Temporal-spatial evolution of tumor habitat analysis: a bibliometric study on research hotspots and trends in medical imaging (2014–2025)

Introduction

Tumor habitat analysis, as an emerging multi-dimensional research approach for deciphering the tumor microenvironment and its surrounding complex ecosystem (1), has demonstrated critical application value in early tumor diagnosis (2-4), prognostic assessment (5-7) and personalized treatment strategy development (8,9). This arises from the growing demand to unravel tumor oncogenesis and progression mechanisms in the field of oncology, coupled with the rapid advancement of multi-omics technologies and bioinformatics analytical methods. In recent years, as tumor habitat analysis techniques have continued to evolve, the volume of high-quality research in this area has increased exponentially. Numerous research teams worldwide have conducted extensive investigations spanning from fundamental theories to clinical applications, achieving significant milestones and contributing to an increasingly diverse body of literature.

Bibliometrics (10) is a discipline that employs mathematical and statistical methods to conduct quantitative analyses of the external characteristics and internal laws of literature. It focuses on core elements such as the quantity of literature, authors, journals, and keywords. Through statistical analysis of the trends, distribution patterns, and interrelationships of these elements over time, it systematically reveals the developmental context of disciplines, research hotspots, and evolutionary trends. For instance, statistical analysis of high-frequency keywords in a specific field can clearly outline current research hotspots, while examining dynamic changes in literature quantity across different stages enables precise identification of disciplinary characteristics at each phase. As a critical analytical tool in scientific research, bibliometrics provides objective and quantitative evidence for scientific research evaluation, optimal resource allocation, and disciplinary strategic planning. It helps researchers quickly gain a comprehensive understanding of a field’s landscape and scientifically formulate directions for subsequent research.

Therefore, this study employs VOSviewer and CiteSpace software (11) to conduct a comprehensive bibliometric analysis of literature related to habitat analysis in oncology from the Web of Science Core Collection (WOSCC) and PubMed databases since January 2014. The objective is to systematically identify current research hotspots and emerging trends in this field, providing a clear reference for subsequent in-depth studies. We present this article in accordance with the BIBLIO reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1303/rc).

Methods

Data sources and search strategy

The statistical analysis data in this study were sourced from the WOSCC. The search strategy was designed as follows: TS= (“Tumor” OR “Cancer” OR “Neoplasm” OR “Growth” OR “Carcinoma” OR “Leukemia” OR “Malignancy”) AND TS= (“Habitat” OR “Microhabitat”) AND TS= (“Medical imaging” OR “Medical image” OR “Radiological image” OR “Imaging modality” OR “MRI” OR “CT” OR “PET” OR “Ultrasound” OR “X-ray”). For the PubMed database, the search query was: (“Tumor”[Title/Abstract] OR “Cancer”[Title/Abstract] OR “Neoplasm”[Title/Abstract] OR “Growth”[Title/Abstract] OR “Carcinoma”[Title/Abstract] OR “Leukemia”[Title/Abstract] OR “Malignancy”[Title/Abstract]) AND (“Habitat”[Title/Abstract] OR “Microhabitat”[Title/Abstract]) AND (“Medical imaging”[Title/Abstract] OR “Medical image”[Title/Abstract] OR “Radiological image”[Title/Abstract] OR “Imaging modality”[Title/Abstract] OR “MRI”[Title/Abstract] OR “CT”[Title/Abstract] OR “PET”[Title/Abstract] OR “Ultrasound”[Title/Abstract] OR “X-ray”[Title/Abstract]). The search was restricted to English-language articles and reviews published from the database’s inception through April 30, 2025.

To ensure the quality of selected studies, two researchers independently screened articles based on their full content, excluding those irrelevant to “tumor” and “habitat analysis”. After cross-verification and deduplication, a final set of 127 publications was retained. Data were exported in “tab-delimited file” format with “Full records and cited references”, including details such as title, keywords, authors, affiliations, addresses, abstracts, and publication dates for each study.

Statistical analysis

The annual publication output was charted using Microsoft Excel. For the core bibliometric analysis, duplicate publications were first identified and removed using CiteSpace (version 6.4). The subsequent analysis and visualization of networks, including countries/regions, authors, and keywords, were performed using VOSviewer (version 1.6.20).

Results

General results

A total of 484 relevant publications were initially identified, including 313 from the WOSCC database and 171 from the PubMed database. Following the inclusion/exclusion criteria (Figure 1), 357 articles were excluded, including duplicates, conference abstracts, book proceedings, and other irrelevant materials. A final set of 127 publications on tumor habitat analysis was included, comprising 112 original research articles (88.2%) and 15 review articles (11.8%). These 127 studies were contributed by 904 authors from 255 institutions across 18 countries, published in 58 journals, and cited 3,938 references from 1,000 unique journals.

Figure 1 Literature inclusion and exclusion criteria.

Trend analysis of publication volume in tumor habitat analysis research

A statistical analysis of publications from the WOSCC and PubMed databases (from their inception to April 30, 2025) revealed that the first article on tumor habitat analysis was published in 2014. The field experienced a period of slow initial growth, with the annual publication volume first exceeding double digits (11 articles) in 2020. In recent years, the pace of research has grown exponentially, a trend likely propelled by advancements in artificial intelligence (AI). The annual publication volume reached a new peak of 36 articles in 2024, and this rapid growth has continued with 30 articles published in the first four months of 2025, indicating that the field’s output is set to continue its high-speed growth. The overall trend in annual publications is illustrated in Figure 2.

Figure 2 Trend of annual publication output in tumor habitat analysis (2014–2025 March).

