CT. Segmentation accuracy was evaluated on true positive rate (TPR), 
 positive predictive value (PPV), Dice similarity coefficient (DSC), false 
 discovery rate (FDR), true negative rate (TNR), and negative predictive va
 lue (NPV). Imaging measures quantifying molecular tumor burden and uptake 
 were extracted from the predicted segmentations. A risk stratification mod
 el was developed for prostate cancer by combining the extracted imaging me
 asures and was evaluated on follow-up prostate-specific antigen (PSA) leve
 ls. A risk stratification model was developed for head and neck cancer pat
 ients by combining imaging measures and American Joint Committee on Cancer
  (AJCC) staging and was evaluated via Kaplan-Meier survival analysis. A pr
 ognostic model was developed to predict pathological response of breast ca
 ncer patients to neoadjuvant chemotherapy using imaging measures from pre-
 therapy and post-therapy PET/CT scans. Prognostic models were evaluated on
  overall accuracy and area under the receiver operating characteristic (AU
 ROC) curve. Statistically significant differences were inferred using a Wi
 lcoxon rank-sum test. \n\nResults: Accuracy metrics and illustrative examp
 les of predicted tumor segmentations are shown in Table 2 and Fig. 1b. The
  risk stratification model yielded an overall accuracy of 0.83 and an AURO
 C of 0.86 in stratifying prostate cancer patients (Fig. 1c). Median follow
 -up PSA levels in the low-intermediate and high risk groups were 1.19 ng/m
 L and 53.20 ng/mL (P < 0.05). Head and neck cancer patients were stratifie
 d into low, intermediate, and high risk groups with significantly differen
 t Kaplan-Meier survival curves by the log-rank test (Fig. 1d). A prognosti
 c model using imaging measures from pre-therapy scans predicted pathologic
 al complete response (pCR) in breast cancer patients with an accuracy of 0
 .72 and an AUROC of 0.72. The model using imaging measures from both pre-t
 herapy and post-therapy scans predicted pCR in breast cancer patients with
  an accuracy of 0.84 and an AUROC of 0.76. \n\nConclusion: A deep semi-sup
 ervised transfer learning approach was developed and demonstrated accurate
  tumor segmentation, quantification, and prognosis on PET/CT of patients a
 cross six cancer types.\n\nTag: Applications, State of the Practice\n\nReg
 istration Category: Workshop Reg Pass\n\nSession Chairs: Lynn Borkon (Fred
 erick National Laboratory for Cancer Research); Sally Ellingson (Universit
 y of Kentucky); Sean Hanlon (National Institutes of Health (NIH), National
  Cancer Institute (NCI)); Patricia Kovatch (Icahn School of Medicine at Mo
 unt Sinai); and Eric Stahlberg (MD Anderson Cancer Center, University of T
 exas)\n\n
END:VEVENT
END:VCALENDAR