Abstract
To provide a reference for finding a reasonable evaluation method for treatment effect of radiofrequency ablation (RFA), computed tomography (CT) image optimized by the intelligent segmentation algorithm was utilized to evaluate the liver condition of hepatocellular carcinoma (HCC) patients after RFA and to estimate the patient’s prognosis. Eighty-eight patients with HCC who needed RFA surgery after diagnosis in our hospital were selected. The CT images before optimization were set as the control group; the CT images after optimization were set as the observation group. Comprehensive diagnosis was taken as the gold standard to compare the ablation range and residual lesions under CT scans before and after surgery. The results showed that the consistency of the two sets of CT images was compared with comprehensive diagnosis under different diameters of the lesion. The difference between the two groups was not statistically considerable when the diameter of the lesion was less than 50 mm (). For lesions larger than 50 mm in diameter, the consistency of the observation group (83%) was remarkably higher than that of the control group (40%), and the difference was substantial (). The kappa value of the observation group was 0.84 and that of the control group was 0.78. The kappa value of observation group was better than the control group, with considerable difference (). In conclusion, the diagnostic effect of CT image based on intelligent segmentation algorithm was superior to conventional diagnosis when the diameter of the lesion was larger than 50 mm. Moreover, the overall improvement rate of patients after RFA treatment was far greater than the recurrence rate, indicating that the clinical adoption of RFA was very meaningful.
1. Introduction
Liver cancer refers to malignant tumors that occur in liver tissue, which is mainly classified into two categories: primary hepatic carcinoma and secondary liver cancer. In terms of the incidence of the two liver cancers, the former is much higher than the latter. The incidence of hepatocellular carcinoma (HCC) in primary liver cancer is relatively high [1–3]. HCC occupies a high ranking in the incidence of various malignant tumors. According to statistical analysis in recent years, HCC has a high incidence in China. Moreover, the incidence on a global scale is also very high, ranking 5th among all malignant tumors [4]. The main pathogenic factors of HCC include genetics, chronic viral hepatitis, and eating habits, but the pathogenesis of HCC is not clear [5]. The main treatment for HCC is surgical resection [6]. Surgical treatment can effectively remove tumor lesions and improve the prognosis of the diseases, but not all patients are suitable to do so. Some HCC patients may have symptoms such as liver cirrhosis during the progression of the disease, leading to insufficient liver function and unable to compensate for the effects of surgical resection, which makes HCC treatment require further development and improvement [7].
Through the continuous development of the medical level, the clinical treatment of HCC has become more and more diversified. Minimally invasive interventional techniques such as chemoembolization, radiofrequency, and laser have been widely used in the clinical treatment of HCC [8–10]. Radiofrequency ablation (RFA) is a local tumor treatment method that has been gradually applied in the past 20 years [11]. Compared with other treatment methods, there are problems such as large wounds, strong radiation, and health effects. RFA is a physical ablation method and has many advantages, such as small treatment wounds, small impact on liver and kidney function, high patient acceptance, high safety, and repeated treatment. It is precisely these advantages that RFA has become a new treatment for liver cancer, and its efficacy can compare to surgical resection [12, 13]. Efficacy evaluation after RFA surgery is indispensable. It is of great significance for improving the efficacy and prolonging the survival of liver cancer patients. The main evaluation method is imaging [14]. Routine dynamic enhanced computed tomography (CT) examination is the most convenient method to evaluate the treatment effect after RFA treatment, and it is widely used in clinical practice [15]. With the rapid development of artificial intelligence technology, deep learning algorithms have been widely used in the optimization of medical images. The processing directions include denoising, segmentation, and recombination, and the processing effect has been clinically recognized. Therefore, deep learning technology plays a vital role in medical image segmentation [16]. Medical image segmentation is also the basis for subsequent image analysis, three-dimensional reconstruction, dose calculation, and other processes, and the segmentation effect can directly affect the diagnosis efficiency [17]. In summary, the medical field requires deep learning algorithms to have good performance.
Therefore, to make CT scans more useful in the evaluation of therapeutic effects after RFA treatment, the CT image optimized by the intelligent segmentation algorithm was used to evaluate the liver condition of HCC patients after RFA treatment and predict the patient’s prognosis. It was hoped to provide a reference basis for finding a reasonable RFA treatment effect evaluation method.
