The role of radiomics in diagnosing gastrointestinal stromal tumors: a review

Full Text

INTRODUCTION

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal neoplasms of the gastrointestinal (GI) tract that originates from the interstitial cells of Cajal and has an annual incidence of 10–20 cases per million people [1]. The most frequent location is the stomach (50%–60%), followed by the small intestine (30%–40%), colon and rectum (5%–10%), and, rarely, esophagus (<1%) [2].

Most GISTs are driven by activating mutations in the gene encoding the type III receptor tyrosine kinase (c-KIT) or the platelet-derived growth factor receptor alpha (PDGFRA). Over 70% of c-KIT mutations occur in exon 11 (KIT exon 11), which encodes the juxtamembrane domain of the receptor. In 7%–10% of cases, mutations affect exon 9 (KIT exon 9), which is typically found in tumors in the small or large intestine and rarely in gastric tumors. A wild-type variant, accounting for 10%–15% of GISTs, lacks detectable mutations in either of these genes [3].

The clinical presentation of GISTs is variable and depends on tumor size and location. The most common symptom is upper GI bleeding, manifesting as hematemesis or melena in 40%–65% of patients. Other symptoms include abdominal discomfort, early satiety, bloating, abdominal pain, bowel obstruction, and perforation. In some cases, the disease remains asymptomatic and is discovered incidentally [4]. During diagnosis, 15%–50% of patients present with evidence of distant metastases [5].

GISTs are a clinically heterogeneous group of tumors exhibiting varying degrees of malignancy. Their biological behavior ranges from benign to aggressive. Several risk stratification systems have been developed to estimate recurrence risk, predict prognosis, and guide treatment strategies. The most widely used systems include the National Institutes of Health (NIH) criteria and Armed Forces Institute of Pathology (AFIP) criteria [6].

Multiple studies have confirmed that tumor location, size, mitotic count, and tumor rupture are independent prognostic factors of GISTs [1]. Joensuu et al. [7] proposed a modified NIH criteria incorporating these four factors into a risk assessment model that categorizes GISTs into very low, low, intermediate, and high risk. This system is widely accepted as a clinical standard for estimating recurrence risk; however, its application is limited to the postoperative setting, once pathologic specimens are available.

Computed tomography (CT) is the preferred method for preoperative diagnosis, tumor staging, and subsequent follow-up and treatment monitoring [8]. With its high sensitivity for tumor detection, CT enables the identification of lesions across a wide size spectrum, which is essential for early diagnosis and timely treatment [9–12]. Contrast-enhanced CT characteristics, such as homogeneous or heterogeneous enhancement and the presence or absence of necrotic areas, are important markers for differential diagnosis [13–18]. Additionally, CT provides critical information on tumor size, location, and invasion of adjacent structures, which is necessary for accurate staging and treatment planning [19].

However, small GISTs (maximum diameter: <2 cm) present a diagnostic challenge. Although CT remains the primary imaging modality for detecting and evaluating tumors, small lesions often exhibit nonspecific imaging features, complicating their identification. These tumors show low contrast enhancement, making them difficult to visualize against the surrounding soft tissues. In some cases, small GISTs may be misidentified as other tumor types or mistaken for the normal gastric wall [16].

Limitations of current diagnostic modalities such as CT and fine-needle aspiration biopsy (FNAB) significantly affect differential diagnosis accuracy. Despite the widespread use of CT for detecting GI tumors, its limited sensitivity may result in missed small lesions or misinterpretation of nonspecific imaging findings, complicating the differentiation of various subepithelial lesions [16]. Although FNAB is a standard method for obtaining tissue samples for histologic analysis, it also has some disadvantages, primarily the limited quantity of available tissue, which may yield nonrepresentative samples and diagnostic uncertainty. Moreover, diagnostic errors can occur due to shielding of intact mucosal layers overlying the tumor, which may hinder tissue acquisition [20].

Considering these limitations, developing and implementing more accurate and effective diagnostic approaches for GISTs is necessary.

Radiomics is a promising direction, with the potential to enhance diagnostic accuracy in GIST detection. By analyzing tumor texture features, radiomic techniques may provide additional insights into tumor heterogeneity and composition, thereby improving the ability to differentiate between tumor types and guiding optimal treatment strategies.

