Preoperative prediction of intrahepatic cholangiocarcinoma lymph node metastasis by means of machine learning: a multicenter study in China

To accurately predict lymph node metastasis in patients with mass-forming intrahepatic cholangiocarcinoma by using CT radiomics features of tumor habitat subregions

BACKGROUND

This study aims to introduce the concept of habitat subregions and construct an accurate prediction model by analyzing refined medical images, to predict lymph node metastasis (LNM) in patients with intrahepatic cholangiocarcinoma (ICC) before surgery, and to provide personalized support for clinical decision-making.

METHODS

Clinical, radiological, and pathological data from ICC patients were retrospectively collected. Using information from the arterial and venous phases of multisequence CT images, tumor habitat subregions were delineated through the K-means clustering algorithm. Radiomic features were extracted and screened, and prediction models based on different subregions were constructed and compared with traditional intratumoral models. Finally, a lymph node metastasis prediction model was established by integrating the features of several subregional models, and its performance was evaluated.

RESULTS

A total of 164 ICC patients were included in this study, 103 of whom underwent lymph node dissection. The patients were divided into LNM- and LNM + groups on the basis of lymph node status, and significant differences in white blood cell indicators were found between the two groups. Survival analysis revealed that patients with positive lymph nodes had significantly worse prognoses. Through cluster analysis, the optimal number of habitat subregions was determined to be 5, and prediction models based on different subregions were constructed. A comparison of the performance of each model revealed that the Habitat1 and Habitat5 models had excellent performance. The optimal model obtained by fusing the features of the Habitat1 and Habitat5 models had AUC values of 0.923 and 0.913 in the training set and validation set, respectively, demonstrating good predictive ability. Calibration curves and decision curve analysis further validated the superiority and clinical application value of the model.

CONCLUSIONS

This study successfully constructed an accurate prediction model based on habitat subregions that can effectively predict the lymph node metastasis of ICC patients preoperatively. This model is expected to provide personalized decision support to clinicians and help to optimize treatment plans and improve patient outcomes.

Research trends and hotspots evolution of artificial intelligence for cholangiocarcinoma over the past 10 years: a bibliometric analysis

Objective

To analyze the research hotspots and potential of Artificial Intelligence (AI) in cholangiocarcinoma (CCA) through visualization.

Methods

A comprehensive search of publications on the application of AI in CCA from January 1, 2014, to December 31, 2023, within the Web of Science Core Collection, was conducted, and citation information was extracted. CiteSpace 6.2.R6 was used for the visualization analysis of citation information.

Results

A total of 736 publications were included in this study. Early research primarily focused on traditional treatment methods and care strategies for CCA, but since 2019, there has been a significant shift towards the development and optimization of AI algorithms and their application in early cancer diagnosis and treatment decision-making. China emerged as the country with the highest volume of publications, while Khon Kaen University in Thailand was the academic institution with the highest number of publications. A core group of authors involved in a dense network of international collaboration was identified. HEPATOLOGY was found to be the most influential journal in the field. The disciplinary development pattern in this domain exhibits the characteristic of multiple disciplines intersecting and integrating.

Conclusion

The current research hotspots primarily revolve around three directions: AI in the diagnosis and classification of CCA, AI in the preoperative assessment of cancer metastasis risk in CCA, and AI in the prediction of postoperative recurrence in CCA. The complementarity and interdependence among different AI applications will facilitate future applications of AI in the CCA field.

Association Between MRI-Based Radiomics Features and Regional Lymph Node Metastasis in Intrahepatic Cholangiocarcinoma and Its Clinical Outcome

BACKGROUND

Regional lymph node metastasis (LNM) assessment is crucial for predicting intrahepatic cholangiocarcinoma (iCCA) prognosis. However, imaging assessment has limitations for identifying LNM.

PURPOSE

To investigate the association between MRI radiomics features, regional LNM status, and prognosis in iCCA.

STUDY TYPE

Retrospective.

SUBJECTS

Two hundred ninety-six patients (male = 197) with surgically confirmed iCCA.

FIELD STRENGTH/SEQUENCE

1.5 T and 3.0 T. DWI, T2WI, and contrast-enhanced T1WI.

ASSESSMENT

Clinical information, radiologic, and MRI-based radiomics features associated with LNM status were collected to establish models. Performance of MRI, PET/CT, and the combined LNM models were compared in training (N = 207) and test (N = 89) datasets. Overall survival (OS) was compared based on LNM status.

STATISTICAL TESTS

The independent features were selected by 5-fold cross-validation. The performance of MRI, PET/CT, and the models was evaluated using the area under receiver operating characteristic curve (AUC). Univariable and multivariable Cox regression were used to identify independent variables for OS. Kaplan-Meier curves were compared with the log-rank test between LNM positive and negative groups. P < 0.05 was considered statistically significant.

