% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@ARTICLE{Starke:284753,
      author       = {S. Starke$^*$ and A. A. Zwanenburg-Bezemer$^*$ and K.
                      Leger$^*$ and F. Lohaus$^*$ and A. Linge$^*$ and G.
                      Kalinauskaite$^*$ and I. Tinhofer$^*$ and N. Guberina$^*$
                      and M. Guberina$^*$ and P. Balermpas$^*$ and J. v. d.
                      Grün$^*$ and U. Ganswindt$^*$ and C. Belka$^*$ and J. C.
                      Peeken$^*$ and S. E. Combs$^*$ and S. Boeke$^*$ and D.
                      Zips$^*$ and C. Richter$^*$ and E. Troost$^*$ and M.
                      Krause$^*$ and M. Baumann$^*$ and S. Löck$^*$},
      title        = {{M}ultitask {L}earning with {C}onvolutional {N}eural
                      {N}etworks and {V}ision {T}ransformers {C}an {I}mprove
                      {O}utcome {P}rediction for {H}ead and {N}eck {C}ancer
                      {P}atients.},
      journal      = {Cancers},
      volume       = {15},
      number       = {19},
      issn         = {2072-6694},
      address      = {Basel},
      publisher    = {MDPI},
      reportid     = {DKFZ-2023-02064},
      pages        = {4897},
      year         = {2023},
      abstract     = {Neural-network-based outcome predictions may enable further
                      treatment personalization of patients with head and neck
                      cancer. The development of neural networks can prove
                      challenging when a limited number of cases is available.
                      Therefore, we investigated whether multitask learning
                      strategies, implemented through the simultaneous
                      optimization of two distinct outcome objectives
                      (multi-outcome) and combined with a tumor segmentation task,
                      can lead to improved performance of convolutional neural
                      networks (CNNs) and vision transformers (ViTs). Model
                      training was conducted on two distinct multicenter datasets
                      for the endpoints loco-regional control (LRC) and
                      progression-free survival (PFS), respectively. The first
                      dataset consisted of pre-treatment computed tomography (CT)
                      imaging for 290 patients and the second dataset contained
                      combined positron emission tomography (PET)/CT data of 224
                      patients. Discriminative performance was assessed by the
                      concordance index (C-index). Risk stratification was
                      evaluated using log-rank tests. Across both datasets, CNN
                      and ViT model ensembles achieved similar results. Multitask
                      approaches showed favorable performance in most
                      investigations. Multi-outcome CNN models trained with
                      segmentation loss were identified as the optimal strategy
                      across cohorts. On the PET/CT dataset, an ensemble of
                      multi-outcome CNNs trained with segmentation loss achieved
                      the best discrimination (C-index: 0.29, $95\%$ confidence
                      interval (CI): 0.22-0.36) and successfully stratified
                      patients into groups with low and high risk of disease
                      progression (p=0.003). On the CT dataset, ensembles of
                      multi-outcome CNNs and of single-outcome ViTs trained with
                      segmentation loss performed best (C-index: 0.26 and 0.26,
                      CI: 0.18-0.34 and 0.18-0.35, respectively), both with
                      significant risk stratification for LRC in independent
                      validation (p=0.002 and p=0.011). Further validation of the
                      developed multitask-learning models is planned based on a
                      prospective validation study, which has recently completed
                      recruitment.},
      keywords     = {Cox proportional hazards (Other) / convolutional neural
                      network (Other) / discrete-time survival models (Other) /
                      head and neck cancer (Other) / loco-regional control (Other)
                      / multitask learning (Other) / progression-free survival
                      (Other) / survival analysis (Other) / tumor segmentation
                      (Other) / vision transformer (Other)},
      cin          = {DD01 / BE01 / E220 / ED01 / FM01 / MU01 / TU01 / HD01},
      ddc          = {610},
      cid          = {I:(DE-He78)DD01-20160331 / I:(DE-He78)BE01-20160331 /
                      I:(DE-He78)E220-20160331 / I:(DE-He78)ED01-20160331 /
                      I:(DE-He78)FM01-20160331 / I:(DE-He78)MU01-20160331 /
                      I:(DE-He78)TU01-20160331 / I:(DE-He78)HD01-20160331},
      pnm          = {315 - Bildgebung und Radioonkologie (POF4-315)},
      pid          = {G:(DE-HGF)POF4-315},
      typ          = {PUB:(DE-HGF)16},
      pubmed       = {pmid:37835591},
      pmc          = {pmc:PMC10571894},
      doi          = {10.3390/cancers15194897},
      url          = {https://inrepo02.dkfz.de/record/284753},
}