Deep Learning Model Based on CT Images to Predict Ki-67 Expression in Renal Carcinoma

doi:10.21203/rs.3.rs-4295534/v1

Download PDF

Research Article

Deep Learning Model Based on CT Images to Predict Ki-67 Expression in Renal Carcinoma

https://doi.org/10.21203/rs.3.rs-4295534/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Objective

To establish a preoperative prediction model of Ki-67 expression in renal carcinoma by combining CT images of renal carcinoma with deep learning technology, and to evaluate its effect in clinical application.

Methods

A retrospective analysis was performed on the CT images and pathological data of 137 patients with renal carcinoma who underwent renal CT plain scan plus enhancement scans and who were diagnosed pathologically from January 2019 to November 2023 at the Affiliated Hospital of Hebei University. All recruited patients were divided into 35 cases with Ki-67 ≥ 10% and 102 cases with Ki-67 < 10% based on Ki-67 expression. Of these, 110 patients were divided into the training group and the test group in a 4:1 ratio, and the remaining 27 cases were used as clinical validation group. Based on the Mobilenetv3-large model, prediction models incorporating plain scan, arterial, venous and excretion phases were constructed using the training set data. The predictive performance of the models, including accuracy, accuracy, sensitivity, specificity, F1 score, and AUC values, were assessed by inputting CT images of the test group. The best-performing model was applied to the clinical validation group to compare the predicted results with the actual results of pathology; the accuracy, sensitivity, specificity and Kappa coefficient of the model were calculated to assess its clinical efficacy.

Results

The Mobilenetv3-large model in the venous phase showed the best performance in terms of accuracy, sensitivity and F1 score, as well as relatively high precision and AUC value, indicating good robustness. In the clinical validation group, the model predicted Ki-67 expression with an accuracy of 0.814, sensitivity and specificity of 0.889 and 0.667 at low and high levels, respectively, and a Kappa coefficient of 0.57.

Conclusion

The Mobilenetv3-large model based on venous-phase CT images is potentially clinically valuable as it can effectively predict Ki-67 expression and provide accurate, efficient, and noninvasive support for the development of personalized treatment plans.

Renal cell cancers

Ki-67

artificial intelligence

deep learning

Mobilenetv3-large

Renal cell carcinoma (RCC) is the most common malignant tumor in the urinary system, with an increasing incidence year by year [1–2]. In the United States, 76,080 new cases of RCC were diagnosed in 2021, and there were 13,780 related deaths[3]. RCC may have an etiological association with factors such as smoking, chemical exposure, unhealthy diet, obesity, chronic diseases, and heredity. The disease denotes cancer originated from the renal tubular epithelial cells and includes three major subtypes: clear cell, papillary, and chromophobe RCC, of which clear cell RCC is most common [4–5]. Treatments for RCC include surgery, immunotherapy, and targeted therapy, but they all come with high recurrence rates and poor prognosis for metastatic RCC.

Ki-67 is a biomarker for assessing tumor cell proliferation and correlates with the malignancy and prognosis of RCC. And its high expression is linked to poor prognosis in a wide range of cancers, including breast cancer and prostate cancer. In RCC, Ki-67 expression, a key indicator for determining prognosis, is remarkably correlated with survival and pathological grading [6]. Recent years have witnessed unprecedented advancements in Artificial Intelligence (AI) technology, and its application in the healthcare field is promising. AI, including machine learning, deep learning and neural networks, facilitates the analysis of complex data and improves diagnostic accuracy and work efficiency [7]. We designed this study to construct noninvasive models for predicting Ki-67 expression using deep learning technology based on preoperative CT images of RCC patients, screen the optimal model and apply it to the clinic, aiming to improve the accuracy of prognostic judgment of RCC patients and formulate personalized treatment plans.

