TY - JOUR
T1 - AI-Based Automated Lipomatous Tumor Segmentation in MR Images
T2 - Ensemble Solution to Heterogeneous Data
AU - Liu, Chih Chieh
AU - Abdelhafez, Yasser G.
AU - Yap, S. Paran
AU - Acquafredda, Francesco
AU - Schirò, Silvia
AU - Wong, Andrew L.
AU - Sarohia, Dani
AU - Bateni, Cyrus
AU - Darrow, Morgan A.
AU - Guindani, Michele
AU - Lee, Sonia
AU - Zhang, Michelle
AU - Moawad, Ahmed W.
AU - Ng, Quinn Kwan Tai
AU - Shere, Layla
AU - Elsayes, Khaled M.
AU - Maroldi, Roberto
AU - Link, Thomas M.
AU - Nardo, Lorenzo
AU - Qi, Jinyi
N1 - Publisher Copyright:
© 2023, The Author(s).
PY - 2023/6
Y1 - 2023/6
N2 - Deep learning (DL) has been proposed to automate image segmentation and provide accuracy, consistency, and efficiency. Accurate segmentation of lipomatous tumors (LTs) is critical for correct tumor radiomics analysis and localization. The major challenge of this task is data heterogeneity, including tumor morphological characteristics and multicenter scanning protocols. To mitigate the issue, we aimed to develop a DL-based Super Learner (SL) ensemble framework with different data correction and normalization methods. Pathologically proven LTs on pre-operative T1-weighted/proton-density MR images of 185 patients were manually segmented. The LTs were categorized by tumor locations as distal upper limb (DUL), distal lower limb (DLL), proximal upper limb (PUL), proximal lower limb (PLL), or Trunk (T) and grouped by 80%/9%/11% for training, validation and testing. Six configurations of correction/normalization were applied to data for fivefold-cross-validation trainings, resulting in 30 base learners (BLs). A SL was obtained from the BLs by optimizing SL weights. The performance was evaluated by dice-similarity-coefficient (DSC), sensitivity, specificity, and Hausdorff distance (HD95). For predictions of the BLs, the average DSC, sensitivity, and specificity from the testing data were 0.72 ± 0.16, 0.73 ± 0.168, and 0.99 ± 0.012, respectively, while for SL predictions were 0.80 ± 0.184, 0.78 ± 0.193, and 1.00 ± 0.010. The average HD95 of the BLs were 11.5 (DUL), 23.2 (DLL), 25.9 (PUL), 32.1 (PLL), and 47.9 (T) mm, whereas of SL were 1.7, 8.4, 15.9, 2.2, and 36.6 mm, respectively. The proposed method could improve the segmentation accuracy and mitigate the performance instability and data heterogeneity aiding the differential diagnosis of LTs in real clinical situations.
AB - Deep learning (DL) has been proposed to automate image segmentation and provide accuracy, consistency, and efficiency. Accurate segmentation of lipomatous tumors (LTs) is critical for correct tumor radiomics analysis and localization. The major challenge of this task is data heterogeneity, including tumor morphological characteristics and multicenter scanning protocols. To mitigate the issue, we aimed to develop a DL-based Super Learner (SL) ensemble framework with different data correction and normalization methods. Pathologically proven LTs on pre-operative T1-weighted/proton-density MR images of 185 patients were manually segmented. The LTs were categorized by tumor locations as distal upper limb (DUL), distal lower limb (DLL), proximal upper limb (PUL), proximal lower limb (PLL), or Trunk (T) and grouped by 80%/9%/11% for training, validation and testing. Six configurations of correction/normalization were applied to data for fivefold-cross-validation trainings, resulting in 30 base learners (BLs). A SL was obtained from the BLs by optimizing SL weights. The performance was evaluated by dice-similarity-coefficient (DSC), sensitivity, specificity, and Hausdorff distance (HD95). For predictions of the BLs, the average DSC, sensitivity, and specificity from the testing data were 0.72 ± 0.16, 0.73 ± 0.168, and 0.99 ± 0.012, respectively, while for SL predictions were 0.80 ± 0.184, 0.78 ± 0.193, and 1.00 ± 0.010. The average HD95 of the BLs were 11.5 (DUL), 23.2 (DLL), 25.9 (PUL), 32.1 (PLL), and 47.9 (T) mm, whereas of SL were 1.7, 8.4, 15.9, 2.2, and 36.6 mm, respectively. The proposed method could improve the segmentation accuracy and mitigate the performance instability and data heterogeneity aiding the differential diagnosis of LTs in real clinical situations.
UR - http://www.scopus.com/inward/record.url?scp=85149016621&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85149016621&partnerID=8YFLogxK
U2 - 10.1007/s10278-023-00785-1
DO - 10.1007/s10278-023-00785-1
M3 - Article
C2 - 36854923
AN - SCOPUS:85149016621
SN - 0897-1889
VL - 36
SP - 1049
EP - 1059
JO - Journal of Digital Imaging
JF - Journal of Digital Imaging
IS - 3
ER -