MAC-ResNet: Knowledge Distillation Based Lightweight Multiscale-Attention-Crop-ResNet for Eyelid Tumors Detection and Classification

Abstract

Eyelid tumors are tumors that occur in the eye and its appendages, affecting vision and appearance, causing blindness and disability, and some having a high lethality rate. Pathological images of eyelid tumors are characterized by large pixels, multiple scales, and similar features. Solving the problem of difficult and time-consuming fine-grained classification of pathological images is important to improve the efficiency and quality of pathological diagnosis. The morphology of Basal Cell Carcinoma (BCC), Meibomian Gland Carcinoma (MGC), and Cutaneous Melanoma (CM) in eyelid tumors are very similar, and it is easy to be misdiagnosed among each category. In addition, the diseased area, which is decisive for the diagnosis of the disease, usually occupies only a relatively minor portion of the entire pathology section, and screening the area of interest is a tedious and time-consuming task. In this paper, deep learning techniques to investigate the pathological images of eyelid tumors. Inspired by the knowledge distillation process, we propose the Multiscale-Attention-Crop-ResNet (MAC-ResNet) network model to achieve the automatic classification of three malignant tumors and the automatic localization of whole slide imaging (WSI) lesion regions using U-Net. The final accuracy rates of the three classification problems of eyelid tumors on MAC-ResNet were 96.8%, 94.6%, and 90.8%, respectively.

Development and evaluation of multimodal AI for diagnosis and triage of ophthalmic diseases using ChatGPT and anterior segment images: protocol for a two-stage cross-sectional study

“…Deep learning models, trained on large datasets, have demonstrated impressive capabilities for diagnosis, especially in the fields of image analysis within radiology (van Leeuwen et al, 2021), pathology (Niazi et al, 2019), and dermatology (Phillips et al, 2019). In ophthalmology, AI studies using image data such as fundus images, anterior segment images, optical coherence tomography and computed tomography images have achieved high accuracy in diagnosing glaucoma (Buisson et al, 2021;Akter et al, 2022), age-related macular degeneration (Yan et al, 2021;Chen et al, 2022), diabetic retinopathy (Son et al, 2020;Li et al, 2022), thyroidassociated ophthalmopathy (Shao et al, 2023a), corneal diseases (Gu et al, 2020;Fang et al, 2022;Tiwari et al, 2022), and ocular tumors (Huang et al, 2022;Shao et al, 2023b). Despite these achievements, an obvious limitation of image-based diagnosis is its inability to consider a patient's medical history, which restricts a comprehensive understanding of the patient's condition.…”