Analysis of publication countries in tumor habitat analysis research

The international collaboration landscape was visualized using VOSviewer (Figure 3), where node size reflects publication volume and connecting lines indicate the strength of cooperation.

Figure 3 International co-authorship network map in tumor habitat analysis research.

This analysis highlights the different roles countries play in this research area. In terms of research output, China is the most prolific nation with 76 publications (405 citations). In terms of collaboration and influence, the United States holds a central position. Despite having fewer publications [30], it leads significantly in total citations [1,091] and possesses the highest total link strength (TLS =10).

To further quantify research impact, the average citations per paper were calculated. The USA demonstrated a substantially higher impact with approximately 36.4 citations per paper (1,091/30), compared to China’s 5.3 citations per paper (405/76). This contrast indicates that while China has a clear advantage in the quantity of research, the United States has produced more impactful work, largely driven by its stronger international collaboration network.

Visually, the network map groups the countries into several distinct collaborative clusters. Notably, the largest cluster (in blue) brings together the two leading nations, the USA and China, along with India and Japan. This indicates a significant collaborative axis exists between the top research powers in North America and Asia. A prominent European cluster (in red) was also identified, comprising Italy, Spain, Switzerland, Belgium, and Norway. Other key clusters include a network of the United Kingdom, the Netherlands, and Austria (in green), and another mixed-regional group of Germany, South Korea, and Australia (in yellow). Countries such as Canada (sky blue) and Brazil (purple) form smaller, more distinct clusters.

The map was generated using VOSviewer, based on a country-level co-authorship analysis with the full counting method. Each node represents a country, where node size corresponds to the number of publications; the lines between nodes indicate the strength of collaborative relationships; and different colors represent distinct collaborative clusters.

Analysis of publication authors in tumor habitat analysis research

A co-authorship analysis was conducted on the author level using the full counting method to map the collaborative landscape of researchers in the field. A total of 904 authors were identified. To determine the core authors, Price’s law was applied using the formula:

M≈0.749Nmax[1]

where Nmax is the maximum number of publications by a single author. Our analysis found that the most prolific author had ten publications (Nmax =10). Based on this, the threshold for a core author was calculated to be approximately 2.37 publications (M≈2.37). Authors with more than three publications were therefore defined as core authors. According to this criterion, 39 authors (4.3% of the total) were identified as the core research group driving this field.

The author collaboration network (Figure 4) reveals a clear division between research productivity and scholarly impact. A “productivity hub” is visible as a large, interconnected green cluster centered around the field’s most prolific authors—Young-Hoon Kim; Ho Sung Kim; and Ji Eun Park—each with ten publications. This group also demonstrates strong collaborative activity, ranking highly in TLS. In contrast, an “impact hub” forms a distinct yellow cluster composed of the most highly-cited researchers. This group includes the top-cited author, Mu Zhou (173 citations), alongside other influential figures like Robert A. Gatenby, RJ Gillies, and Balagurunathan (all with 88 citations).

Figure 4 Author collaboration network-based intellectual structure network (VOSviewer) in tumor habitat analysis research.

This distinct separation of the main clusters visually underscores a key finding: the authors driving the highest publication volume are largely different from those whose work has garnered the most citations. Additionally, the network map highlights other smaller, self-contained groups, such as a tight red cluster (Xinhua Hu; Dongming Liu; and Wenbin Zhang, et al.) and a blue cluster (Jie Tian; Yongming Dai; and Yunfei Zhang, et al.), which both appear to be specific and closely-knit research teams operating independently of the main networks.

An author co-citation analysis was conducted to map the intellectual foundations of the tumor habitat analysis field. To select an appropriate threshold, a sensitivity analysis was performed at minimum co-citation counts of 5, 10, and 15 (12). Based on a comparison of key network metrics and the resulting network structures, a threshold of 10 co-citations was ultimately chosen (as detailed in Appendix 1). This value provided the optimal balance between network coverage and interpretability, meeting preset criteria for node count and network structure. This resulted in a network of 51 key researchers who are considered foundational to this domain. The analysis identified several central figures whose work is frequently cited together, forming the intellectual foundation of the field. J Wu emerged as the most influential author with 64 co-citations, followed by Jpb Oconnor (46 co-citations), RA Gatenby (45 co-citations), and RJ Gillies (45 co-citations). The co-citation network, visualized in Figure 5, groups these authors into four primary schools of thought. The blue cluster is notably prominent, featuring the top-cited author, J Wu alongside the highly influential researchers RA Gatenby and RJ Gillies indicating a core focus on the foundational concepts of the tumor microenvironment. Another significant cluster (in red) is led by Jpb Oconnor, JE Park, and Mu Zhou, representing a body of work related to radiomics and imaging biomarkers. The green cluster, featuring authors such as S Napel and R Verma, and the yellow cluster, which includes A Zwanenburg and E Sala represent other key but more specialized research directions. The intellectual connections and groupings of these authors illustrate the key theoretical cornerstones and research directions that have defined the field.

Figure 5 Author co-citation-based intellectual structure network (VOSviewer) in tumor habitat analysis research.

The network was generated using VOSviewer, based on an author-level co-authorship analysis with the full counting method and a clustering resolution of 1.0. Each node represents an author, where node size corresponds to the number of publications; the lines between nodes indicate co-authorship links; and different colors represent primary collaborative clusters.

The network was generated using VOSviewer, covering all authors with a minimum of 10 co-citations, and the analysis was conducted with the full counting method and a clustering resolution of 1.0. Each node represents an author, where node size corresponds to the author’s total co-citations; the lines between nodes represent the frequency of co-citation; and different colors group authors into distinct intellectual clusters or schools of thought.