SEARCH METHODOLOGY

A comprehensive search of published data was conducted using PubMed, Google Scholar, and eLibrary to provide an in-depth overview of current knowledge in the field of radiomics applied to GISTs. The search strategy incorporated combinations of keywords and subject terms related to radiomics, GISTs, and gastrointestinal tumor diagnostics: гастроинтестинальная стромальная опухоль / gastrointestinal stromal tumor, GIST, диагностика / diagnosis, радиомика / radiomics, компьютерная томография / computed tomography, and endoscopic biopsy.

Eligible publications included peer-reviewed articles in Russian and English published within the past 5 years.

Following the screening process, 25 original studies were included, each presenting recent advances and findings in radiomics for GIST diagnostics.

RADIOMICS APPROACH

Over the years, radiomics has become one of the most rapidly advancing domains in medical diagnostics, particularly in oncology [21]. It integrates image processing technologies, machine learning, and statistical analysis to extract data from medical imaging modalities such as CT, magnetic resonance imaging (MRI), and positron emission tomography (PET) [22]. Radiogenomics enhances this framework by linking imaging features with tumor genotypes [23].

A key feature of radiomics is noninvasive analysis based on data obtained from various imaging techniques (Fig. 1) [24]. These modalities provide high spatial resolution and detailed structural and functional information about GI organs, making them valuable for the diagnosis and monitoring of GISTs [25].

 

Fig. 1. Example of tumor segmentation using texture analysis software (3D Slicer): venous-phase computed tomography scan of a patient with a gastric gastrointestinal stromal tumor.

 

Solid scientific evidence and clinical research outcomes induce increasing interest in applying radiomics for diagnosing GISTs [26]. This review aimed to analyze and synthesize recent advances in radiomics with a focus on its potential utility in diagnosing GISTs. Various image analysis techniques, correlations between radiomic features and molecular or genetic profiles of tumors, the role of multimodal approaches, and current challenges in standardization and validation of radiomics methods were explored [27].

Applications of Radiomics in the Diagnosis of Gastrointestinal Stromal Tumors

Studies exploring radiomics in the context of GIST diagnostics are summarized in Appendix 1.

The reviewed studies employed diverse methods and imaging protocols and varying sets of radiomic features and statistical tools. Most investigations used a retrospective design. The number of included GIST cases ranged from 41 to 1143. In most studies, radiomic features were extracted from CT images (n = 20); in three studies, MRI was used, and in only two studies, features were derived from ultrasound imaging. CT-based radiomic analysis was performed using different enhancement phases: venous phase in nine studies (36%), multiphasic imaging (dual or triple phase) in eight (32%), and unenhanced CT in two (8%). Three-dimensional region of interest (3D ROI) segmentation was used in 80% of studies and two-dimensional (2D) ROI segmentation in 20%; one study did not specify the segmentation approach.

RADIOGENOMICS

Mutation Status Identification

c-KIT 11 gene mutation. Mutations in exon 11 of the c-KIT gene account for >70% of all c-KIT mutations and are considered the primary oncogenic drivers in GISTs. Tumors harboring c-KIT 11 mutations are known to respond most favorably to targeted therapy with imatinib mesylate [28]. Therefore, identifying the mutational profile is critical for patients with GISTs who may benefit from molecularly targeted tyrosine kinase inhibitor therapy. Additionally, the presence of c-KIT 11 mutation has been associated with a worse prognosis [29]. However, mutation profiling requires genotyping based on surgically obtained tissue samples, a process that is typically invasive and expensive. The high costs and invasiveness of genotyping procedures limit their routine use for many patients [30].

Xu et al. (2018) [31] demonstrated the potential of CT texture analysis to noninvasively differentiate GISTs without c-KIT 11 mutations. They identified standard deviation as an independent radiomic predictor for distinguishing non-c-KIT 11-mutated GISTs. Although this study provided a valuable contribution to imaging-based diagnostics, it was limited by a relatively small sample size (69 cases in the training cohort and 17 in the validation cohort, including only 4 non-c-KIT 11 GISTs), which could have influenced the findings. Moreover, the analysis included only 30 radiomic features, limiting the model’s informativeness.

Subsequently, these limitations were addressed in a study by Liu et al. [30], which included a more representative sample, a broader array of radiomic features, and an evaluation of semantic CT image features. Nevertheless, the study had its drawbacks, which included a retrospective design, heterogeneous imaging protocols performed on three different CT scanners, and lack of image preprocessing, all of which may have affected data reproducibility.