Study population

A retrospective analysis was performed for patients who underwent abdominal CT plain scan + enhancement scans at the Affiliated Hospital of Hebei University, were surgically treated at the Department of Urology of the hospital, and were pathologically diagnosed with renal cell carcinoma (RCC) after surgery from January 2019 to November 2019. Eventually 137 patients were enrolled after strict screening with inclusion and exclusion criteria, including 124 cases of renal clear cell carcinoma, 7 cases of chromophobe renal cell carcinoma, 4 cases of papillary renal cell carcinoma, 1 case of collecting duct renal cell carcinoma, and 1 case of sarcomatoid renal cell carcinoma.

Inclusion criteria

(1) Having complete clinical data; (2) Having preoperative four-phase CT scan images (including plain scan phase, arterial phase, venous phase, and excretion phase) from the Affiliated Hospital of Hebei University with high image quality and no artifacts; (3) Being diagnosed for the first time and not having undergone any form of treatment such as radiotherapy, chemotherapy, immunotherapy and surgical treatment prior to the CT scan; and (4) Being scanned by the same CT scanning equipment with the same scanning parameters.

Exclusion criteria

(1) Not having CT plain scan + enhancement scans in our hospital before surgery or not having pathologic examination results; (2) Having unclear CT scan images and poorly displayed lesions; (3) Have a tumor that has developed metastases; and (4) Have missing or incomplete clinical data.

Some existing studies [8–10] defined Ki-67 ≥ 10% as high grade and Ki-67 < 10% as low grade: ① There were 35 cases in the high-grade group, including 28 males and 7 females, aged 28–75 years old, with a mean age of (58.00 ± 8.94); 14 cases were on the left side and 21 cases were on the right side, and the maximum diameter of the tumors ranged from 1.8–13.0cm, with a mean maximum diameter of 5.11 ± 2.57; ② There were 102 cases in the low-grade group, including 57 males and 45 females, aged 30–76 years old, with a mean age of (57.46 ± 10.84); 46 cases were on the left side and 56 cases were on the right side, and the maximum diameter of the tumors ranged from 1.0–12.5cm, with a mean maximum diameter of 4.35 ± 2.39. The details of the clinical data are shown in Table 1.

Table 1

Clinical characteristics of 137 patients with renal carcinoma
Characteristics	Low-grade (Ki-67 < 10%) (n = 102)	High-grade (Ki-67 ≥ 10%) (n = 35)	P
Gender			0.011
Male	57	28
Female	45	7
Age (yrs.)	57.46 ± 10.84	58.00 ± 8.94	0.79
Tumor side			0.60
Left	46	14
Right	56	21
Maximum diameter of tumor (cm)	4.35 ± 2.39	5.11 ± 2.57	0.113
Pathologic type			0.746
Renal clear cell carcinoma	92	32
Chromophobe renal cell carcinoma	6	1
Papillary renal cell carcinoma	3	1
Other types	1	1

Images preparation and segmentation

The initially screened CT images were manually outlined for the region of interest (ROI) of the tumor by two experienced urologists using the labelme tool to form the volume of interest (VOI) of the whole tumor. The outline was based on the maximum diameter of the tumor and extended outward by 1–2 mm. After outlining, the images were saved as a json file, on which the tumor region was cropped according to the annotations in the file, and they were uniformly resized to 224*224 pixels (Fig. 1, Fig. 2).

Model construction and data amplification

The enrolled 137 patients were divided into a model construction group of 110 cases (26 cases of high grade and 84 cases of low grade) and a clinical validation group of 27 cases (9 cases of high grade and 18 cases of low grade). Patients in the model construction group underwent model construction, training and optimization, while those in the clinical validation group underwent clinical validation of model efficacy. The model construction group was further subdivided into the training group of 87 cases (20 cases of high grade and 67 cases of low grade) and the test group of 23 cases (6 cases of high grade and 17 cases of low grade). CT images were categorized and stored by phase and grade, with 3533 images in arterial phase, 3656 images in venous phase, 3278 images in excretion phase, 2619 images in planar phase, and 13086 images in combined four phases in the training group, whereas 944 images in arterial phase, 975 images in venous phase, 931 images in excretion phase, 793 images in planar phase, and 3643 images in combined four phases in the test group. In response to the small number of high-grade patients, high-grade CT images in the training group were augmented by brightness enhancement, random flip, and 20° random rotation, and the test group was augmented by brightness enhancement. After augmentation, the training group had 5372 images in the arterial phase, 5558 images in the venous phase, 5090 images in the excretion phase, 3999 images in the plain scan phase, and 20019 images in the combined four phases; the test group had 1187 images in the arterial phase, 1234 images in the venous phase, 1167 images in the excretory phase, 1005 images in the plain phase, and 4593 images in the combined four phases.