Analysis of publishing institutions and journals in tumor habitat analysis research

A citation analysis of organizations was conducted in VOSviewer using the full counting method to identify the most influential research institutions (Table 1). The University of Ulsan (South Korea) emerged as a dominant hub for both productivity and collaboration, leading with 11 publications and the highest TLS [261]. Fudan University (China) followed, ranking second in both publications [10] and TLS [151]. In contrast, research impact was led by US institutions. The H. Lee Moffitt Cancer Center & Research Institute was the most influential, accumulating the highest number of citations [256] from just 5 publications. This indicates that while the University of Ulsan is a primary center for research output and networking, key US cancer centers are the main drivers of highly-cited, impactful work in the field. To identify the core journals, a citation analysis of sources was conducted in VOSviewer using the full counting method. The results, summarized in Table 2, reveal a distinction between publication volume and scholarly impact. Frontiers in Oncology was the most productive journal with 15 articles and also possessed the highest TLS [67], indicating its central role in the network. In contrast, Cancers emerged as the most influential journal, accumulating the most citations [193] from fewer publications [9]. This suggests that while Frontiers in Oncology is the primary outlet for research volume, articles published in Cancers have garnered the greatest scholarly impact.

Top 10 cited references in tumor habitat analysis research

An analysis of the most cited literature in clinical research on tumor habitat analysis revealed that the top-cited article is Gillies’ 2016 publication entitled “Quantitative Imaging in Cancer Evolution and Ecology”. This study established a theoretical framework for cancer ecology and evolution, offering a novel perspective for research; it also emphasized the significance of quantitative imaging, providing theoretical support for precision diagnosis and personalized therapy. The top 10 most cited references are listed in Table 3.

Keyword co-occurrence analysis in tumor habitat analysis research

As condensed representations of research themes, keywords reveal the core topics, hotspots, and trends within a field. A co-occurrence analysis of keywords was conducted using VOSviewer with the full counting method. The resulting network map (Figure 6) included all keywords that appeared at least 3 times (21-24). The network consists of 51 nodes and 1,048 links, where a larger node size indicates a higher keyword frequency.

Figure 6 Keyword co-occurrence network map (VOSviewer) in tumor habitat analysis research.

The co-occurrence network showed 51 nodes and 1,048 connections with larger nodes indicating higher keyword frequencies. The top 10 keywords ranked by frequency are Habitat analysis Radiomics MRI Heterogeneity Glioblastoma Machine learning Cancer Survival Breast cancer and Prognosis reflecting current research priorities including habitat characterization imaging technology combined with machine learning tumor heterogeneity exploration specific tumor types (e.g., glioblastoma breast cancer) and prognostic prediction. This indicates a future trend toward interdisciplinary approaches in tumor research integrating microscale habitat analysis with macroscale clinical applications to enhance precision predictive power and personalized diagnosis and treatment across the entire research spectrum.

The network was generated using VOSviewer, based on a keyword co-occurrence analysis with the full counting method, a clustering resolution of 1.0, and including all keywords with a minimum of 3 occurrences. Each node represents a keyword, where node size is proportional to the keyword’s frequency; the lines between nodes represent co-occurrence relationships; and different colors group the keywords into thematic clusters, which correspond to research hotspots.

Thematic classification of included studies

To further clarify how current research hotspots correspond to practical thematic applications, the studies included in this bibliometric analysis were systematically categorized by their main application areas. Table 4 presents a summary of the references grouped under four key thematic subsections—preoperative diagnosis, treatment efficacy evaluation, prognostic prediction, and genotyping—as well as an additional subsection highlighting challenges and future directions.

This thematic classification aligns with the keyword co-occurrence findings, demonstrating how frequently cited research clusters around tumor habitat characterization, radiomics, machine learning integration, specific tumor types (e.g., glioblastoma, breast cancer), and clinical outcome prediction. Together, these results emphasize the interdisciplinary nature of tumor habitat analysis and its growing impact on precision oncology.

Discussion

This study analyzed research literature from the WOSCC and PubMed databases on the application of habitat analysis in medical imaging, covering the period from its initial introduction in 2014 to April 2025. The findings reveal rapid development in this field, with both the number of publications and their citation frequency steadily increasing. Medical imaging generates vast amounts of image data, which provide a solid foundation for AI applications. The explosive growth of imaging and clinical data, together with recent breakthroughs in AI research, has demonstrated substantial potential for AI-powered habitat analysis to be applied in clinical practice. This approach offers new tools for clinicians to identify high-risk patients, tailor precise and personalized treatment regimens, and predict patient outcomes more accurately.

In terms of research distribution, studies on habitat analysis in medical imaging are predominantly concentrated in countries such as China, the United States, and South Korea. Among these, China has produced a wealth of clinical data and has achieved significant practical progress by combining advanced computational techniques with large-scale case studies. Overall, the literature included in our bibliometric analysis focuses mainly on the following key themes.

Application of tumor habitat analysis in preoperative diagnosis

In current clinical practice, early-stage solid tumors are commonly managed through direct surgical resection, while locally advanced tumors often require neoadjuvant chemoradiotherapy before surgery to achieve tumor downstaging, enhance resectability, and reduce the risk of postoperative local recurrence. Within this workflow, habitat analysis based on tumor heterogeneity has demonstrated significant value for improving preoperative prediction and supporting the development of individualized treatment strategies (25-27).