2. Methods
2.1. Research Objects
In this study, 88 HCC patients who needed RFA surgery after being diagnosed by examination in our hospital from March 2018 to April 2020 were selected. Among them, there were 50 males and 38 females, aged from 35 to 74 years old, with an average of 58.91 ± 9.24 years old. Comprehensive scanning examination showed that the tumor diameter ranged from 9 to 58 mm, with an average of 32.89 ± 8.20 mm. There were 44 patients with lesion diameter <30 mm, 36 patients with lesion diameter 30–50 mm, and 8 patients with lesion diameter >50 mm. Before treatment, all patients received dynamic enhanced spiral CT scan. After that, intelligent segmentation algorithm was used to separate and optimize the liver lesions and normal tissues of all patients. The CT images of all patients were classified into two groups according to whether they were optimized or not. The CT images before optimization were set as the control group, and the optimized CT images were set as the observation group. The comprehensive diagnosis of laboratory examination, imaging examination, and pathological results was taken as the “gold standard.” The results of the two groups of imaging examination were analyzed and compared to evaluate their adoption value. The study was approved by the Medical Ethics Committee. Inclusion criteria: (1) all patients met HCC-related diagnostic criteria [18] and were confirmed by pathological results; (2) all patients had single-focus primary HCC; (3) all the patients were between 18 and 78 years old; (4) none of the patients had received other preoperative antitumor therapy; (5) patients and their family members knew about this study and signed the consent form; (6) liver function was classified according to the Child-Pugh grading standard [19] (Figure 1), and the Child-Pugh grading of liver function was all grade A or B; (7) contrast-enhanced CT (CECT) examination was completed in all patients after RFA. Exclusion criteria: (1) patients with incomplete follow-up after RFA; (2) patients with incomplete preservation of clinical data; (3) the image quality of CECT was not ideal.
2.2. Research Methods
2.2.1. CT Image Examination
CT scans were performed on all patients before and after RFA surgery. The patient was in a supine position with his arms raised above his head, and then a CT scan was performed. Philips Brilliance 16-layer spiral CT scanning instrument (Japan) was adopted for inspection. Figure 2 shows the detailed operation process and index parameters.
2.2.2. CT Image Based on Intelligent Segmentation Algorithm
In this study, a new CT processing method was adopted, namely, intelligent CT image segmentation of liver tumor. The method integrated the high-level knowledge of medical experts into the image segmentation algorithm to perform image segmentation. According to the characteristics of different stages of medical image segmentation and the applicability of different algorithms, the multiscale watershed transformation and fuzzy clustering method were combined to harvest the best results in general.
The process of segmentation has two steps. The first step is separating the liver part from the abdominal CT image, mainly using the characteristics of the image gray level. The multiscale watershed algorithm and morphological filtering method were adopted and combined with human anatomy knowledge, and the parameters of morphological filtering and watershed algorithm were dynamically adjusted to accurately extract the liver. There is no need for manual repair later. The second step, which is also the main processing part of this research, is extracting the liver lesions. Since medical images have greater fuzziness, the expert’s knowledge and fuzzy C-means (FCM) are integrated.
First, the traditional fuzzy clustering method is improved, and the objective function is modified. After modification, the objective function is improved, as shown in the following equation:
In equation (1), are the two-dimensional space coordinates of the image pixels, are the two-dimensional coordinates of T cluster centers, is the Euclidean distance of the image gray domain, is the distance between the image pixels and the cluster centers on the two-dimensional space plane Euclidean distance, and , which is a constant.
Then, new membership functions and iterative equations for cluster centers were derived to prove that the membership function and clustering center derived from the new objective function can reach a local minimum. If the initial cluster center and the initial membership function are selected correctly, both may reach the global minimum. The membership function iteration equation is derived as follows.
It is assumed that is known, and the Lagrange multiplier method is used to obtain the following equation:
Find the result of a member function that minimizes so that there is the following equation:
The result is as follows:
Since , equation (4) is substituted in equation (3), then there is the following equation:
Then, the following equation is obtained:
Equation (6) is substituted in equation (4), then the final expression equation (7) of member functions is obtained:
The iteration equation of clustering center is deduced as follows.
The modified objective function has two clustering centers, one is in the brightness domain and the other is in the space domain, whose iterative equations are then solved.
Lagrange multiplier method is used to find the value of when the objective function (1) gets the minimum value. The derivative of is taken by (2). Make the derivative equal to zero, then (8) is obtained:
After a series of calculation and arrangement, the following equation is obtained:
Then, the distance function is improved. The noise of the image is mainly represented by the stripes and particle spots on the image, so the CT noise is mainly small stripes and particle noise. The method of reducing the impact of noise is applied to FCM algorithm, and the following equation is obtained:
In equation (10), is used to select the influence of surrounding pixels on the distance . 0 ≤ β ≤ 1. is the gray value of surrounding pixels. When β = 0, it is the gray distance of pixels in the usual sense.