In 2023, Guo et al. [32] employed an oversampling technique to increase the sample size of GISTs lacking c-KIT 11 mutations, balancing the dataset. They developed a radiomic nomogram that showed excellent diagnostic performance in the validation cohort, with an area under the curve (AUC) exceeding 0.8, indicating a strong discriminatory capacity between mutation subtypes.

Zhang et al. [33] used a larger patient cohort than that used in previous studies. Radiomic features were extracted from contrast-enhanced CT images and evaluated for their predictive value in identifying specific c-KIT 11 genotypes. The model exhibited strong predictive performance, particularly for identifying c-KIT 11 557/558 deletions. In the training cohort, the AUC values ranged from 0.759 to 0.956, whereas those in the validation cohort ranged from 0.688 to 0.870. The limitations of this study included its retrospective design and the absence of a demonstrated correlation between contrast-enhanced CT findings in patients with c-KIT 11 mutations and the underlying clinical mechanisms of this association.

Understanding the limitations of each study and acknowledging their contributions to the scientific field are essential for guiding future studies and refining methods. This does not diminish the significance of the conducted studies but rather underscores the need for continued research in radiomics and radiogenomics to enhance diagnostic and prognostic tool accuracy in oncology.

c-KIT 9 gene mutation. c-KIT exon 9 mutation is less prevalent than exon 11 mutations, accounting for approximately 9% of GISTs [34]. Tumors with c-KIT 9 mutations are characterized by greater aggressiveness and invasiveness than those harboring c-KIT 11 mutations. Clinical studies have shown that c-KIT 9-mutated GISTs exhibit a distinct therapeutic response compared with exon 11-mutated tumors [35]. Therefore, timely identification of this mutation is critical for accurate diagnosis and optimal treatment planning. In 2023, Wei et al. [36] developed a radiomic nomogram demonstrating a high predictive accuracy for c-KIT 9 mutation, with an AUC of 0.902 and 0.907 in the training and testing cohorts, respectively. The nomogram integrated the radiomics score (Rad-Score) with clinical risk factors such as nongastric tumor location and presence of distant metastases. However, the retrospective design of the study and manual segmentation of images may have introduced bias, and the rarity of this mutation limited the sample size.

Risk Stratification

According to the 2008 NIH criteria, the malignant potential of GISTs is classified into very low, low, intermediate, and high. This classification is widely accepted as the clinical standard for risk stratification and provides a practical tool for predicting recurrence risk [7].

GISTs classified as very low- or low-risk tumors are considered potentially malignant, although they are often managed clinically as benign tumors. In contrast, intermediate- and high-risk GISTs are treated as malignant neoplasms. The treatment involves the use of imatinib mesylate and other agents administered preoperatively or postoperatively to prevent recurrence or metastasis [37].

Histopathologic examination remains the gold standard for assessing the malignant potential of GISTs. It provides key parameters such as tumor size, location, and mitotic index; the latter is often assessed using core needle biopsy. However, this invasive approach carries risks including tumor cell dissemination and hemorrhage. Therefore, early risk stratification is crucial for determining the optimal treatment strategy [38].

Zhang et al. [39] analyzed 140 arterial-phase CT scans of verified GISTs. The results demonstrated high diagnostic performance for preoperative prediction of intermediate- and high-risk tumors, with AUCs of 0.809 and 0.935, respectively. These findings support the feasibility of risk stratification into four prognostic categories.

In 2019, Wang et al. [40] combined the NIH categories into two groups: low-risk (very low to intermediate) and high-risk tumors. Patients with a mitotic index ≤5 per 50 high-power fields were classified as having low mitotic activity, whereas those with a mitotic index >5 were categorized as having high mitotic activity. Two radiomic models were developed: one for predicting tumor malignancy and the other for estimating the mitotic index. Both models demonstrated strong predictive power, with high AUC values indicating robust discriminatory ability between risk categories.

Moreover, several studies have highlighted the considerable potential of CT texture analysis for preoperative prediction of GIST malignancy risk [41, 42].

Combined models. In clinical practice, subjective symptoms and CT findings help clinicians predict the risk stratification of GISTs. CT-based semantic features, such as tumor size, location, margin characteristics, hemorrhage, and necrosis, are commonly used to assess malignancy risk. Several studies have attempted to predict GIST risk stratification using preoperative contrast-enhanced CT with intravenous bolus administration [43–45]. CT parameters such as tumor size, growth pattern, and increased vascular supply have been identified as characteristic features of high-risk tumors.