Network model selection and training

The CT data of each phase of the training group were input into the Mobilenetv3-large network for network training, and the plain scan phase model, arterial phase model, venous phase model, excretion phase model and combined four-phase model were constructed respectively. Adaptive moment estimation (Adam) and cross-entropy loss function were used for all network training. The final hyperparameters of the trained models in this study were learning rate = 0.0001, batch size = 16, and number of optimizations = 100.

All models were created using the Python 3.7 programming language and subsequently compiled and trained using Pytorch 1.10 and Cuda 11.2. Desktop workstations with RTX 3060 GPUs were used for these models and Jupyter software was used for the integrated development environment (IDE) tool.

Network model testing and optimal model screening

Following the initial construction and training of the models for each phase, their performance was tested using data from the test group. The CT data of each phase of the test group were input into the constructed models of each phase, thus outputting their respective accuracy, precision, sensitivity, specificity, F1-score, and AUC value in terms of prediction. The optimal model is selected based on the evaluation output. Eventually, the optimal model was screened based on the evaluation output.

Clinical validation of the optimal model

To validate the clinical application of the optimal model, we input the corresponding CT data of 27 patients in the validation group into the optimal model and output the diagnostic results of their Ki-67 high and low grades, so as to validate the predictive performance of the model. To begin with, three corresponding CT images of patients in the verification group were randomly selected, which were subsequently pre-processed by labeling, cropping, etc. and imported into the optimal model, and finally their outputs were compared with their actual postoperative pathology results. Consistent results indicate a correct diagnosis, otherwise an incorrect diagnosis. If each patient has more than or equal to 2 correctly diagnosed CT images, its Ki-67 expression is considered to be correctly predicted, in which case the accuracy of prediction and other corresponding results should be calculated.

Statistical analysis

All data were statistically analyzed using the SPSS 26.0 software (SPSS Inc., Chicago, IL, USA). Measurement data were expressed as mean ± standard deviation (Xˉ±S), and the confusion matrix plots and ROC curves of the optimal model were drawn using Python 3.7. The accuracy, precision, sensitivity, specificity, F1-score and AUC value of the model were calculated, and its predictive efficacy was comprehensively evaluated based on the results. An AUC value of 0.5 indicates no clinical diagnostic value, and a larger AUC value between 0.5 and 1 indicates a better predictive efficacy of the model. In the process of clinical verification, the accuracy, sensitivity and specificity of the optimal model to predict the verification group were calculated, and the consistency between the prediction results of the model and the actual pathological results was evaluated using Kappa value. A Kappa value ≤ 0.2 indicates weak consistency, a value between 0.21–0.40 indicates weak consistency, a value between 0.41–0.60 indicates moderate consistency, a value between 0.61–0.80 indicates significant consistency, and a value between 0.81–0.99 indicates optimal consistency. The 95% confidence intervals were then calculated, and the efficacy of the optimal model for clinical application was assessed combined with all the results obtained.

Predictive model performance results

Following the establishment and training of the predictive models for each of the four phases and the combined four-phase, the CT images of the relevant test groups were used to test the prediction performance of each model. The prediction performance of each model in the prediction of both low- and high-grade groups was tested separately, which was mainly evaluated by the indexes of prediction accuracy, precision, sensitivity, specificity, F1-score and AUC value. The results are shown in Table 2. To exclude the influencing factors of high-grade and low-grade and thus better evaluate the prediction performance among models, we compared the prediction accuracy, precision, sensitivity, F1-score and AUC value of both low- and high-grade groups in each model by summing them up and taking the arithmetic mean. The results are shown in Table 3 and Fig. 3.