Finally, the improved FCM algorithm is used to extract the lesions. According to the principle of minimizing the objective function, the ending condition of the FCM algorithm is the minimum value of equation (1). Without reducing the accuracy of the algorithm, to reduce the amount of calculation, the termination condition of the iteration within the allowable error range is replaced by the following equation:
In equation (11), i is the number of iterations and ε is the iteration termination parameter with a value between (0, 1).
The specific image processing performance is shown in Figure 3.
2.2.3. Treatment Process
(1) The Choice of Position. Right liver disease corresponded to left decubitus position, while left liver disease corresponded to supine position.
(2) Position of Needle Puncture. Right liver disease corresponded to intercostal needle puncture, while left liver disease corresponded to subcostal puncture needle.
(3) Specific Operation Process. Radiofrequency acupuncture was punctured to the farthest end of the lesion under the guidance of B-ultrasound or CT, and the radiofrequency source was turned on for treatment. It was noted that the thoracic lobe and abdominal cavity organs should be avoided during needle insertion. The treatment methods of tumor lesions of different sizes were different, and the specific operations are shown in Table 1.
2.3. Observation Indexes
First, the gold standard for disease diagnosis in this study was the results of comprehensive judgment based on CT imaging examination, laboratory examination, needle biopsy, and follow-up (one to two years of follow-up).
All patients were evaluated by CT scan one month after RFA surgical treatment, including whether the ablation range (maximum layer and maximum diameter of the lesion) was reduced, and whether there were residual lesions.
Follow-up was performed every three months to observe whether there was tumor recurrence, which lasted for two consecutive years.
2.4. Statistical Methods
SPSS 22.0 was employed to carry out statistics, the count data were expressed by the rate (%), the Y2 test was performed, and the rank count data was subjected to the rank sum test. The measurement data were expressed by the mean plus or minus standard deviation, and the t-test was performed. indicates that the difference was statistically considerable. The consistency of the diagnosis of the two groups of CT images was compared by kappa test. A kappa value of 0.81∼1.00 meant excellent consistency, 0.61∼0.80 meant good consistency, 0.41∼0.60 meant moderate consistency, and <0.4 meant poor consistency.
3. Results
3.1. The Size Distribution of the Lesions Shown by the CT Images of Each Group before RFA
The statistical analysis of the CT image scan results of the observation group and the control group and the results under the comprehensive examination are shown in Table 2. For the diameter of the lesion that was less than 30 mm, the distribution of the number of people in the observation group was consistent with the comprehensive examination, both of which was 44 (100%), while that of the control group was 37 (84%). For the diameter of the lesion that was 30–50 mm, there were 36 people from comprehensive examination result, 35 (97%) people in the observation group, and 36 people (92%) in the control group. For the diameter of the lesion that was greater than 50 mm, there were 8 people from comprehensive examination result, 9 people (89%) in the observation group, and 12 people (75%) in the control group. Through comparative analysis, the results of the observation group and the comprehensive inspection were closer than that of the control group. However, the comparison of the similarity between the two groups of CT images and the comprehensive examination was not statistically considerable ().
3.2. Comparison of the Situation after Treatment
There were 88 tumor lesions found in this study. The comparison between the two groups of CT images in judging the lesions of different diameters and the complete ablation of the overall lesions after one month of RFA treatment are shown in Table 3. Through the CT image analysis of the control group, 83 cases (94%) were completely ablated, and 5 cases (6%) were left in the lesion. Through analysis of the CT images of the observation group, 78 cases (87%) were completely ablated, and 10 cases (13%) had residual lesions. There was no substantial difference between the two groups of CT images in the overall complete ablation rate of the lesion () (Figure 4). Then, the consistency of the judgment results of the two groups was compared with the results of the comprehensive diagnosis under different diameters, as illustrated in Figure 5. The comparison between the two groups was not statistically considerable in judging diameter of the lesion that was less than 50 mm (). For the lesion diameter that was greater than 50 mm, the consistency of the observation group (83%) was remarkably higher than that of the control group (40%), and the contrast was statistically considerable (). Moreover, the kappa values of the observation group and the control group were 0.84 and 0.78, respectively, both of which were in good agreement, and the comparison was statistically considerable ().
