Periocular malignancy represents an increasing burden and currently requires disfiguring surgery in an attempt to cure patients. Basal cell carcinoma (BCC) is the commonest cancer worldwide and morphoeic BCC (mBCC) is an aggressive subtype. Sebaceous gland carcinoma (SGC) is a rare, but life-threatening condition that often requires blinding surgery to prevent mortality, especially in the pagetoid subtype. MBCC has a high risk of local recurrence compared to the more indolent nodular subtype reflected by a different set of driver genes including FLNB and HECTD4. Surrounding mBCC stroma is abnormal, containing mutations in EPHA3 and GLI3. Four common dysregulated pathways detected using both whole exome and RNA sequencing for mBCC were; 'hedgehog (Hh) signalling pathway', 'BCC', 'Natural killer cell mediated cytotoxicity' and 'Fc Epsilon RI signalling pathway'. Hh mutational profile for nodular BCC was not reflected in the RNA and protein expression. In contrast, Hh overexpression is seen in the tumour and stroma of morphoeic tissue with the latter potentially being partly responsible for its aggressive nature and risk of recurrence that may warrant removal to prevent recurrence. SGC has a low overall mutational burden, no UV signature and defective mismatch repair signature. Driver genes included TP53, RB1 and the dynein family is a novel driver possibly involved in chromatid segregation as marked chromosomal instability was demonstrated on copy number analysis. Correlation of whole exome and RNA sequencing data demonstrated upregulated 'cell cycle', 'ubiquitin mediated proteolysis' and 'wnt signalling'. Subtype analysis of pagetoid and nodular SGC revealed the histone gene cluster family as important to both. Oncomir hsa-miR-21 was overexpressed in both and loss of hsa-miR-199a occurs in pagetoid. Increased protein expression of HIST1H2BD was seen in both subtypes as was Hh expression. These novel SGC findings support a chromosomally unstable cancer with the ability to invade extracellular matrix.

Philosophiae Doctor - PhD / Purpose: High risk human papillomavirus (HR-HPV) is well established as a causative agent of squamous cell carcinoma (SCC) of the orophaynx. HR-HPV has also been reported in periocular cancers and precancers, but controversy exists about its overall incidence and clinicopathologic profile. The purpose of this study is to evaluate the role of HR-HPV infection in periocular cancers and precancers, using multiple methods of detection. Design: Retrospective observational case series with laboratory investigations. Methods: Sequential surgical samples of 87 carcinomas (invasive SCC, SCC in situ and sebaceous carcinoma) from three different periocular sites (conjunctiva, lacrimal sac and the eyelid) diagnosed over a 15-year period (2000-2015) were selected for evaluation. Unstained paraffin sections of 87 cases of periocular carcinomas were analyzed with immunohistochemistry (IHC) for p16 as a screening test. p16 positive conjunctival- and lacrimal sac SCC were further evaluated for HR-HPV using DNA in situ hybridization (DNA ISH), and a subset was also analyzed by DNA Polymerase Chain Reaction (DNA PCR). p16 positive periocular sebaceous carcinomas (SC) were analyzed with PCR, and a subset of 18cases was further studied with a novel method of mRNA ISH, an advanced technique with an enhanced sensitivity and specificity. Relevant patient clinical information was obtained from review of the electronic medical records.

Many real-world face recognition tasks are under unconstrained conditions such as off-angle pose variations, illumination variations, facial occlusion, facial expression, etc. In this work, we are focusing on the real-world scenarios where only the periocular region of a face is visible such as in the ISIS case. In Part I of the dissertation, we will showcase the face recognition capability based on the periocular region, which we call the periocular face recognition. We will demonstrate that face matching using the periocular region directly is more robust than the full face in terms of age-tolerant face recognition, expression-tolerant face recognition, pose-tolerant face recognition, as well as contains more cues for determining the gender information of a subject. In this dissertation, we will study direct periocular matching more comprehensively and systematically using both shallow and deep learning methods. Based on this, in Part II and Part III of the dissertation, we will continue to explore an indirect way of carrying out the periocular face recognition: periocular-based full face hallucination, because we want to capitalize on the powerful commercial face matchers and deep learning-based face recognition engines which are all trained on large-scale full face images. The reproducibility and feasibility of re-training for a proprietary facial region, such as the periocular region, is relatively low, due to the nonopen source nature of commercial face matchers as well as the amount of training data and computation power required by the deep learning based models. We will carry out the periocular-based full face hallucination based on two proposed reconstructive dictionary learning methods, including the dimensionally weighted K-SVD (DW-KSVD) dictionary learning approach and its kernel feature space counterpart using Fastfood kernel expansion approximation to reconstruct high-fidelity full face images from the periocular region, as well as two proposed generative deep learning approaches that build upon deep convolutional generative adversarial networks (DCGAN) to generate the full face from the periocular region observations, including the Gang of GANs (GoGAN) method and the discriminant nonlinear many-to-one generative adversarial networks (DNMM-GAN) for applications such as the generative open-set landmark-free frontalization (Golf) for faces and universal face optimization (UFO), which tackles an even broader set of problems than periocular based full face hallucination. Throughout Parts I-III, we will study how to handle challenging realworld scenarios such as unconstrained pose variations, unconstrained illumination conditions, and unconstrained low resolution of the periocular and facial images. Together, we aim to achieve unconstrained periocular face recognition through both direct periocular face matching and indirect periocular-based full face hallucination. In the final Part IV of the dissertation, we will go beyond and explore several new methods in deep learning that are statistically efficient for generalpurpose image recognition. Methods include the local binary convolutional neural networks (LBCNN), the perturbative neural networks (PNN), and the polynomial convolutional neural networks (PolyCNN).