Table 2

Predictive performance of models constructed from images in each phase
	Grade	Accuracy	AUC value	Precision	Sensitivity	Specificity	F1-score
Plain scan phase model	Low	0.701	0.752	0.786	0.698	0.706	0.739
Plain scan phase model	High	0.701	0.759	0.602	0.706	0.698	0.650
Arterial phase model	Low	0.767	0.889	0.695	0.976	0.547	0.812
Arterial phase model	High	0.767	0.899	0.955	0.547	0.976	0.696
Venous phase model	Low	0.784	0.824	0.773	0.888	0.640	0.827
Venous phase model	High	0.784	0.822	0.805	0.640	0.888	0.714
Excretion phase model	Low	0.730	0.767	0.720	0.788	0.665	0.701
Excretion phase model	High	0.730	0.783	0.741	0.665	0.788	0.739
Combined four-phase model	Low	0.774	0.871	0.723	0.929	0.598	0.813
Combined four-phase model	High	0.774	0.882	0.882	0.598	0.929	0.713

Table 3

Average performance of models constructed from images in each phase
	Accuracy	Average AUC	Average precision	Average sensitivity	Average F1-score
Plain scan phase model	0.701	0.755	0.694	0.702	0.694
Arterial phase model	0.767	0.894	0.825	0.761	0.754
Venous phase model	0.784	0.823	0.789	0.764	0.770
Excretion phase model	0.730	0.775	0.730	0.726	0.726
Combined four-phase model	0.774	0.877	0.802	0.764	0.763

Predictive model performance analysis

The prediction performance of the models was compared in terms of average prediction performance, as shown in Table 3. The average prediction accuracy, sensitivity, precision, F1-score, and AUC value for the plain scan phase model of Mobilenetv3-large were 0.701, 0.702, 0.694, 0.694, and 0.755, respectively; the above metrics for the arterial phase model of Mobilenetv3-large were 0.767, 0.761, 0.825, 0.754 and 0.894, respectively. For the venous phase model of Mobilenetv3-large, these metrics were 0.784, 0.764, 0.789, 0.770, and 0.823, respectively; and for the excretion phase model of Mobilenetv3-large, these metrics were 0.730, 0.726, 0.730, 0.726, and 0.775, respectively. The above indicators for the combined four-phase model of Mobilenetv3-large were 0.774, 0.764, 0.802, 0.763 and 0.877, respectively.

It was concluded that the venous phase model of Mobilenetv3-large had the highest average accuracy, sensitivity and F1-score, and its average precision and AUC value were also relatively high among the five models and had the best degree of robustness. Therefore, the venous phase model was judged to be the optimal model, which can be used for subsequent validation and analysis of the data in the clinical validation group. The confusion matrix and ROC curve of the venous stage model are shown in Fig. 4.

Clinical validation and analysis of the optimal model

The optimal model, the venous phase model of Mobilenetv3-large, was used for the prediction of Ki-67 expression in the validation group of 27 patients (18 low-grade and 9 high-grade). Three CT images of each patient in the venous phase were randomly selected and input into the venous phase model, and two or more correctly predicted CT images were determined to be correctly predicted. The results show that the model predicted in the validation group with an accuracy of 0.814, a sensitivity of 0.889 and a specificity of 0.667 for predicting the low-grade, and a sensitivity of 0.667 and a specificity of 0.889 for predicting the high-grade. The prediction results are shown in detail in Table 4. The Kappa value of the consistency test between the predicted and actual pathologic diagnosis of Ki-67 expression of the venous phase model of Mobilenetv3-large was 0.57 (P = 0.003), indicating a moderate consistency between the prediction results of the optimal model and the actual pathological diagnosis of the patient, with 95% confidence intervals of (0.24, 0.90).