A Chinese research group developed a nomogram by integrating radiomic features with clinically relevant variables, including tumor size and ulceration. The nomogram achieved AUCs of 0.930 and 0.931 in the training and validation cohorts, respectively, outperforming the standalone radiomics and clinical models [46].

Wang et al. [47] compared four predictive models: two radiomic models based on CT images obtained during the arterial and venous phases; a clinical model incorporating tumor size and the presence of necrosis; and a clinical model based on conventional CT features assessed by five radiologists. Radiomic models based on arterial and venous phases demonstrated superior diagnostic accuracy for predicting malignancy risk, with AUCs significantly higher than those of models relying solely on conventional clinical parameters.

In another study, a combined nomogram for risk stratification incorporating additional clinical data (tumor size and mitotic index) achieved predictive performance nearly equivalent to established clinical standards, with an AUC of approximately 0.965 [48].

Wang et al. [49] introduced a novel combined model integrating demographic data, CT imaging features, radiomic biomarkers, and immunohistochemical characteristics to predict GIST risk. The nomogram included patient sex, Ki-67 proliferation index, lesion morphology, and presence of necrosis. It demonstrated excellent performance in the training and validation cohorts, with AUCs of 0.921 and 0.913, respectively.

Thus, a comprehensive approach, integrating radiomic, clinical, and CT-derived features, enables a more nuanced evaluation of tumor characteristics, including size, location, imaging morphology, and biological behavior. This approach provides a more comprehensive understanding of tumor characteristics. Although clinical and CT features offer limited or nonobvious information, radiomic analysis enables the identification of predictive markers that support more accurate diagnosis. Tumor size and necrosis are recognized as key predictors of malignancy and are directly associated with recurrence and metastatic potential.

Moreover, the integration of multimodal data improves predictive accuracy by mitigating errors inherent to unidimensional analysis. Combined models are more adept at interpreting complex clinical data, enhancing classification performance across varying risk categories.

The models incorporating diverse data sources are more robust to variations in patient populations and imaging conditions, making them more reliable across different clinical settings.

The use of multiple data sources reduces the risk of overfitting, which is a common limitation of models trained solely on radiomic features, particularly when data volume is limited.

Magnetic resonance imaging. MRI enhances tumor characterization by utilizing multiple imaging sequences, facilitating a more accurate evaluation of its biological behavior. Diffusion-weighted imaging (DWI) reflects restrictions in water molecule diffusion and motion. Some studies have demonstrated that DWI texture features may be biomarkers of intratumoral heterogeneity and predict metastatic potential in GISTs [50].

Yang et al. [51] were the first to develop a diagnostic model based on MRI-derived radiomic features for predicting the mitotic index in GIST patients. The nomogram integrated radiomic features with maximum tumor diameter and location, yielding strong discrimination in training (AUC: 0.878) and validation cohorts (AUC: 0.903).

Mao et al. [52] developed radiomic models incorporating three MRI sequences: DWI with apparent diffusion coefficient (ADC) mapping, T2-weighted imaging (T2WI), and T1-weighted imaging (T1WI) as noninvasive methods for malignancy risk stratification of GISTs. The ADC sequence outperformed T1WI and T2WI in identifying high-risk tumors (AUC: 0.85), indicating high diagnostic efficacy.

Additionally, Yang et al. [53] conducted a preliminary study on a hybrid model that combined radiomics and deep learning applied to MRI for predicting the mitotic index. The model showed promising predictive accuracy in the training and test cohorts. This may be attributable to superior soft tissue contrast of the MRI compared with CT, allowing for greater information density and more discriminative deep learning feature extraction.

Ultrasound imaging. Radiomics enhances the diagnostic capabilities of all imaging modalities used in diagnosing GIST. Notably, in China, Zhuo et al. [54] developed the first radiomic model based on two-dimensional transabdominal ultrasound for malignancy risk stratification in GISTs. The resulting nomogram incorporated radiomic and ultrasound-derived predictors, including maximum tumor diameter. Radiomic analysis of ultrasound images demonstrated significant correlation with GIST malignancy risk. This radiomic nomogram outperformed the clinical ultrasound nomogram and standalone radiomic model, achieving an AUC of 0.90 in the validation cohort.

