MRI-based automatic segmentation of rectal cancer using 2D U-Net on two independent cohorts.
Franziska KnuthIngvild Askim AddeBao Ngoc HuynhAurora Rosvoll GroendahlRené Mario WinterAnne NegårdStein Harald HolmedalSebastian MeltzerAnne Hansen ReeNorman John CarrSvein DuelandKnut Håkon HoleTherese SeierstadKathrine Røe RedalenCecilia Marie FutsaetherPublished in: Acta oncologica (Stockholm, Sweden) (2021)
T2w MR-based DL models demonstrated high performance for automatic tumor segmentation, at the same level as published data on interobserver variation. DWI did not improve results further. Using DL models on unseen cohorts requires caution, and one cannot expect the same performance.
Keyphrases
- deep learning
- rectal cancer
- convolutional neural network
- contrast enhanced
- diffusion weighted imaging
- artificial intelligence
- diffusion weighted
- machine learning
- magnetic resonance imaging
- locally advanced
- big data
- magnetic resonance
- electronic health record
- computed tomography
- squamous cell carcinoma
- neural network
- meta analyses