深度學習在腦部影像的臨床應用

Clinical Application of Deep Learning for Brain Image

蕭輔仁(Furen Xiao) ； 廖珩君(Heng-Chun Liao) ； 廖俊智(Chun-Chih Liao)

《台灣醫學》 24卷1期 (2020/01) Pp. 50-57

https://doi.org/10.6320/FJM.202001_24(1).0005

摘要

人工智慧（artificial intelligence, Al）在醫療上的運用已然成為顯學，日常為病人服務過程中產生出的大量資料也成為科學家有興趣的目標。有了深度學習技術的加持後，醫學影像分析發生了驚人的進展。AI逐漸成為我們臨床工作的好夥伴，不久之後常規的胸部X光或腦部CT一做完，電腦馬上就可產生初步報告，就和現在的心電圖機一樣。腦部影像中比較常被使用到的就是CT和MRI：已經有很多人成功的利用深度學習在CT上偵測出腦中風及各種外傷性病灶；腦部MRI的應用則更加廣泛，電腦可以在影像中自動偵測腦瘤、血管病灶、退化性病灶，劃出它們的輪廓甚至做出診斷、嚴重度分級、產生報告。更進一步的，AI也可以產生虛擬的CT或MRI，運用在臨床的教育訓練上。做為一個AI時代的醫療人員，除了初步瞭解人工智慧、深度學習這些技術到底是什麼之外，我們更應該知道目前這些技術能夠做到什麼，進一步發想出未來可能的臨床應用。

關鍵字

深度學習 ； 腦部影像 ； 放射手術

並列摘要

In recent years, artificial intelligence have revolutionized the medical profession. Data generated from daily diagnostic and treatment processes are considered ＂gold mines＂ from which valuable information and knowledge can be extracted. With the aid of deep learning, the process of medical image analysis has undergone revolutionary changes. Some experts even predict that human radiologists and pathologists will be replaced by machines within years. Although such change may occur within decades instead of years, artificial intelligence (AI) has already been an integral part of medical service. Real-time computer-generated reports are already provided on modern electrocardiography machines. Similar computer-aided diagnosis may be implemented on X-ray or computed tomography (CT) machines in the near future. CT and magnetic resonance imaging (MRI) are the two most widely used diagnostic equipments for brain diseases. Many researchers have successfully detected stroke and traumatic lesions on CT images using AI techniques. On MRI images, a wider range pathologies including tumors, vascular lesions and degenerative lesions can be decected. Based on these algorithms, it is possible to develop systems of computer-aided lesion segmentation ((or other words more 'friendly' to medical professionals, such as still use detection or edge detection?)), volumetry, grading and report generation. Furthermore, AI can be used to generate images of given disease type appearing at any location, providing materials for clinical training. The technical details of deep learing or AI is beyond the scope of this article. However, as we have entered the era of AI, knowing current medical applications and its potentials will open a new avenue for ＂non-conventional＂ solutions to unsolved problems.

並列關鍵字

deep learning ； brain image ； radiosurgery

參考文獻

Jha S, Topol EJ: Adapting to Artificial Intelligence: Radiologists and Pathologists as Information Specialists. JAMA 2016;316:2353-4. doi: 10.1001/jama.2016.17438

Menze BH, Jakab A, Bauer S, et al: The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS). IEEE Trans Med Imaging 2015;34:1993-2024. doi: 10.1109/TMI.2014.2377694

Krizhevsky A, Sutskever I, Hinton GE: Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems. 2012;1097-105.

Wang G, Li W, Ourselin S, Vercauteren T: Automatic brain tumor segmentation using cascaded anisotropic convolutional neural networks. International MICCAI Brainlesion Workshop. Springer, 2017;178-90.