A notable example is the study by Zhang et al. (28) which applied gadoxetic acid–enhanced magnetic resonance imaging (MRI) to cluster tumor and peritumoral tissues into four distinct habitats. By quantifying habitat volume fractions, they noninvasively predicted microvascular invasion (MVI) and relapse-free survival in hepatocellular carcinoma (HCC) patients. Their results highlighted that habitat features within a 3-mm peritumoral margin, along with tumor diameter, were independent risk factors for MVI. Integrating these variables into a nomogram improved the accuracy of preoperative risk stratification.

Similarly, Liu et al. (29) employed a radiomics-based habitat approach to differentiate malignant brain tumor types using preoperative MRI data. By extracting quantitative features from tumor and peritumoral regions, their model improved diagnostic confidence and provided actionable information for surgical planning and clinical decision-making.

Other research has combined multiparametric imaging modalities for a more comprehensive view of tumor heterogeneity. For instance, one study integrated MR perfusion, diffusion, and positron emission tomography (PET) imaging, registering tumor masks in three-dimensional (3D) Slicer and clustering voxel values with Otsu’s algorithm into eight habitats. The resulting habitat features captured distinct perfusion, hypoxia, and diffusion properties that corresponded to histopathological markers such as cellular density and necrosis, offering valuable insights for preoperative assessment of high-grade gliomas (30).

Beyond MRI, computed tomography (CT)-based habitat analysis has also been explored. In one study (31), contrast-enhanced CT volumes were segmented into three habitats—VOIA (heterogeneous/weak enhancement), VOIB (homogeneous/intense enhancement), and VOIC (peritumoral microinvasion zone). A combined model incorporating clinical factors and CT-based habitats showed improved differentiation between chronic obstructive pulmonary disease (COPD)-related peripheral bronchial carcinoma and inflammatory masses. Another example applied a reference tissue-based probabilistic classification method using muscle as a reference to analyze tumor habitats via apparent diffusion coefficient (ADC) and simulated diffusion-weighted imaging (simDWI) parameters. In seven patients with lumbar high-grade myxofibrosarcoma, this method effectively identified distinct regions of protein-rich fluid, hypercellular tissue, and necrosis, and showed good agreement with pathology and FDG-PET findings before and after radiotherapy (32).

In the context of hepatocellular carcinoma-cholangiocarcinoma, Xiao et al. (33) demonstrated the potential of habitat imaging models for predicting tumor composition. Using K-means clustering to classify voxels based on enhancement patterns, they calculated habitat volume fractions and developed an MRI-based model through least absolute shrinkage and selection operator (LASSO) regression. The HCC-like habitat fractions were positively correlated with the actual HCC component percentage and were linked to genes involved in cell migration and metastasis pathways. The model achieved R-squared values of 0.611 in the training set and 0.541 in the validation set, supporting its utility for preoperative diagnosis and risk stratification.

Taken together, these studies highlight the diverse imaging modalities, clustering strategies, and predictive models currently applied in preoperative tumor habitat analysis. By quantifying spatial heterogeneity and linking imaging-derived habitats to key pathological or genetic markers, habitat analysis provides valuable noninvasive biomarkers to guide surgical planning, assess recurrence risk, and enable more personalized treatment decisions.

Application of tumor habitat analysis in efficacy evaluation

Tumor habitat analysis has emerged as an important tool for evaluating treatment efficacy, addressing limitations of traditional response assessment methods that mainly rely on changes in overall tumor size and often overlook intratumoral heterogeneity. By characterizing distinct subregions within the tumor, habitat analysis can reveal variations in treatment response across different habitats, enabling earlier and more precise evaluation of therapeutic outcomes. For example, during radiotherapy or chemotherapy, tumor cells within different habitats may exhibit variable sensitivities to treatment modalities. Monitoring changes in habitat features—such as volume reduction, shifts in metabolic activity, or alterations in cellular density—can therefore provide early predictive information about treatment effectiveness and facilitate timely adjustments to therapeutic strategies. In patients with locally advanced cervical cancer, habitat analysis has been shown to help identify subregions resistant to chemoradiotherapy, thereby guiding more targeted interventions, improving therapeutic efficacy, and reducing the risk of overtreatment or treatment delays (6,34).

Several studies illustrate how habitat analysis contributes to treatment efficacy evaluation across different cancer types. For instance, Chen et al. (35) proposed an innovative habitat radiomics framework to differentiate non-small cell lung cancer from benign inflammatory diseases preoperatively. Their approach combined PET and CT modalities by adaptively generating habitats based on four-dimensional feature vectors, including PET standardized uptake value, CT intensity, and local entropy metrics from both modalities. They compared an adaptive clustering–based habitat model with conventional threshold-based and non-habitat models. The adaptive clustering habitat radiomics model, the area under the curve (AUC) is 0.7270±0.0147 significantly outperformed the conventional models in the test set (P<0.001). Moreover, integrating adaptive habitat features with conventional non-habitat features further improved performance (AUC =0.7329), highlighting the added diagnostic value of PET/CT habitat analysis in preoperative differentiation and treatment planning.

In another example, Benvenuti et al. (6) retrospectively analyzed 22 patients with high-grade osteosarcoma undergoing neoadjuvant chemotherapy. MRI images were acquired before and after treatment, and tumor volumes were automatically subdivided into three clusters using the K-means algorithm. Each cluster represented regions with distinct imaging characteristics. Baseline analysis revealed that certain cluster volumes in the poor response group were significantly larger than in the good-to-moderate response group. After chemotherapy, histogram-based parameters, such as kurtosis, showed strong discriminative ability in distinguishing between responders and non-responders. These findings demonstrate the potential of MRI-based habitat analysis for predicting neoadjuvant chemotherapy response in osteosarcoma patients, providing valuable noninvasive biomarkers to support clinical decision-making. However, the authors noted that larger cohorts are needed to validate these preliminary results.

