基於超音波影像特徵之腫瘤偵測及邊緣擷取-以2D甲狀腺及3D乳房影像為例

Tumor Detection and Segmentation based on Sonographic Features with Examples of 2D Thyroid Images and 3D Breast Images

邱齡瑩(Ling-Ying Chiu)

指導教授 ： 陳正剛

國立臺灣大學/工學院/工業工程學研究所/博士(2020年)

https://doi.org/10.6342/NTU202003153

摘要

欲在超音波影像上執行腫瘤的電腦輔助診斷，需要明確定義腫瘤的位置和邊界。然而，腫瘤本身的生物學特性、超音波成像的物理性質和品質、操作者的主觀認知與操作條件等種種因素，都使得辨識腫瘤的位置及邊界更加困難。
本研究的主題主要聚焦於超音波影像當中腫瘤的自動偵測以及邊緣分割。我們應用了兩種超音波影像：甲狀腺和乳房，分別應用於探討自動邊界分割和位置偵測的問題。我們將提出之方法應用於實際臨床案例的2D和3D超音波影像上，以評估所提方法之效能。
在邊界分割問題上，我們提出了一種新穎的半自動分割方法，使用Variance-Reduction的統計方法來對甲狀腺腫瘤的邊界進行分割，且不需對影像進行預處理。而在位置偵測問題上，我們提出了一種全自動電腦輔助偵測演算法，將two-phase merge-filter 演算法應用於自動三維乳房超音波影像。
研究結果顯示，我們提出的分割方法對於超音波影像上甲狀腺腫瘤的邊界分割是可靠且有效的。另外，我們提出的電腦輔助偵測系統具備極高的靈敏度，並伴隨相當低的偽陽性值，因而具有高度潛力成為自動三維乳房超音波影像的良好輔助工具。

關鍵字

自動偵測 ； 自動分割 ； 自動三維乳房超音波 ； 乳房腫瘤 ； 電腦輔助偵測 ； two-phase merge-filter ； 腫瘤位置 ； 腫瘤邊界 ； 甲狀腺腫瘤 ； 超音波影像 ； Variance-Reduction statistic

並列摘要

To perform computer-aided diagnosis of tumors on ultrasound images, the location and boundary of tumors should be clearly defined. However, the identification of tumors location and boundary are difficult issues due to the biological characteristics of the tumors, the physics and quality of ultrasound imaging, and the subjective factors and operating conditions of the operator.
The main focus of this research is on the automatic detection and segmentation of tumors in ultrasound images. Two types of ultrasound images, thyroid and breast, were used to explore the issues of automatic boundary segmentation and location detection, respectively. The performances of the proposed methods were then applied on 2D and 3D ultrasound images of actual clinical cases.
In boundary segmentation, a novel and semi-automatic method was proposed for segmenting the boundary of thyroid nodule based on the Variance-Reduction (V-R) statistics without image preprocessing. In location detection, we proposed a fully automatic computer-aided detection (CADe) algorithm applying on automated three-dimensional breast ultrasound (ABUS) images with a two-phase merge-filter algorithm.
It was shown that the segmentation method was reliable and effective in segmenting thyroid nodule boundary on ultrasound images. Moreover, the proposed CADe system had a great potential of becoming a good companion tool of the ABUS imaging by ensuring high sensitivity with a relatively small number of false-positives.

並列關鍵字

Automatic detection ； automatic segmentation ； automated three-dimensional breast ultrasound ； breast tumor ； computer-aided detection ； two-phase merge-filter ； tumor location ； tumor boundary ； thyroid nodule ； ultrasound image ； Variance-Reduction statistic

參考文獻

[1] Lim H, Devesa SS, Sosa JA, Check D, Kitahara CM. Trends in thyroid cancer incidence and mortality in the United States, 1974-2013. JAMA. 2017;317(13):1338-1348.

[2] Davies L, Welch HG. Current thyroid cancer trends in the United States. JAMA Otolaryngol Head Neck Surg. 2014;140(4):317-322.

[3] Baker SG. The central role of receiver operating characteristic (ROC) curves in evaluating tests for the early detection of cancer. J Natl Cancer Inst. 2003;95(7):511-515.

[4] Nabhan F, Ringel MD. Thyroid nodules and cancer management guidelines: comparisons and controversies. Endocr Relat Cancer. 2017;24(2):R13-R26.

[5] Chen K-Y, Chen C-N, Wu M-H, Ho M-C, Tai H-C, Kuo W-H, Huang W-C, Wang Y-H, Chen A, Chang K-J. Computerized quantification of ultrasonic heterogeneity in thyroid nodules. Ultrasound Med Biol. 2014;40(11):2581-2589.