摘要详情
182 / 2024-06-14 23:03:08
Automatic brain tumor identification using label-free nonlinear optical microscopy and deep learning
brain tumor identification, deep learning, nonlinear optical microscopy
全文待审
Surgical resection is the primary treatment for brain cancer, where maximizing cancer removal while preserving adjacent healthy tissue is crucial. This task is challenging due to the invasive nature of brain tumors. Current intraoperative histopathology methods, such as frozen section and smear preparation, have drawbacks like labeling requirements and being time-consuming and labor-intensive. In this paper, we propose an automatic, label-free method for brain tumor identification using nonlinear optical microscopy (NOM) combined with deep learning. NOM is employed to capture dual-modal images of brain tissue, including two-photon excitation fluorescence (TPEF) and fluorescence lifetime imaging (FLIM) images. These images serve as input for a U-net network, which accurately identifies tumor regions. This method was tested on brain tissue slices from tumor-bearing nude mice and ND2:SmoA1 transgenic mice. Results show that NOM can reveal cerebral structures and metabolic conditions. By extracting structural features and microenvironment information from NOM images, U-net could distinguish between tumor and peritumor regions with an average accuracy of 0.98 and evaluate the granular layer. The deep learning-based, label-free NOM approach promises automatic identification of cancerous versus non-cancerous tissues and is expected to integrate with endoscopy, offering a promising tool for characterizing cancer and cerebral structures during neurosurgery.
重要日期
-
会议日期
09月08日
2024
至09月12日
2024
-
09月15日 2024
初稿截稿日期
-
09月15日 2024
注册截止日期
主办单位
ShenZhen University
联系方式
- Organizing Committee
- ic******@163.com
- +86********
历届会议
-
2009年05月11日 中国 哈尔滨市
2009年纳米光子学国际会议