Collectively, these studies illustrate how tumor habitat analysis offers a more nuanced and spatially resolved assessment of treatment response. By capturing intratumoral heterogeneity and linking it to therapy outcomes, habitat-based imaging biomarkers can improve response prediction, guide personalized treatment adjustments, and ultimately contribute to more precise oncology care.

Application of tumor habitat analysis in prognostic prediction

Tumor habitat analysis, by leveraging advanced imaging modalities and clustering algorithms, provides detailed spatial information on intratumoral heterogeneity, offering promising prognostic biomarkers for various malignancies (36-38). By characterizing distinct subregions within a tumor, habitat analysis reveals biological variations that can be linked to molecular subtypes, survival outcomes, and treatment response, thus supporting more personalized prognostic assessment.

Chaudhury et al. (20) demonstrated this potential in breast cancer by applying texture dynamics analysis to dynamic contrast-enhanced MRI (DCE-MRI) data. They defined four distinct habitats based on contrast enhancement patterns and quantified habitat heterogeneity using texture features. Their findings showed that texture features from the rapid-delayed washout habitat provided high predictive accuracy for ER status and lymph node metastasis after neoadjuvant chemotherapy, emphasizing the clinical relevance of habitat-level heterogeneity for predicting patient prognosis.

Similarly, Lee et al. (39) investigated the association between spatial habitat features derived from multiparametric MRI and survival outcomes in glioblastoma. In a cohort of 74 glioblastoma patients, 27 spatial features were extracted from the identified habitats. Statistical analysis revealed significant correlations between specific habitat features and molecular subtypes as well as 12-month survival status. Their predictive model achieved a sensitivity of 0.86, specificity of 0.64, and overall accuracy of 0.75 for predicting 12-month survival, demonstrating the prognostic utility of spatial habitat characterization in management.

Extending these findings to a larger multicenter context, Del Mar Álvarez-Torres et al. (40) validated the prognostic significance of vascular habitat hemodynamic markers in glioblastoma. Analyzing MRI data from 184 patients across seven European centers using high-throughput screening methodology, they found that maximum relative cerebral blood volume in high-angiogenic tumor, low-angiogenic tumor, and infiltrative peritumoral edema habitats was significantly correlated with overall survival. This study confirmed the robustness of these habitat-specific hemodynamic markers as prognostic biomarkers, with strong inter-center consistency.

In addition, Beig et al. (14) developed a progression-free survival (PFS) prediction model for glioblastoma based on radiomics features derived from Gd-T1-weighted MRI tumor habitats. Processing imaging data from 203 patients, they identified 25 radiomics features related to biological processes such as cell differentiation, adhesion, and angiogenesis. These features were integrated into a risk regression score (RRS) that, when combined with clinical and molecular data, effectively predicted PFS. This work highlights how habitat-informed radiomics can deepen our understanding of tumor biology and improve individualized prognostic evaluation.

Collectively, these studies underscore the potential of tumor habitat analysis to provide robust, reproducible, and biologically meaningful prognostic indicators. By mapping spatial intratumoral heterogeneity and linking it to clinical outcomes, habitat-based models offer valuable tools for risk stratification and treatment planning, supporting precision oncology in diverse tumor types.

Application of tumor habitat analysis in genotyping

Tumor habitat analysis has shown significant potential in advancing tumor genotyping, offering a novel radiological approach that complements molecular assays in the era of precision medicine. By capturing spatial heterogeneity within tumors through advanced imaging and clustering algorithms, habitat analysis provides noninvasive insights into the underlying genetic variations that drive tumor behavior and treatment response.

In glioblastoma, for example, one study (41) extracted radiomics features from distinct habitats—such as the enhancing tumor region, necrotic core, and peritumoral edema—using multiparametric MRI. These features demonstrated strong associations with glioblastoma molecular subtypes and relevant gene expression patterns. In particular, specific texture characteristics correlated with genes regulating cell differentiation, adhesion, and angiogenesis, underscoring the value of habitat-informed radiomics in predicting genotypic profiles noninvasively.

Expanding this application to other tumor types, Veeraraghavan et al. (42) investigated high-grade serous ovarian cancer by integrating habitat-derived radiomics features with clinical-genomic data. In a retrospective study of 75 patients, they developed an integrated radio-genomic marker model that outperformed other approaches in predicting PFS and platinum resistance classification. Importantly, they identified the cluster dissimilarity feature (cluDiss) as a robust marker linked to PFS and relevant biological pathways, suggesting its utility for patient stratification in future multicenter trials.

Similarly, Zhao et al. (43) explored the link between tumor habitats and driver mutations in colorectal cancer (CRC). Using pre-treatment 18F-fluorodeoxyglucose positron emission tomography (¹⁸F-FDG PET) scans, they extracted habitat-specific radiomics features to predict KRAS, NRAS, and BRAF mutation status in 62 patients. Their habitat-based model achieved AUCs of 0.759 and 0.701 in the training and validation cohorts, respectively, demonstrating good calibration and clinical applicability for noninvasive mutation prediction in CRC patients.

These studies collectively highlight how habitat analysis enhances our understanding of the relationship between intratumoral spatial heterogeneity and genotypic diversity. By identifying distinct subregions that reflect variations in cellular composition and microenvironment, habitat-based models offer a more nuanced perspective on tumor biology, which is often difficult to capture with biopsy alone. This improved insight lays the foundation for developing more accurate predictive biomarkers, enabling personalized targeted therapies and more precise patient stratification (9,44,45).

