Diabetic retinopathy (DR) is a common complication of diabetes mellitus that can result in significant vision loss or blindness if left untreated, making early detection and effective management crucial. Ultra-widefield fundus (UWF) imaging, which captures up to 200 degrees of the retina, provides a broader and more comprehensive view for diagnosing and monitoring DR compared to traditional color fundus photography (CFP), which captures only 30 to 60 degrees. Despite its advantages, annotating UWF images remains labor-intensive, and the availability of open datasets for UWF imaging is limited compared to CFP. To address these challenges, two key methods are applied: 1) Multitask Learning with Pseudo-labeling, which enables the model to perform multiple tasks simultaneously and generalize more effectively by distilling knowledge from unlabeled data across various open datasets; and 2) UWF transformation, which makes UWF images resemble CFP images by focusing on common retinal structures despite differences in field of view. Our model achieved first place in two tasks-identification of diabetic macular edema and diabetic retinopathy-at MICCAI 2024 UWF4DR, demonstrating the effectiveness of these approaches in overcoming challenges in UWF image analysis. Our code, along with our data augmentation library that includes UWF preprocessing, is available at https://github.com/medi-whale/UWF4DR_mediwhale .