Ye Changkong, Zhang Wenyan, Pang Zijuan, Wang Wei
Changkong Ye, Bachelor's Degrees. Department of Ultrasound, Beihai people's Hospital, Beihai 536000, China.
Wenyan Zhang, Bachelor's Degrees. Department of Ultrasound, Beihai people's Hospital, Beihai 536000, China.
Pak J Med Sci. 2021;37(6):1693-1698. doi: 10.12669/pjms.37.6-WIT.4885.
To explore the therapeutic effects of ultrasound-guided microwave ablation and radio frequency ablation for liver cancer patients.
Seventy-eight patients with microwave ablation were rolled into the experimental group and 56 patients with radio frequency ablation were in the control group. This study was conducted from March 1, 2019 to June 30, 2020 in our hospital. Based on Convolutional Neural Networks (CNN) and Migration feature (MF), a new ultrasound image diagnosis algorithm CNNMF was constructed, which was compared with AdaBoost and PCA-BP based on Principal component analysis (PCA) and back propagation (BP), and the accuracy (Acc), specificity (Spe), sensitivity (Sen), and F1 values of the three algorithms were calculated. Then, the CNNMF algorithm was applied to the ultrasonic image diagnosis of the two patients, and the postoperative ablation points, complications and ablation time were recorded.
The Acc (96.31%), Spe (89.07%), Sen (91.26%), and F1 value (0.79%) of the CNNMF algorithm were obviously larger than the AdaBoost and the PCA-BP algorithms (P< 0.05); in contrast with the control group. The number of ablation points in the experimental group was obviously larger, and the ablation time was obviously shorter (P<0.05); the experimental group had one case of liver abscess and two cases of wound pain after surgery, which were both obviously less than the control group (four cases; five cases) (P<0.05).
In contrast with traditional algorithms, the CNNMF algorithm has better diagnostic performance for liver cancer ultrasound images. In contrast with radio frequency ablation, microwave ablation has better ablation effects for liver cancer tumors, and can reduce the incidence of postoperative complications in patients, which is safe and feasible.
探讨超声引导下微波消融和射频消融治疗肝癌患者的疗效。
将78例行微波消融的患者纳入实验组，56例行射频消融的患者作为对照组。本研究于2019年3月1日至2020年6月30日在我院进行。基于卷积神经网络（CNN）和迁移特征（MF）构建了一种新的超声图像诊断算法CNMF，并将其与基于主成分分析（PCA）和反向传播（BP）的AdaBoost和PCA-BP算法进行比较，计算三种算法的准确率（Acc）、特异性（Spe）、敏感性（Sen）和F1值。然后将CNMF算法应用于两组患者的超声图像诊断，并记录术后消融点、并发症及消融时间。
CNMF算法的Acc（96.31%）、Spe（89.07%）、Sen（91.26%）和F1值（0.79%）明显大于AdaBoost和PCA-BP算法（P<0.05）；与对照组相比，实验组的消融点数明显更多，消融时间明显更短（P<0.05）；实验组术后发生肝脓肿1例、伤口疼痛2例明显少于对照组（4例、5例）（P<0.05）。
与传统算法相比，CNMF算法对肝癌超声图像具有更好的诊断性能。与射频消融相比，微波消融对肝癌肿瘤具有更好的消融效果，可降低患者术后并发症的发生率，安全可行。