Challenges and future directions in tumor habitat analysis

Despite its significant potential for elucidating the complex intratumoral microenvironment, tumor habitat analysis still faces several major challenges. First, the lack of standardized protocols limits cross-study comparability and validation. This standardization gap is directly reflected in our keyword co-occurrence analysis, where different imaging modality keywords such as ‘CT’, ‘MRI’, and ‘DCE-MRI’ appear as separate, weakly connected nodes rather than a cohesive group. Their disparate network metrics (e.g., ‘MRI’: 74 occurrences, 270 TLS vs. ‘CT’: 8 occurrences, 33 TLS) further suggest a fragmented research front, underscoring the need for more unified data acquisition and processing standard (46-48). Furthermore, administrative and regulatory barriers hinder data sharing in multicenter collaborations, impeding the development of large-scale databases needed for robust statistical analyses.

Looking ahead, the future directions for tumor habitat analysis are clearly indicated by the trends emerging from our bibliometric data. A key trend is the move toward interdisciplinary precision diagnostics. Our analysis provides concrete indicators for this, revealing strong co-occurrence links between keywords from different conceptual clusters, such as ‘Artificial Intelligence’/‘Machine Learning’ and ‘Diagnosis’ (Links =17 & 30). This integration is also supported by the high link strength between ‘Habitat Analysis’ and ‘Diagnosis’ (Link Strength =55), connecting distinct research themes. This emerging interdisciplinarity points to a significant potential for translational impact, a trend also supported by our data. Keywords for emerging technologies with high translational value, such as ‘Artificial Intelligence’ [average publication year (APY): 2022.2], ‘Machine Learning’ (APY: 2022.9), and ‘Habitat Radiomics’ (APY: 2024.6), are among the most recent and highly connected in the field. To realize this potential, future efforts should focus on expanding multicenter collaborations, establishing industry-wide standards for the entire research pipeline, and developing secure data-sharing frameworks using technologies like federated learning. Additionally, developing more efficient and interpretable algorithms, such as those using attention mechanisms or SHapley Additive exPlanations (SHAP) values, will be vital to increasing clinical trust and accelerating the translation of habitat analysis into routine oncology practice.

Limitations

This study has several limitations that should be considered. First, the analysis was restricted to English-language literature from the WOSCC and PubMed databases, which may exclude relevant publications in other languages or from other databases, potentially leading to selection bias. Additionally, bibliometric methods are effective at identifying quantitative trends, influential actors, and network structures, but they do not analyze the intellectual content of the articles in depth. This study maps the “what” (e.g., key topics, authors) but not the “why” or “how” of the scientific discourse; for example, it does not differentiate whether a citation is supportive, critical, or perfunctory. Moreover, the metrics used, such as citation counts, serve as quantitative proxies for research impact but do not definitively equate to research quality, as this analysis does not account for nuances like self-citation or the context of citations. Finally, the literature search for this study concluded in April 2025, and due to the dynamic nature of academic publishing, the most recent articles published after this date were not included.

Conclusions

A comprehensive evaluation of research achievements and trends in tumor habitat analysis reveals its broad application prospects across key clinical areas of oncology, including preoperative diagnosis, treatment efficacy evaluation, prognostic prediction, and genotyping. It holds promise to bring transformative impacts to clinical oncology and facilitate the advancement of precision medicine. However, beyond widely recognized challenges such as multi-modal data integration, data-sharing barriers due to privacy concerns, and poor model interpretability, our analysis specifically highlights critical hurdles in clinical translation and technical standardization that are characteristic of this interdisciplinary field. In summary, while tumor habitat analysis demonstrates enormous potential in clinical oncology, overcoming these challenges is essential for its widespread adoption. Efforts to strengthen multi-modal data integration, improve privacy protection mechanisms to enable data sharing, develop highly interpretable models, and promote clinical translation research are critical to unlocking its full potential, ultimately supporting better patient outcomes and enhanced efficacy in cancer treatment.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the BIBLIO reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1303/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1303/prf