Table 4

Results of clinical validation of the venous phase model of Mobilenetv3-large
Optimal model prediction results	Postoperative pathological Ki-67 expression results		Total
Optimal model prediction results	Low-grade	High-grade	Total
Low-grade	16	3	19
High-grade	2	6	8
Total	18	9	27

A growing number of early-stage RCCs are being diagnosed with advances in imaging technology. Despite the increase in early intervention treatments, RCC-specific mortality has not significantly improved. This has resulted in the need of a more effective approach to improve patient survival. Artificial intelligence (AI) has been increasingly used in the medical field in recent years. It is in its early stages of application in the field of RCC, but its successful application in other medical fields foretells a great potential in the field of RCC. In this study, a model for preoperative prediction of Ki-67 expression in RCC patients was created and validated by deep learning (DL) technology based on combined four-phase CT images of RCC patients. Thanks to the advantages of non-invasiveness, reduction of complications, and ease of acceptance, the model assists subordinate hospitals in understanding the Ki-67 expression of RCC patients. It is expected to be applied to clinical practice in the future to facilitate the development of individualized treatment plans.

Ki-67 is a broadly recognized marker for tumor prognosis [11–13]. In theory, histopathological changes are well characterized by imaging techniques. Radiomic features can quantify the image pixel and gray scale distribution to mirror molecular pathological changes. In this sense, Ki-67 expression prediction based on CT images is feasible [14]. Studies have been conducted to predict Ki-67 expression using AI technology, primarily centered on other tumors. But in any case this shows the promise of AI technology in the field of RCC [15–17]. In this study, Ki-67 expression was predicted by deep learning (DL) technology based on combined four-phase CT images of RCC patients. The results yielded a prediction accuracy of more than 70% for all five models, and the optimal model, the venous phase model of Mobilenetv3-large, showed good prediction efficacy with average accuracy, sensitivity, precision, F1-score and AUC value of 0.784, 0.764, 0.789, 0.770 and 0.823, respectively. No additional studies were found to similarly utilize AI technology to predict Ki-67 expression in RCC, highlighting a key innovative point of the present study.

Mobilenetv3 is a type of DL convolutional neural network that has achieved satisfactory achievements in the medical field. Take the study of Huang et al. [18] as an example, they constructed a recognition model for digitized pathology slide images of breast cancer by using Mobilenetv3 network combined with bilinear structure, with a classification accuracy as high as 0.88. Mobilenetv3-large, as a branch of the Mobilenetv3 network, prioritizes improving prediction accuracy. We made an initial attempt to model Ki-67 expression prediction in RCC patients using the Mobilenetv3-large network, yielding prediction accuracies above 70% in all cases and up to 78.4% in the venous phase.

There is a wealth of relevant research on AI in the field of RCC, such as the differentiation of benign and malignant renal masses. Baghdadi et al. [19] collected CT images of 212 patients with pathologically diagnosed renal oncocytoma and chRCC and developed a model to discriminate benign renal oncocytoma from chRCC using convolutional neural network with 95% accuracy, 100% sensitivity and 89% specificity. Apart from that, AI also functions in the prediction of pathologic grading of RCC. Xu et al. [20] developed a Fuhrman grading prediction model for ccRCC using DL algorithm. They collected CT images of 706 patients with ccRCC, with 592 patients as the training group and 114 patients as the validation group, and defined patients with grade Ⅰ and Ⅱ as the low-grade group and patients with grade Ⅲ and Ⅳ as the high-grade group. The results yielded an accuracy of 82% for the model with an AUC of 0.882. AI has also been applied in the identification of pathologic types of RCC. Han et al. [21] undertook the first study of classifying RCC subtypes based on DL algorithm, where triphasic CT images of 169 RCC patients were collected and used to train the established DL model. The results yielded an accuracy of 85% for the DL algorithm with an AUC of 0.9. There are even applications of AI in the prediction of pathological staging, gene mutation, and prognosis of renal carcinoma [22–24]. The majority of current studies, however, maintain a focus on the prediction of ccRCC, considering that there are fewer predictions related to pathological immunohistochemistry and that the main pathologic type of RCC is ccRCC. Though most of the patients enrolled in this study were suffering from ccRCC whereas few patients with pRCC and chRCC were enrolled, we have covered the pathologic types of ccRCC, pRCC, and chRCC, resulting in a more comprehensive variety of predictions, which may be more suitable for clinical application. In the future, more enrolled patients as well as pRCC and chRCC patients are needed to further improve the predictive efficacy of the AI models so that they are more clinically applicable.