Jia et al. [55] showed that high-risk features identified by CT and endoscopic ultrasound could not predict malignancy potential in GISTs measuring 1–2 cm. Conversely, a radiomic model based on contrast-enhanced CT images identified small GISTs (1–2 cm) with high malignant potential.

Ki-67. The Ki-67 antigen, a marker of cellular proliferation, is widely used in clinical oncology to predict outcomes in aggressive malignancies, including lung cancer, breast cancer, and glioma [56]. In a meta-analysis, Li et al. [57] reported a direct correlation between malignancy risk and Ki-67 overexpression levels, confirming its potential as an additional prognostic indicator for malignant GISTs.

Zhang et al. [58] developed and validated a radiomic nomogram based on non-contrast CT data to predict Ki-67 expression levels preoperatively in patients with GISTs. A multicenter cohort was used to analyze and validate the radiomic features. The combined nomogram, integrating a radiomic feature with tumor size as a clinical variable, demonstrated high predictive accuracy across multiple validation phases (AUC: 0.801 in the training cohort, 0.828 in internal validation, and 0.784 in external validation).

In 2022, Feng et al. [59] introduced a nomogram based on contrast-enhanced CT (arterial and venous phases). Tumor segmentation and radiomic feature extraction were followed by multivariate logistic regression analysis along with CT features (tumor size, growth pattern, and ulceration). However, the study found no significant difference in predictive performance between the radiomics-only (AUC: 0.772) and combined models (AUC: 0.760).

A multivariate analysis by Liu et al. [60] showed that maximum tumor size is an independent risk factor unrelated to Ki-67 levels. Their logistic regression model identified tumor size and a radiomics-based model as independent predictors of high Ki-67 expression. These variables were used to develop a radiomic nomogram. The model based on three-phase contrast-enhanced CT yielded higher AUC values than models using only arterial and venous-phase data. The earlier study by Zhang et al. [58] utilized non-contrast CT and 2D ROI segmentation, which may have limited the assessment of tumor heterogeneity.

Differential Diagnosis

The stomach is the most common site of GISTs; thus, differential diagnosis should distinguish them from other mesenchymal tumors such as leiomyomas and schwannomas.

Zhang et al. [61] developed a combined model using two-dimensional endoscopic ultrasound images and radiomic features to differentiate GISTs from other gastric tumors (leiomyomas and schwannomas). The model demonstrated high sensitivity and specificity of 91% and 90.6%, respectively. Notably, for subepithelial lesions <20 mm, sensitivity and specificity exceeded 91%, with an AUC of 0.96, indicating that the model is highly effective in differential diagnosis and may help avoid unnecessary surgical resection.

A study conducted in the Netherlands aimed to differentiate GISTs from other intra-abdominal tumors of varying origins. Diagnostic accuracy improved significantly when texture-based imaging features were combined with clinical variables such as patient age, sex, and tumor location (AUC: 0.84) [62], further supporting the utility of integrated predictive models.

Assessment of Therapeutic Response in Gastrointestinal Stromal Tumors

Wang et al. [63] developed a radiomic nomogram to predict recurrence-free survival in intermediate- and high-risk patients with GIST undergoing adjuvant chemotherapy. The model demonstrated strong predictive performance, with AUCs of 0.80, 0.84, and 0.76 for 3-, 5-, and 7-year recurrence-free survival, respectively, in the training cohort and 0.78, 0.80, and 0.76 in the validation cohort. The radiomic nomogram outperformed the clinicopathologic nomogram in recurrence-free survival prediction [C-index: 0.864 (95% CI, 0.817–0.911) vs. 0.733 (95% CI, 0.675–0.791)].

In a separate study, Ao et al. [64] constructed a combined nomogram to predict recurrence or metastasis risk based on clinical and radiomic features. The model exhibited high prognostic accuracy, with AUCs of 0.833 in the training cohort and 0.937 in the validation cohort.

CONCLUSION

Although interest in using radiomics to enhance diagnostic accuracy and malignancy risk stratification in GISTs and other intra-abdominal tumors is growing, the number of scientific publications on this topic remains limited. Most existing studies have focused on specific applications of radiomics, such as mitotic index assessment, recurrence risk prediction, and malignant potential estimation in GISTs. Radiomics has substantial potential for improving diagnostic and therapeutic outcomes; however, additional multicenter and multidisciplinary studies are warranted to validate and standardize radiomic approaches.