Funding: The study was partially support by Shenzhen Science and Technology Program of China (No. JCYJ20220818101804009), Shenzhen Key Laboratory for Magnetic Resonance Imaging (No. SYSPG20241211173949066), and Guangdong Innovation Platform of Translational Research for Cerebrovascular Diseases.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1303/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are 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.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Napel S, Mu W, Jardim-Perassi BV, et al. Quantitative imaging of cancer in the postgenomic era: Radio(geno)mics, deep learning, and habitats. Cancer 2018;124:4633-49. [Crossref] [PubMed]
2. Cho HH, Kim H, Nam SY, et al. Measurement of Perfusion Heterogeneity within Tumor Habitats on Magnetic Resonance Imaging and Its Association with Prognosis in Breast Cancer Patients. Cancers (Basel) 2022;14:1858. [Crossref] [PubMed]
3. Li M, Ding N, Yin S, et al. Tumour habitat-based radiomics analysis enhances the ability to predict prostate cancer aggressiveness with biparametric MRI-derived features. Front Oncol 2025;15:1504132. [Crossref] [PubMed]
4. Tan R, Sui C, Wang C, et al. MRI-based intratumoral and peritumoral radiomics for preoperative prediction of glioma grade: a multicenter study. Front Oncol 2024;14:1401977. [Crossref] [PubMed]
5. Wu H, Tong H, Du X, et al. Vascular habitat analysis based on dynamic susceptibility contrast perfusion MRI predicts IDH mutation status and prognosis in high-grade gliomas. Eur Radiol 2020;30:3254-65. [Crossref] [PubMed]
6. Benvenuti G, Marzi S, Vidiri A, et al. Prediction of tumor response to neoadjuvant chemotherapy in high-grade osteosarcoma using clustering-based analysis of magnetic resonance imaging: an exploratory study. Radiol Med 2025;130:13-24. [Crossref] [PubMed]
7. Wu J, Gensheimer MF, Zhang N, et al. Tumor Subregion Evolution-Based Imaging Features to Assess Early Response and Predict Prognosis in Oropharyngeal Cancer. J Nucl Med 2020;61:327-36. [Crossref] [PubMed]
8. Wang H, Zeng L, Wu H, et al. Preoperative vascular heterogeneity based on dynamic susceptibility contrast MRI in predicting spatial pattern of locally recurrent high-grade gliomas. Eur Radiol 2024;34:1982-93. [Crossref] [PubMed]
9. Sujit SJ, Aminu M, Karpinets TV, et al. Enhancing NSCLC recurrence prediction with PET/CT habitat imaging, ctDNA, and integrative radiogenomics-blood insights. Nat Commun 2024;15:3152. [Crossref] [PubMed]
10. Gläser J, Glänzel W, Scharnhorst A. Same data—different results? Towards a comparative approach to the identification of thematic structures in science. Scientometrics 2017;111:981-98.
11. van Eck NJ, Waltman L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 2010;84:523-38. [Crossref] [PubMed]
12. He Y, Xie Z, Li J, et al. Global Trends in mHealth and Medical Education Research: Bibliometrics and Knowledge Graph Analysis. JMIR Med Educ 2024;10:e52461. [Crossref] [PubMed]
13. Sala E, Mema E, Himoto Y, et al. Unravelling tumour heterogeneity using next-generation imaging: radiomics, radiogenomics, and habitat imaging. Clin Radiol 2017;72:3-10. [Crossref] [PubMed]
14. Beig N, Bera K, Prasanna P, et al. Radiogenomic-Based Survival Risk Stratification of Tumor Habitat on Gd-T1w MRI Is Associated with Biological Processes in Glioblastoma. Clin Cancer Res 2020;26:1866-76. [Crossref] [PubMed]
15. Zhou M, Hall L, Goldgof D, et al. Radiologically defined ecological dynamics and clinical outcomes in glioblastoma multiforme: preliminary results. Transl Oncol 2014;7:5-13. [Crossref] [PubMed]
16. Ismail M, Hill V, Statsevych V, et al. Shape Features of the Lesion Habitat to Differentiate Brain Tumor Progression from Pseudoprogression on Routine Multiparametric MRI: A Multisite Study. AJNR Am J Neuroradiol 2018;39:2187-93. [Crossref] [PubMed]
17. Zhou M, Chaudhury B, Hall LO, et al. Identifying spatial imaging biomarkers of glioblastoma multiforme for survival group prediction. J Magn Reson Imaging 2017;46:115-23. [Crossref] [PubMed]
18. Algohary A, Shiradkar R, Pahwa S, et al. Combination of Peri-Tumoral and Intra-Tumoral Radiomic Features on Bi-Parametric MRI Accurately Stratifies Prostate Cancer Risk: A Multi-Site Study. Cancers (Basel) 2020;12:2200. [Crossref] [PubMed]
19. Fang M, Kan Y, Dong D, et al. Multi-Habitat Based Radiomics for the Prediction of Treatment Response to Concurrent Chemotherapy and Radiation Therapy in Locally Advanced Cervical Cancer. Front Oncol 2020;10:563. [Crossref] [PubMed]
20. Chaudhury B, Zhou M, Goldgof DB, et al. Heterogeneity in intratumoral regions with rapid gadolinium washout correlates with estrogen receptor status and nodal metastasis. J Magn Reson Imaging 2015;42:1421-30. [Crossref] [PubMed]
21. Zhang X, Estoque RC, Xie H, et al. Bibliometric analysis of highly cited articles on ecosystem services. PLoS One 2019;14:e0210707. [Crossref] [PubMed]
22. Wang Z, Bi H, Li D, et al. Bibliometric Analysis of Exosome Research in Spinal Cord Injury (2000-May 2024). Trends, Collaborations, and Emerging Insights. Drug Des Devel Ther 2025;19:6829-48. [Crossref] [PubMed]
23. Gu W. A bibliometric analysis of programmed cell death in oral cancer literature: research patterns and emerging trends (2000-2024). Discov Oncol 2025;16:585. [Crossref] [PubMed]
24. Zheng W, Wu Y, Wang Y, et al. Bibliometric and visualized analysis of the top-100 highly cited articles on immunotherapy for endometrial cancer. Medicine (Baltimore) 2023;102:e34228. [Crossref] [PubMed]
