• ChoiceNet: CNN learning through choice of multiple feature map representations

      Rayhan, Farshid; email: f.rayhan@manchester.ac.uk; Galata, Aphrodite; Cootes, Tim F. (Springer London, 2021-07-11)
      Abstract: We introduce a new architecture called ChoiceNet where each layer of the network is highly connected with skip connections and channelwise concatenations. This enables the network to alleviate the problem of vanishing gradients, reduces the number of parameters without sacrificing performance and encourages feature reuse. We evaluate our proposed architecture on three independent tasks: classification, segmentation and facial landmark localisation. For this, we use benchmark datasets such as ImageNet, CIFAR-10, CIFAR-100, SVHN CamVid and 300W.
    • Mechanical, biological and tribological behaviour of fixation plates 3D printed by electron beam and selective laser melting

      Al-Tamimi, Abdulsalam Abdulaziz; Hernandez, Miguel A.; Omar, Abdalla; Morales-Aldana, David Felipe; Peach, Chris; Bartolo, Paulo; email: paulojorge.dasilvabartolo@manchester.ac.uk (Springer London, 2020-07-06)
      Abstract: Commercially available fixation plates are built using metallic biocompatible materials such as titanium and its alloys and stainless steel. However, these plates show a stiffness mismatch comparing to bone, leading to stress shielding and bone loss. In this paper, we investigate the combined use of topology optimisation and additive manufacturing to print fixation plates with reduced stiffness and improved biological performance. Ti-6Al-4 V plates were topology optimised considering different loading conditions and volume reductions and printed using electron beam melting and selective laser melting. The effect of processing conditions on the mechanical properties, microhardness, wear resistance and surface roughness was analysed. Results show acceptable wear resistance values for a medical device and a reduction of stress shielding by increasing volume reduction. It is also shown that no polishing is required as 3D printed plates are able to support cell attachment and proliferation. In comparison to commercial plates, 3D printed ones show significantly better biological performance. For the same design, SLM plates present higher mechanical properties, while EBM plates present better cell attachment and proliferation.