RESULTS

Intrahepatic duct dilatation, enhancement pattern, and CA19-9 were independent clinicoradiologic features. The radiomics model was constructed by the independent radiomics features extracted from T2WI and delay phase T1WI. The combined LNM model showed AUC of 0.888, 0.884, and 0.811 in training, validation, and test cohorts with a positive net benefit. PET/CT exhibited similar sensitivity to the combined LNM model (0.750 vs. 0.733, P > 0.999) while the combined LNM model showed higher specificity (0.703 vs. 0.630, P = 0.039) in the test cohort. High risk of regional LNM was significantly associated with worse OS (median: 24 months) than low risk (median: 30 months, P < 0.0001).

DATA CONCLUSIONS

The combined LNM model has the strongest correlation with LNM status for mass-forming iCCA patients.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

Development and validation of a machine-learning model to predict lymph node metastasis of intrahepatic cholangiocarcinoma: A retrospective cohort study

Lymph node metastasis in intrahepatic cholangiocarcinoma significantly impacts overall survival, emphasizing the need for a predictive model. This study involved patients who underwent curative liver resection between different time periods. Three machine learning models were constructed with a training cohort (2010-2016) and validated with a separate cohort (2019-2023). A total of 170 patients were included in the training set and 101 in the validation cohort. The lymph node status of patients not undergoing lymph node dissection was predicted, followed by survival analysis. Among the models, the support vector machine (SVM) had the best discrimination, with an area under the curve (AUC) of 0.705 for the training set and 0.754 for the validation set, compared to the random forest (AUC: 0.780/0.693) and the logistic regression (AUC: 0.703/0.736). Kaplan-Meier analysis indicated that patients in the positive lymph node group or predicted positive group had significantly worse overall survival (OS: p < 0.001 for both) and disease-free survival (DFS: p < 0.001 for both) compared to negative groups. An online user-friendly calculator based on the SVM model has been developed for practical application.

Lymph node metastasis of intrahepatic cholangiocarcinoma: the present and prospect of detection and dissection

Intrahepatic cholangiocarcinoma (ICC) ranks as the second most primary liver cancer that often goes unnoticed with a high mortality rate. Hepatectomy is the main treatment for ICC, but only 15% of patients are suitable for surgery. Despite advancements in therapeutic approaches, ICC has an unfavorable prognosis, largely due to lymph node metastasis (LNM) that is closely linked to the elevated recurrence rates. Consequently, the identification of precise and suitable techniques for the detection and staging of LNM assumes paramount importance for ICC therapy. While preoperative imaging plays a crucial role in ICC diagnosis, its efficacy in accurately diagnosing LNM remains unsatisfactory. The inclusion of lymph node dissection as part of the hepatectomy procedures is significant for the accurate pathological diagnosis of LNM, although it continues to be a topic of debate. The concept of sentinel lymph node in ICC has presented a novel and potentially valuable approach for diagnosing LNM. This review aims to explore the current state and prospects of LNM in ICC, offering a promising avenue for enhancing the clinical diagnosis and treatment of ICC to improve patient prognosis.

Comparison of Machine Learning Models Using Diffusion-Weighted Images for Pathological Grade of Intrahepatic Mass-Forming Cholangiocarcinoma

Is the radiomic approach, utilizing diffusion-weighted imaging (DWI), capable of predicting the various pathological grades of intrahepatic mass-forming cholangiocarcinoma (IMCC)? Furthermore, which model demonstrates superior performance among the diverse algorithms currently available? The objective of our study is to develop DWI radiomic models based on different machine learning algorithms and identify the optimal prediction model. We undertook a retrospective analysis of the DWI data of 77 patients with IMCC confirmed by pathological testing. Fifty-seven patients initially included in the study were randomly assigned to either the training set or the validation set in a ratio of 7:3. We established four different classifier models, namely random forest (RF), support vector machines (SVM), logistic regression (LR), and gradient boosting decision tree (GBDT), by manually contouring the region of interest and extracting prominent radiomic features. An external validation of the model was performed with the DWI data of 20 patients with IMCC who were subsequently included in the study. The area under the receiver operating curve (AUC), accuracy (ACC), precision (PRE), sensitivity (REC), and F1 score were used to evaluate the diagnostic performance of the model. Following the process of feature selection, a total of nine features were retained, with skewness being the most crucial radiomic feature demonstrating the highest diagnostic performance, followed by Gray Level Co-occurrence Matrix lmc1 (glcm-lmc1) and kurtosis, whose diagnostic performances were slightly inferior to skewness. Skewness and kurtosis showed a negative correlation with the pathological grading of IMCC, while glcm-lmc1 exhibited a positive correlation with the IMCC pathological grade. Compared with the other three models, the SVM radiomic model had the best diagnostic performance with an AUC of 0.957, an accuracy of 88.2%, a sensitivity of 85.7%, a precision of 85.7%, and an F1 score of 85.7% in the training set, as well as an AUC of 0.829, an accuracy of 76.5%, a sensitivity of 71.4%, a precision of 71.4%, and an F1 score of 71.4% in the external validation set. The DWI-based radiomic model proved to be efficacious in predicting the pathological grade of IMCC. The model with the SVM classifier algorithm had the best prediction efficiency and robustness. Consequently, this SVM-based model can be further explored as an option for a non-invasive preoperative prediction method in clinical practice.