25. Yang Y, Cheng J, Cui C, et al. Multi-dimensional interpretable deep learning-radiomics based on intra-tumoral and spatial habitat for preoperative prediction of thymic epithelial tumours risk categorisation. Acta Oncol 2025;64:391-405. [Crossref] [PubMed]
26. Zhu Y, Wang J, Xue C, et al. Deep Learning and Habitat Radiomics for the Prediction of Glioma Pathology Using Multiparametric MRI: A Multicenter Study. Acad Radiol 2025;32:963-75. [Crossref] [PubMed]
27. Ma Q, Wang J, Tu Z, et al. Prediction model of axillary lymph node status using an automated breast volume ultrasound radiomics nomogram in early breast cancer with negative axillary ultrasound. Front Immunol 2025;16:1460673. [Crossref] [PubMed]
28. Zhang Y, Yang C, Qian X, et al. Evaluate the Microvascular Invasion of Hepatocellular Carcinoma (≤5 cm) and Recurrence Free Survival with Gadoxetate Disodium-Enhanced MRI-Based Habitat Imaging. J Magn Reson Imaging 2024;60:1664-75. [Crossref] [PubMed]
29. Liu D, Chen J, Ge H, et al. Differentiation of malignant brain tumor types using intratumoral and peritumoral radiomic features. Front Oncol 2022;12:848846. [Crossref] [PubMed]
30. Bailo M, Pecco N, Callea M, et al. Decoding the Heterogeneity of Malignant Gliomas by PET and MRI for Spatial Habitat Analysis of Hypoxia, Perfusion, and Diffusion Imaging: A Preliminary Study. Front Neurosci 2022;16:885291. [Crossref] [PubMed]
31. Zuo W, Li J, Zuo M, et al. Prediction of the benign and malignant nature of masses in COPD background based on Habitat-based enhanced CT radiomics modeling: A preliminary study. Technol Health Care 2024;32:2769-81. [Crossref] [PubMed]
32. Xing S, Freeman CR, Jung S, et al. Probabilistic classification of tumour habitats in soft tissue sarcoma. NMR Biomed 2018;31:e4000. [Crossref] [PubMed]
33. Xiao Y, Huang P, Zhang Y, et al. Component prediction in combined hepatocellular carcinoma-cholangiocarcinoma: habitat imaging and its biologic underpinnings. Abdom Radiol (NY) 2024;49:1063-73. [Crossref] [PubMed]
34. Liu H, Hou CJ, Wei M, et al. High-risk habitat radiomics model based on ultrasound images for predicting lateral neck lymph node metastasis in differentiated thyroid cancer. BMC Med Imaging 2025;25:16. [Crossref] [PubMed]
35. Chen L, Liu K, Zhao X, et al. Habitat Imaging-Based (18)F-FDG PET/CT Radiomics for the Preoperative Discrimination of Non-small Cell Lung Cancer and Benign Inflammatory Diseases. Front Oncol 2021;11:759897. [Crossref] [PubMed]
36. Wang C, Wu F, Wang F, et al. The Association Between Tumor Radiomic Analysis and Peritumor Habitat-Derived Radiomic Analysis on Gadoxetate Disodium-Enhanced MRI With Microvascular Invasion in Hepatocellular Carcinoma. J Magn Reson Imaging 2025;61:1428-39. [Crossref] [PubMed]
37. Li H, Sui Y, Tao Y, et al. Coupling Habitat Radiomic Analysis with the Diversification of the Tumor ecosystem: Illuminating New Strategy in the Assessment of Postoperative Recurrence of Non-Muscle Invasive Bladder Cancer. Acad Radiol 2025;32:821-33. [Crossref] [PubMed]
38. Feng Q, Xiang W, Fan Y, et al. Enhancement of the nontumor component in newly diagnosed glioblastoma as a more accurate predictor of local recurrence location: a multicenter study. Quant Imaging Med Surg 2025;15:299-313. [Crossref] [PubMed]
39. Lee J, Narang S, Martinez J, et al. Spatial Habitat Features Derived from Multiparametric Magnetic Resonance Imaging Data Are Associated with Molecular Subtype and 12-Month Survival Status in Glioblastoma Multiforme. PLoS One 2015;10:e0136557. [Crossref] [PubMed]
40. Del Mar Álvarez-Torres M, Juan-Albarracín J, Fuster-Garcia E, et al. Robust association between vascular habitats and patient prognosis in glioblastoma: An international multicenter study. J Magn Reson Imaging 2020;51:1478-86. [Crossref] [PubMed]
41. Moon HH, Park JE, Kim N, et al. Prospective longitudinal analysis of physiologic MRI-based tumor habitat predicts short-term patient outcomes in IDH-wildtype glioblastoma. Neuro Oncol 2025;27:841-53. [Crossref] [PubMed]
42. Veeraraghavan H, Vargas HA, Jimenez-Sanchez A, et al. Integrated Multi-Tumor Radio-Genomic Marker of Outcomes in Patients with High Serous Ovarian Carcinoma. Cancers (Basel) 2020;12:3403. [Crossref] [PubMed]
43. Zhao H, Su Y, Wang Y, et al. Using tumor habitat-derived radiomic analysis during pretreatment (18)F-FDG PET for predicting KRAS/NRAS/BRAF mutations in colorectal cancer. Cancer Imaging 2024;24:26. [Crossref] [PubMed]
44. Wu J, Meng H, Zhou L, et al. Habitat radiomics and deep learning fusion nomogram to predict EGFR mutation status in stage I non-small cell lung cancer: a multicenter study. Sci Rep 2024;14:15877. [Crossref] [PubMed]
45. Chen K, Sui C, Wang Z, et al. Habitat radiomics based on CT images to predict survival and immune status in hepatocellular carcinoma, a multi-cohort validation study. Transl Oncol 2025;52:102260. [Crossref] [PubMed]
46. Wang X, Xu C, Grzegorzek M, et al. Habitat radiomics analysis of pet/ct imaging in high-grade serous ovarian cancer: Application to Ki-67 status and progression-free survival. Front Physiol 2022;13:948767. [Crossref] [PubMed]
47. Verma R, Correa R, Hill VB, et al. Tumor Habitat-derived Radiomic Features at Pretreatment MRI That Are Prognostic for Progression-free Survival in Glioblastoma Are Associated with Key Morphologic Attributes at Histopathologic Examination: A Feasibility Study. Radiol Artif Intell 2020;2:e190168. [Crossref] [PubMed]
48. Yang Y, Han Y, Zhao S, et al. Spatial heterogeneity of edema region uncovers survival-relevant habitat of Glioblastoma. Eur J Radiol 2022;154:110423. [Crossref] [PubMed]