3.3. Progression of the Lesion during the Two-Year Follow-Up
During the two years of follow-up after treatment, the two groups of CT image analysis showed that the recurrence of lesions was 9 in the control group (10%) and 12 (14%) in the observation group, and the result of comprehensive diagnosis was 11 (13%). After comparison, the differences between the three groups of examination results were not statistically considerable (), as shown in Figure 6. The CT images of typical ablation and patients with recurrence are shown in Figures 7 and 8. Figures 7(a) and 8(a) show the two sets of CT images of the same patient at the sixth month after surgery. Both showed that the tumor in the right lobe of the liver was completely ablated. Figures 7(b) and 8(b) show the lesions of the same patient at the ninth month postoperative review. The recurrence of the original tumor edge was not obvious in Figure 7(b). In Figure 8(b), it was obvious that the tissue density around the lesion increased.
(a)
(b)
(a)
(b)
4. Discussion
As a commonly used minimally invasive treatment for liver cancer in clinical practice, RFA, its mechanism of action is radiofrequency therapy, which is a kind of hyperthermia [16]. The treatment mechanism is that the cluster electrode of the radiofrequency needle tip emits medium and high frequency radiofrequency waves (460 kHz) to excite tissue cells to perform plasma oscillations. The ions collide with each other to generate heat up to 80–100°C, which causes the coagulation and necrosis of tumor tissue that can effectively and quickly kill local tumor cells [17]. RFA can be treated repeatedly with the recurrence and metastasis of cancer. The patient has a high quality of life and a good immune status after treatment. In recent years, its effectiveness and safety have been recognized by foreign experts [18, 19]. Based on this study, it conducted a two-year follow-up survey and found that the proportion of recurrences in all patients (13%) was far lower than the number of patients with complete necrosis (87%). Moreover, the condition of the relapsed patients was basically improved after treatment again. This shows that RFA therapy can achieve good therapeutic effects for the treatment of HCC.
Many researchers used RFA for the radical treatment of HCC, and the results showed that the treatment effect of RFA was quite good. Moreover, for HCC with lesion diameter less than 50 mm, the therapeutic effect of RFA was basically the same as that of surgical resection [20, 21]. In this study, the ablation rate of HCC patients with lesions less than 50 mm was also remarkably high (75% vs 25%), which was consistent with the above research results. Some researchers suggested the following possible reasons for this phenomenon. First, the larger the size of the lesion, the more irregular the three-dimensional morphology, which was easy to cause cauterization and omission, and the higher the probability of residual cancer cells. Second, HCC with larger diameter was biologically more aggressive and prone to intrahepatic metastasis [22, 23]. Therefore, the diameter of the cancer focus has a certain reference value for the selection of cases before RFA and the prediction of the short-term efficacy. Of course, the detection of postoperative efficacy of RFA is also essential, and imaging is the main means of examination. In this study, conventional CT images and improved intelligent segmentation algorithm-based CT images were used to study and analyze the effect of RFA in the treatment of primary liver cancer, which were then compared with the gold standard of comprehensive inspection. The results showed that the CT image diagnosis results based on intelligent segmentation algorithm were relatively consistent with the comprehensive diagnosis results (kappa = 0.84), which indicated that the adoption of the improved intelligent segmentation algorithm in CT images of liver cancer lesions had certain significance. Other researchers came up with the same new method for segmenting medical images, and the experimental results showed that the improved fuzzy clustering algorithm was more compact and accurate than the traditional algorithm. It was also verified that this method can effectively segment CT images of liver tumors [13]. The study also showed that the consistency between conventional CT diagnosis results and comprehensive diagnosis results was high, kappa = 0.78, which was at a very good level. Therefore, the effect of conventional CT examination was also good. Some studies proposed that the sensitivity, specificity, accuracy, and consistency of kappa of CECT in examining the postoperative treatment effect of RFA were 85.71%, 94.92%, 93.15%, and 0.785, respectively, showing high accuracy and clinical value [24], which was very similar to the results of this study.
5. Conclusion
CT images based on intelligent segmentation algorithm were utilized to evaluate the effect of RFA in the treatment of HCC. The results showed that the diagnostic effect of CT image based on intelligent segmentation algorithm was superior to conventional diagnosis when the diameter of the lesion was larger than 50 mm. Moreover, RFA was more effective in treating liver cancer with a lesion diameter of less than 50 mm than that of liver cancer with a diameter greater than 50 mm, and the overall improvement rate was much greater than the recurrence rate, indicating that the clinical adoption of RFA was very meaningful. However, the set of observed indicators in this study was not comprehensive, resulting in a lack of completeness of the study. In future studies, it will optimize this aspect and provide more research support for the adoption of RFA therapy. It is deemed that RFA therapy will have an excellent development prospect in the future.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare no conflicts of interest.