MT-TransUNet: Mediating Multi-Task Tokens in Transformers for Skin Lesion Segmentation and Classification
This work addresses automated skin cancer diagnosis for medical applications by integrating segmentation and classification, though it is incremental as it builds on existing Transformer-based methods.
The paper tackled the problem of skin cancer diagnosis by proposing a multi-task framework for simultaneous lesion segmentation and classification, achieving state-of-the-art results on ISIC-2017 and PH2 datasets with improved computational efficiency, such as reducing model parameters from 130M to 48M and inference time from 2.02s to 0.17s per image.
Recent advances in automated skin cancer diagnosis have yielded performance on par with board-certified dermatologists. However, these approaches formulated skin cancer diagnosis as a simple classification task, dismissing the potential benefit from lesion segmentation. We argue that an accurate lesion segmentation can supplement the classification task with additive lesion information, such as asymmetry, border, intensity, and physical size; in turn, a faithful lesion classification can support the segmentation task with discriminant lesion features. To this end, this paper proposes a new multi-task framework, named MT-TransUNet, which is capable of segmenting and classifying skin lesions collaboratively by mediating multi-task tokens in Transformers. Furthermore, we have introduced dual-task and attended region consistency losses to take advantage of those images without pixel-level annotation, ensuring the model's robustness when it encounters the same image with an account of augmentation. Our MT-TransUNet exceeds the previous state of the art for lesion segmentation and classification tasks in ISIC-2017 and PH2; more importantly, it preserves compelling computational efficiency regarding model parameters (48M~vs.~130M) and inference speed (0.17s~vs.~2.02s per image). Code will be available at https://github.com/JingyeChen/MT-TransUNet.