Table 71 Recent studies describing fully automated LV or RV segmentation algorithm based on Convolutional Neural Networks (CNN)
Author, year | Segmented structure | Data used for training/validation | Validation methods/remarks |
|---|---|---|---|
Tran, 2017 [203] | LV + RV | LV: MICCAI-2009 (n = 45) RV: RVSC-2012 (n = 45) | Validation: auto vs manual Metrics: DICE, HD, ACD |
Bai, 2018 [2] | LV + RV Including LV, RV, LA, RA from long-axis cine | 4875 subjects UK Biobank cohort Multi-center, single vendor | Validation: auto vs manual Metrics: DICE, HD, APD LV: EDV, ESV, EF, SV, CO, LV mass RV: EDV, ESV, EF, CO |
Bernard, 2018 [198] | LV + RV | ACDC-2017 (n = 150) | Nine methods compared Validation: auto vs manual Metrics: DICE, HD, CLBR LV: EDV, ESV, EF, LV mass RV: EDV, ESV, EF |
Khened, 2018 [200] | LV + RV | KAGGLE-2015 (n = 1140) ACDC-2017 (n = 150) LVSC-2011 (n = 200) | Validation: auto vs manual Metrics: DICE, CLBR Patient diagnosis |
Tan, 2018 [201] | LV | LVSC-2011 (n = 200) KAGGLE-2015 (n = 1140) | Validation: auto vs manual Metrics: DICE, JI, HD EDV, ESV |
Tao, 2018 [3] | LV | Training: 400 subjects Testing: 150 subjects Multi-center, multi-vendor Multiple patient categories: MI (n = 322), DCM (n = 168), HCM (n = 23), DCM (n = 23), PH (n = 10) other (n = 27), normal (n = 23) | Validation: auto vs manual Metrics: DICE EDV, ESV, EF, LV mass |
Vigneault, 2018 [204] | LV + RV from multiple views | 53 subjects HCM (n = 42), healthy (n = 21) ACDC-2017 (n = 150) | Validation: auto vs manual Metrics: DICE |
Backhaus, 2019 [212] | LV + RV | Evaluation of SuiteHEART software (Neosoft) 300 randomly selected patients used for validation Single center 1.5 T and 3 T data | Validation: auto vs manual LV: EDV, ESV, SV, EF, LV mass RV: EDV, ESV, EF |
Bhuva, 2019 [197] | LV | Training data: 599 subjects Test data 110 patients, 5 disease categories: myocardial infarction (n = 32), LVH (n = 17), cardiomyopathy (n = 17), other pathology (n = 14), healthy volunteers (n = 30) Multi-center, multi-vendor, 1.5 T + 3 T Scan-rescan data Data availability: | Validation based on comparing scan-rescan reproducibility of automated vs manual analysis EDV, ESV, SV, EF, LV mass Detectable change in EF |
Curiale, 2019 [199] | LV | MICCAI-2009 (n = 45) Cardiac Atlas Project (n = 95) | Validation: auto vs manual Metrics: DICE EDV, ESV, EF, LV mass |
Tong, 2019 [202] | LV + RV | ACDC-2017 (n = 150) | Validation: auto vs manual Metrics: Dice, HD LV: EDV, ESV, EF, LV mass RV: EDV, ESV, EF |