eprintid: 10193741 rev_number: 9 eprint_status: archive userid: 699 dir: disk0/10/19/37/41 datestamp: 2024-06-24 14:23:25 lastmod: 2024-06-24 14:23:25 status_changed: 2024-06-24 14:23:25 type: article metadata_visibility: show sword_depositor: 699 creators_name: Lee, Matthew creators_name: Sanchez-Matilla, Ricardo creators_name: Stoyanov, Danail creators_name: Luengo, Imanol title: DIPO: Differentiable Parallel Operation Blocks for Surgical Neural Architecture Search ispublished: inpress divisions: UCL divisions: B04 divisions: C05 divisions: F48 keywords: Neural Architecture Search, Surgical Image Segmentation, Surgical Keypoint Detection, Surgical Instrument Detection note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. abstract: Deep learning has been used across a large number of computer vision tasks, however designing the network architectures for each task is time consuming. Neural Architecture Search (NAS) promises to automatically build neural networks, optimised for the given task and dataset. However, most NAS methods are constrained to a specific macro-architecture design which makes it hard to apply to different tasks (classification, detection, segmentation). Following the work in Differentiable NAS (DNAS), we present a simple and efficient NAS method, Differentiable Parallel Operation (DIPO), that constructs a local search space in the form of a DIPO block, and can easily be applied to any convolutional network by injecting it in-place of the convolutions. The DIPO block's internal architecture and parameters are automatically optimised end-to-end for each task. We demonstrate the flexibility of our approach by applying DIPO to 4 model architectures (U-Net, HRNET, KAPAO and YOLOX) across different surgical tasks (surgical scene segmentation, surgical instrument detection, and surgical instrument pose estimation) and evaluated across 5 datasets. Results show significant improvements in surgical scene segmentation (+10.5% in CholecSeg8K, +13.2% in CaDIS), instrument detection (+1.5% in ROBUST-MIS, +5.3% in RoboKP), and instrument pose estimation (+9.8% in RoboKP). date: 2024-05-28 date_type: published publisher: Institute of Electrical and Electronics Engineers (IEEE) official_url: http://dx.doi.org/10.1109/jbhi.2024.3406065 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 2287324 doi: 10.1109/JBHI.2024.3406065 medium: Print-Electronic lyricists_name: Stoyanov, Danail lyricists_id: DSTOY26 actors_name: Stoyanov, Danail actors_name: Woolley, Clare actors_id: DSTOY26 actors_id: CWOOL20 actors_role: owner actors_role: impersonator funding_acknowledgements: 203145/Z/16/Z [Wellcome/EPSRC Centre for Interventional and Surgical Sciences] full_text_status: public publication: IEEE Journal of Biomedical and Health Informatics pagerange: 1-13 event_location: United States issn: 2168-2194 citation: Lee, Matthew; Sanchez-Matilla, Ricardo; Stoyanov, Danail; Luengo, Imanol; (2024) DIPO: Differentiable Parallel Operation Blocks for Surgical Neural Architecture Search. IEEE Journal of Biomedical and Health Informatics pp. 1-13. 10.1109/JBHI.2024.3406065 <https://doi.org/10.1109/JBHI.2024.3406065>. (In press). Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10193741/1/Stoyanov_DIPO_at_JBHI_Final-2.pdf