Related Experiment Video
Updated: Jan 11, 2026
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
3.2K
AttenUNeT X with iterative feedback mechanisms for robust deep learning skin lesion segmentation
1School of Computer Science and Engineering, VIT University, Vellore, Tamil Nadu, India.
Scientific Reports
|November 19, 2025
Summary
This study introduces AttenUNeT X, an improved U-Net model for precise skin cancer segmentation. The novel approach enhances feature refinement and attention, leading to more accurate early diagnosis of skin lesions.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate skin lesion segmentation is crucial for early skin cancer detection.
- Existing segmentation models may struggle with complex lesion features and boundaries.
Purpose of the Study:
- To introduce AttenUNeT X, a novel U-Net architecture extension for enhanced skin lesion segmentation.
- To improve the accuracy and reliability of automated skin cancer diagnosis through advanced image analysis.
Main Methods:
- Developed AttenUNeT X, integrating a feedback mechanism, an Order Statistics Layer (OSL), and enhanced attention modules.
- Trained and validated the model on ISIC 2018, PH2, and ISIC 2017 datasets.
- Employed a preprocessing pipeline including hair removal, resizing, normalization, and data augmentation.
Main Results:
- Achieved a Dice coefficient of 0.9211, IoU of 0.8533, and pixel accuracy of 0.9824 on the ISIC 2018 dataset.
- Demonstrated strong performance across multiple datasets (ISIC 2018, PH2, ISIC 2017), indicating robust generalizability.
- The model effectively refines spatial features and prioritizes diagnostically relevant regions for improved segmentation.
Conclusions:
- AttenUNeT X significantly enhances skin lesion segmentation accuracy.
- The proposed model shows strong potential for clinical deployment in dermatological workflows for early skin cancer diagnosis.