Radiomics enables the evaluation of microscopic tissue changes that cannot be discerned through conventional imaging interpretation. Despite encouraging preliminary findings, more systematic studies using standardized imaging protocols and analytical algorithms are required to generalize and confirm these results. An important direction for future research is the development of international standards for radiomic studies, which would enable the application, comparison, and generalization of findings across diverse clinical settings.

Radiomics may also help identify genetic and molecular tumor features, which are essential for guiding treatment decisions. Its ability to extract detailed morphologic data from imaging studies opens new opportunities for advancing personalized medicine. This includes distinguishing GIST from other tumor types and assessing their malignant potential more accurately.

Therefore, radiomics has substantial potential in the differential diagnosis and risk stratification of GISTs, surpassing traditional diagnostic approaches. However, broader application of radiomics for comparison with other intra-abdominal tumors requires further research focusing on methodological refinement, standardization, and clinical integration of radiomic models.

ADDITIONAL INFORMATION

Appendix 1. Summary of studies on the use of radiomics in the diagnosis of gastrointestinal stromal tumors. doi: 10.17816/DD631596-4232091

Funding source. This article was not supported by any external sources of funding.

Disclosure of interests. The authors declare that they have no relationships, activities or interests (personal, professional or financial) with third parties (commercial, non-commercial, private) whose interests may be affected by the content of the article, as well as no other relationships, activities or interests over the past three years that must be reported.

Authors’ contribution. E.A. Martirosyan — literature review, collection and analysis of literary sources, writing and editing the article; G.G. Karmazanovsky — research conception and design, approval of the final manuscript, editing the article, responsibility for the integrity of all parts of the article; E.V. Kondratyev, E.A. Sokolova, V.A. Nechaev — editing the article; V.N. Galkin — approval of the final manuscript; E.S. Kuzmina — approval of the final manuscript, advisory support; A.V. Glotov — advisory support. Thereby, all authors provided approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

About the authors

Elina A. Martirosyan

A.V. Vishnevsky National Medical Research Center of Surgery; S.S. Yudin City Clinical Hospital

Author for correspondence.
Email: robatik2009@mail.ru
ORCID iD: 0000-0002-1854-9638
SPIN-code: 8006-8917
Russian Federation, Moscow; Moscow

Grigory G. Karmazanovsky

A.V. Vishnevsky National Medical Research Center of Surgery; The Russian National Research Medical University named N.I. Pirogov

Email: karmazanovsky@ixv.ru
ORCID iD: 0000-0002-9357-0998
SPIN-code: 5964-2369

MD, Dr. Sci. (Medicine), Professor, academician of the Russian Academy of Sciences

Russian Federation, Moscow; Moscow

Evgeny V. Kondratyev

A.V. Vishnevsky National Medical Research Center of Surgery

Email: kondratev@ixv.ru
ORCID iD: 0000-0001-7070-3391
SPIN-code: 2702-6526

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow

Elena A. Sokolova

A.V. Vishnevsky National Medical Research Center of Surgery

Email: elena83.sokolova@yandex.ru
ORCID iD: 0000-0002-5667-7833
SPIN-code: 9197-6568
Russian Federation, Moscow

Valentin A. Nechaev

S.S. Yudin City Clinical Hospital

Email: nechaevva1@zdrav.mos.ru
ORCID iD: 0000-0002-6716-5593
SPIN-code: 2527-0130

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow

Evgeniya S. Kuzmina

S.S. Yudin City Clinical Hospital

Email: kuz011@mail.ru
ORCID iD: 0009-0007-2856-5176
SPIN-code: 9668-5733
Russian Federation, Moscow

Vsevolod N. Galkin

S.S. Yudin City Clinical Hospital

Email: galkinvn2@zdrav.mos.ru
ORCID iD: 0000-0002-6619-6179
SPIN-code: 3148-4843

MD, Dr. Sci. (Medicine), Professor

Russian Federation, Moscow

Andrey V. Glotov

A.V. Vishnevsky National Medical Research Center of Surgery

Email: andrew.glotov@mail.ru
ORCID iD: 0000-0002-6904-9318
SPIN-code: 4947-4382
Russian Federation, Moscow

References

Supplementary files