Dense volumetric detection and segmentation of mediastinal lymph nodes in chest CT images

TY - CONF

T1 - Dense volumetric detection and segmentation of mediastinal lymph nodes in chest CT images

AU - Oda, Hirohisa

AU - Roth, Holger R.

AU - Bhatia, Kanwal K.

AU - Oda, Masahiro

AU - Kitasaka, Takayuki

AU - Iwano, Shingo

AU - Homma, Hirotoshi

AU - Takabatake, Hirotsugu

AU - Mori, Masaki

AU - Natori, Hiroshi

AU - Schnabel, Julia A.

AU - Mori, Kensaku

PY - 2018/2/27

Y1 - 2018/2/27

N2 - We propose a novel mediastinal lymph node detection and segmentation method from chest CT volumes based on fully convolutional networks (FCNs). Most lymph node detection methods are based on filters for blob-like structures, which are not specific for lymph nodes. The 3D U-Net is a recent example of the state-of-the-art 3D FCNs. The 3D U-Net can be trained to learn appearances of lymph nodes in order to output lymph node likelihood maps on input CT volumes. However, it is prone to oversegmentation of each lymph node due to the strong data imbalance between lymph nodes and the remaining part of the CT volumes. To moderate the balance of sizes between the target classes, we train the 3D U-Net using not only lymph node annotations but also other anatomical structures (lungs, airways, aortic arches, and pulmonary arteries) that can be extracted robustly in an automated fashion. We applied the proposed method to 45 cases of contrast-enhanced chest CT volumes. Experimental results showed that 95.5% of lymph nodes were detected with 16.3 false positives per CT volume. The segmentation results showed that the proposed method can prevent oversegmentation, achieving an average Dice score of 52.3 ± 23.1%, compared to the baseline method with 49.2 ± 23.8%, respectively.

AB - We propose a novel mediastinal lymph node detection and segmentation method from chest CT volumes based on fully convolutional networks (FCNs). Most lymph node detection methods are based on filters for blob-like structures, which are not specific for lymph nodes. The 3D U-Net is a recent example of the state-of-the-art 3D FCNs. The 3D U-Net can be trained to learn appearances of lymph nodes in order to output lymph node likelihood maps on input CT volumes. However, it is prone to oversegmentation of each lymph node due to the strong data imbalance between lymph nodes and the remaining part of the CT volumes. To moderate the balance of sizes between the target classes, we train the 3D U-Net using not only lymph node annotations but also other anatomical structures (lungs, airways, aortic arches, and pulmonary arteries) that can be extracted robustly in an automated fashion. We applied the proposed method to 45 cases of contrast-enhanced chest CT volumes. Experimental results showed that 95.5% of lymph nodes were detected with 16.3 false positives per CT volume. The segmentation results showed that the proposed method can prevent oversegmentation, achieving an average Dice score of 52.3 ± 23.1%, compared to the baseline method with 49.2 ± 23.8%, respectively.

KW - Computer-Aided Diagnosis

KW - Fully convolutional network

KW - Imbalance weighting

U2 - 10.1117/12.2287066

DO - 10.1117/12.2287066

M3 - Abstract

AN - SCOPUS:85046290545

T2 - Medical Imaging 2018: Computer-Aided Diagnosis

Y2 - 12 February 2018 through 15 February 2018

ER -