Application of AI on cholangiocarcinoma

Cholangiocarcinoma, classified as intrahepatic, perihilar, and extrahepatic, is considered a deadly malignancy of the hepatobiliary system. Most cases of cholangiocarcinoma are asymptomatic. Therefore, early detection of cholangiocarcinoma is significant but still challenging. The routine screening of a tumor lacks specificity and accuracy. With the application of AI, high-risk patients can be easily found by analyzing their clinical characteristics, serum biomarkers, and medical images. Moreover, AI can be used to predict the prognosis including recurrence risk and metastasis. Although they have some limitations, AI algorithms will still significantly improve many aspects of cholangiocarcinoma in the medical field with the development of computing power and technology.

New Insights into Surgical Management of Intrahepatic Cholangiocarcinoma in the Era of “Transplant Oncology”

Intrahepatic cholangiocarcinoma (iCCA) represents the second most frequent type of primary liver neoplasm. The diagnosis and treatment of patients with iCCA involves many challenges. To date, surgical resection with negative margins is the main curative option, achieving an acceptable long-term survival. Despite enabling a considerable improvement in the outcome, iCCA recurrence after surgery is still common. Tumor extension and the histological subtype, as well as vascular and lymph node involvements, are key factors used to define the prognosis. In this narrative review, we aimed to discuss the potential benefits of using different surgical strategies in the field of iCCA, including vascular resection, the mini-invasive approach, liver transplantation, the mechanism used to enable future liver remnant augmentation, and lymph node dissection. We also discussed the new protocols developed in the field of systemic treatment, including immunotherapy and molecular targeted therapy. Recent advancements in the diagnosis, surgical treatment, and understanding of tumor biology have changed the landscape in terms of treatment options. Creating a multidisciplinary tumor board is essential to achieving the best patient outcomes. Further investigational trials are required with the intent of tailoring the treatments and establishing the right patient population who would benefit from the use of new therapeutics algorithms.

The Application of Artificial Intelligence to Investigate Long-Term Outcomes and Assess Optimal Margin Width in Hepatectomy for Intrahepatic Cholangiocarcinoma

BACKGROUND

Intrahepatic cholangiocarcinoma (ICC) is associated with poor long-term outcomes, and limited evidence exists on optimal resection margin width. This study used artificial intelligence to investigate long-term outcomes and optimal margin width in hepatectomy for ICC.

METHODS

The study enrolled patients who underwent curative-intent resection for ICC between 1990 and 2020. The optimal survival tree (OST) was used to investigate overall (OS) and recurrence-free survival (RFS). An optimal policy tree (OPT) assigned treatment recommendations based on random forest (RF) counterfactual survival probabilities associated with each possible margin width between 0 and 20 mm.

RESULTS

Among 600 patients, the median resection margin was 4 mm (interquartile range [IQR], 2-10). Overall, 379 (63.2 %) patients experienced recurrence with a 5-year RFS of 28.3 % and a 5-year OS of 38.7 %. The OST identified five subgroups of patients with different OS rates based on tumor size, a carbohydrate antigen 19-9 [CA19-9] level higher than 200 U/mL, nodal status, margin width, and age (area under the curve [AUC]: training, 0.81; testing, 0.69). The patients with tumors smaller than 4.8 cm and a margin width of 2.5 mm or greater had a relative increase in 5-year OS of 37 % compared with the entire cohort. The OST for RFS estimated a 46 % improvement in the 5-year RFS for the patients younger than 60 years who had small (<4.8 cm) well- or moderately differentiated tumors without microvascular invasion. The OPT suggested five optimal margin widths to maximize the 5-year OS for the subgroups of patients based on age, tumor size, extent of hepatectomy, and CA19-9 levels.

CONCLUSIONS

Artificial intelligence OST identified subgroups within ICC relative to long-term outcomes. Although tumor biology dictated prognosis, the OPT suggested that different margin widths based on patient and disease characteristics may optimize ICC long-term survival.