Data volume is influential on the performance of DL technology applied to image analysis, i.e., an increase in data volume improves the model prediction performance [25]. Data enhancement techniques allow augmentation of data without changing the data labels and increase image heterogeneity. In this study, the prediction performance was improved by the data augmentation technique because of a lower data volume in the high-grade group. However, the high-grade group was still lower than the low-grade group in terms of F1-score and sensitivity after data augmentation. This suggests that while data augmentation techniques may compensate for data deficiencies, there is still a need to collect more high-grade patient data to make the data volume more balanced and improve predictive performance.

There are still certain limitations to this study: small patient volume, insufficient data volume, single-center study, RCC type dominated by ccRCC, and potential absence of scientific validity of Ki-67 threshold setting. In this regard, multicenter studies should be conducted in the future in which data from different hospitals and devices are added to improve the generalizability of the model. In addition, MRI images may be employed in the future to improve predictive efficacy.

The Mobilenetv3-large model based on venous-phase CT images enables accurate, efficient, and noninvasive preoperative prediction of Ki-67 expression in renal cell carcinoma for better individualized treatment plans, highlighting a certain applied clinical value.

Authors’ Contributions

Shen D and Cui ZY had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. Shi BY and Ma T contributed to the study concept and design. Zhou B, Gao C contributed to the acquisition of data, analysis, and interpretation of data. Liu YB did the critical revision of the manuscript for important intellectual content. All authors made a significant contribution to the conception, study design, surgery, acquisition of data, analysis, and interpretation, or in all these areas; took part in drafting, revising, and critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Acknowledgements

We are very grateful to the patients who agreed to provide clinical data to us.

Funding

This study is supported by Medical Science Foundation of Hebei University (No.: 2021X07).

Conflict of interest

All authors declare that they have no competing interests and any accompanying images. All authors agreed to publish this work.

Ethical approval and consent to participate

Ethical approval for this study was obtained from the Ethics Committee of The Affiliated Hospital of Hebei University. Since several patients had low literacy levels and did not understand enough about the disease, they authorized their direct relatives as legal representatives. Informed consent was obtained from all patients or their appointed representatives in accordance with the ethics committee guidelines. This study was conducted under the Declaration of Helsinki.

Consent for publication

Written informed consent was obtained from patients for the publication of this manuscript.

Data availability

The datasets generated and analysed during the current study are not publicly available due to privacy but are available from the corresponding author on reasonable request.

