Explainable AI for Brain Tumour Auditing and Tagging

用于脑肿瘤审核和标记的可解释人工智能

• 批准号: 10072780
• 金额: $ 6.36万
• 依托单位: THERMOSOL LIMITED
• 依托单位国家: 英国
• 项目类别: Grant for R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10072780
• 关键词: Explainable; AI; Brain; Tumour; Auditing

Brain tumors are a serious health issue; early detection is crucial for successful treatment. Magnetic Resonance Imaging (MRI) is the most accurate non-surgical method for identifying brain tumors, but manual analysis by radiologists is time-consuming and prone to errors. Deep learning models have shown promising results in detecting brain tumors, but their black-box nature and privacy concerns restrict their adoption in real-world scenarios. To address these issues, a new framework called BAT is proposed, which uses state-of-the-art explainable AI models to detect autonomously, segment brain tumors using MRI images, and provide concise explanations for the decision. BAT can be used for clinical auditing or tagging without requiring patient personal information. The key innovation of BAT is its ability to automate tumor detection, segmentation, and stage prediction while ensuring anonymity and interpretability.

脑肿瘤是一个严重的健康问题；早期检测对于成功治疗至关重要。磁共振成像（MRI）是鉴定脑肿瘤的最准确的非手术方法，但是放射科医生进行的手动分析既耗时又容易出现误差。深度学习模型在检测脑肿瘤方面表现出了令人鼓舞的结果，但是他们的黑盒性质和隐私涉及在现实世界中的采用限制。为了解决这些问题，提出了一个新的称为BAT的框架，该框架使用可解释的AI模型来自主检测，使用MRI图像分割脑肿瘤，并为该决定提供简洁的解释。 BAT可用于临床审核或标记，而无需患者个人信息。 BAT的关键创新是其自动化肿瘤检测，分割和舞台预测的能力，同时确保匿名和解释性。