國立陽明交通大學 機械工程學系講者：國立陽明交通大學機械工程研究所 林敬凱博士生
講題：The application of convolutional neural networks in the diagnosis of pulmonary malignancy
時間：2022/11/17 (周四) 3:30 pm - 5:30 pm
地點：工程五館 B1 國際會議廳
摘要：
Lung cancer is the most common cause of cancer related death in the worldwide. The outcome of the disease depends on staging, therefore proper staging must be performed to determine the treatment plan. The transthoracic approach with CT- guided biopsy traditionally is the first choice due to its having the highest diagnostic accuracy. However, because of its high complication rate, which may lead to patient morbidity and mortality, the use of bronchoscopic techniques has gradually increased. Endobronchial ultrasound (EBUS) is a revolutionary tool for lung cancer diagnosis and lymph node staging. In the previous publications, the diagnostic yields of EBUS alone is around 70%. When combining rapid on-site cytologic evaluation (ROSE), the diagnostic accuracy of EBUS procedures can significant improved. However, few cytologists are willing to perform ROSE during the bronchoscopy procedure. We developed a Convolutional Neural Network (CNN) system in the interpretation of cytologic specimens by EBUS procedures. Malignant cells were successfully classified with more than 90% accuracy. We obtained an accuracy of 86.7% in the differentiating the cell type of lung cancer (adenocarcinoma, squamous cell carcinoma, and small cell lung cancer). During the EBUS procedures, several possible targets are frequently found in the same station. Distinguishing the most likely target lesion is a critical issue for determining the diagnostic yield of the procedure. We constructed a new CNN system in the judgement of EBUS image pattern, and the diagnostic accuracy was 82% and an area under the curve was 0.8812. For the guidance of bronchoscopy and EBUS procedures, we demonstrate an innovative electromagnetic targeting system utilizing a passive magnetic-flux- concentrator for tracking endobronchoscope. The guide sheath with the silicon-steel collar travels between the emitting coil and the receiving electromagnets-arrays, the magnetic flux is concentrated by the silicon-steel collar and thereby the induced voltage is changed. According to experimental results, the system is successfully verified to be able to target the endobronchoscope in the bronchial-tree model. The targeting errors on the x-, y- and z-axes are 9 mm, 10 mm, and 5 mm, respectively.
講者學經歷：
- 中國醫藥大學醫學系 學士 87.07-94.06
- 國立交通大學機械系 博士生 105.09~至今
- 國立台灣大學附設癌醫中心醫院綜合內科部 主治醫師 109.02-至今
- 國立台灣大學醫學院內科學 臨床講師 109.02-至今
- 國泰綜合醫院內科 住院醫師 94.09-95.09
- 國立台灣大學醫學院附設醫院內科 住院醫師 95.10-98.06
- 國立台灣大學醫學院附設醫院胸腔內科 總醫師 98.07-100.06
- 台大醫院新竹分院內科 主治醫師 100.07-106.07
- 台大醫院新竹分院內科加護病房 病房主任 101.09~105.09
- 國立台灣大學醫學院附設醫院內科 主治醫師 106.08-109.02
研究專長：醫學、胸腔暨重症、微奈米工程