Sung H, Ferlay J, Siegel R L, et al (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries [J]. CA Cancer J Clin 71(3):209-249.
Siegel R L, Miller K D, Fuchs H E, et al (2021) Cancer Statistics, 2021 [J]. CA Cancer J Clin 71(1):7-33.
Pullen R L, Jr (2021) Renal cell carcinoma, part 1 [J]. Nursing 51(7):34-40.
Miller D D, Brown E W (2018) Artificial Intelligence in Medical Practice: The Question to the Answer? [J]. Am J Med 131(2):129-133.
Cui S, Tseng H H, Pakela J, et al (2020) Introduction to machine and deep learning for medical physicists [J]. Med Phys 47(5):e127-e147.
Chrabańska M, Rynkiewicz M, Kiczmer P, et al (2021) Expression of CD44, MMP-2, MMP-9, and Ki-67 as the Prognostic Markers in Non-Clear Cell Renal Cell Carcinomas-A Prospective Cohort Study [J]. J Clin Med 11(17) pii: jcm11175196. doi: 10.3390/jcm11175196.
Xu L, Yang C, Zhang F, et al (2022) Deep Learning Using CT Images to Grade Clear Cell Renal Cell Carcinoma: Development and Validation of a Prediction Model [J]. Cancers (Basel) 14(11): 2574.
Lin F, Cui E M, Lei Y, et al (2019) CT-based machine learning model to predict the Fuhrman nuclear grade of clear cell renal cell carcinoma [J]. Abdom Radiol (NY) 44(7):2528-2534.
Ding J, Xing Z, Jiang Z, et al (2018) CT-based radiomic model predicts high grade of clear cell renal cell carcinoma [J]. Eur J Radiol 03:51-56.
Lin F, Ma C, Xu J, et al (2020) A CT-based deep learning model for predicting the nuclear grade of clear cell renal cell carcinoma [J]. Eur J Radiol 129:109079.
Zheng Z, Chen Z, Xie Y, et al (2021) Development and validation of a CT-based nomogram for preoperative prediction of clear cell renal cell carcinoma grades [J]. Eur Radiol 31(8):6078-6086.
Sasaguri K, Takahashi N (2018) CT and MR imaging for solid renal mass characterization [J]. Eur J Radiol 99:40-54.
Cheville J C, Lohse C M, Zincke H, et al (2003) Comparisons of outcome and prognostic features among histologic subtypes of renal cell carcinoma [J]. Am J Surg Pathol 27(5):612-624.
Shinagare A B, Krajewski K M, Braschi-Amirfarzan M, et al (2017) Advanced Renal Cell Carcinoma: Role of the Radiologist in the Era of Precision Medicine [J]. Radiology 284(2):333-351.
Zhang G M, Shi B, Xue H D, et al (2019) Can quantitative CT texture analysis be used to differentiate subtypes of renal cell carcinoma? [J]. Clin Radiol 74(4):287-294.
Kocak B, Yardimci A H, Bektas C T, et al (2018) Textural differences between renal cell carcinoma subtypes: Machine learning-based quantitative computed tomography texture analysis with independent external validation [J]. Eur J Radiol 107:149-157.
Han S, Hwang S I, Lee H J (2019) The Classification of Renal Cancer in 3-Phase CT Images Using a Deep Learning Method [J]. J Digit Imaging 32(4):638-643.
Huang J, Mei L, Long M, et al (2019) BM-Net: CNN-Based MobileNet-V3 and Bilinear Structure for Breast Cancer Detection in Whole Slide Images [J]. Bioengineering (Basel) 9(6):261.
Baghdadi A, Aldhaam N A, Elsayed A S, et al (2020) Automated differentiation of benign renal oncocytoma and chromophobe renal cell carcinoma on computed tomography using deep learning [J]. Bju International 125(4):553-560
Xu L, Yang C, Zhang F, et al (2022) Deep Learning Using CT Images to Grade Clear Cell Renal Cell Carcinoma: Development and Validation of a Prediction Model [J]. Cancers (Basel) 14(11):2574.
Han S, Hwang S I, Lee H J (2019) The Classification of Renal Cancer in 3-Phase CT Images Using a Deep Learning Method [J]. J Digit Imaging 32(4):638-643.
Lee S, Jung J, Park I, et al (2019) A deep learning and similarity-based hierarchical clustering approach for pathological stage prediction of papillary renal cell carcinoma [J]. Comput Struct Biotechnol J 18:2639-2646.
Kocak B, Durmaz E S, Ates E, et al (2019) Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning-Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status [J]. AJR Am J Roentgenol 212(3):W55-W63.
Schulz S, Woerl A C, Jungmann F, et al (2021) Multimodal Deep Learning for Prognosis Prediction in Renal Cancer [J]. Front Oncol 11:788740.
Yang H, Lin J, Liu H, et al (2023) Automatic analysis framework based on 3D-CT multi-scale features for accurate prediction of Ki67 expression levels in substantial renal cell carcinoma [J]. Insights Imaging 14(1):130.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Deep Learning Model Based on CT Images to Predict Ki-67 Expression in Renal Carcinoma

Status:

Version 1

Abstract

Figures

Introduction

Material and Methods

Study population

Images preparation and segmentation

Model construction and data amplification

Network model selection and training

Network model testing and optimal model screening

Clinical validation of the optimal model

Statistical analysis

Results

Predictive model performance results

Predictive model performance analysis

Clinical validation and analysis of the optimal model

Discussion

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1