by Christos Evangelou, MSc, PhD – Medical Writer and Editor 

Using a new AI-powered digital pathology method, researchers at Nagasaki University analyzed the patterns of collagen fibers within tumors from patients with colon cancer and identified fibrosis as a predictor of lymph node metastasis with 89.5% accuracy. The method also identified patients likely to experience recurrence after chemotherapy with 100% specificity.

The report was published in Modern Pathology.

Study Rationale

Yuko Akazawa, MD, PhD, Associate Professor at Nagasaki University and the corresponding author of the study, noted that, despite advances in surgical techniques and chemotherapy, colon cancer remains one of the most common malignancies worldwide, with significant variability in outcomes even among patients with the same TNM stage. For patients with T3 colon adenocarcinoma, where cancer has grown through the muscle wall but has not reached nearby organs, current staging systems provide limited information about the risk of lymph node metastases or recurrence after treatment.

“T3 patients’ prognosis varies,” Akazawa explained. “Tumor fibrosis is understudied, despite the fact that fibrosis occupies a significant proportion of colon cancer, sometimes surpassing tumor cells in prevalence; however, current prognostic predictions predominantly focus on TNM factors and the differentiation of cancer cells.”

She further explained that the focus on TNM staging represents a missed opportunity, as additional histological factors such as fibrosis play a crucial role in cancer progression, creating what researchers describe as “highways” that facilitate tumor cell migration to lymphatic vessels and blood vessels, enabling metastasis.

“Going through patients’ slides, I have always wondered why we ignore fibrosis findings in the pathology reports,” Akazawa said. “Probably because it was difficult for our eyes to distinguish the fibrosis patterns.”

Methodology

The research team used the FibroNest platform, a cloud-based AI system that was originally developed to analyze liver fibrosis. This automated platform can quantify 350 different characteristics of tissue fibrosis, analyzing collagen content, fiber morphology, and architectural patterns that are invisible to the human eye.

“The FibroNest system can identify subtle variations in fibrosis patterns that pathologists cannot detect,” noted Akazawa, whose team had previously used this platform to distinguish liver cirrhosis patterns in patients with and without hepatocellular carcinoma.

The team analyzed tissue samples from 38 matched patients with stage T3 moderately differentiated colon adenocarcinoma. Using propensity score matching to eliminate confounding variables, researchers compared 19 patients with lymph node metastases to 19 without metastases. Tissue sections were stained with Azan-Mallory staining, which highlights collagen fibers in blue, and then digitally analyzed at 20× magnification.

The AI system evaluated three categories of fibrosis characteristics: collagen content and general properties, morphometric traits of individual collagen fibers (including length, width, area, and density), and architectural features such as fiber organization and texture patterns. These measurements were integrated into what the researchers termed “Phenotypic Fibrosis Composite Score” or Ph-FCS.

Key Findings

The study showed differences in the patterns of fibrosis between patients with and without lymph node metastases. Using the Ph-FCS, the researchers were able to distinguish between metastatic and non-metastatic cases with 89.5% sensitivity and 89.5% specificity, achieving an area under the curve of 0.95.

In addition, the research team identified kurtosis, a statistical measure describing the distribution of gray-scale intensities in tissue images, as the single most powerful predictor of lymph node metastasis. Akazawa explained that this parameter reflects whether collagen distribution around tumor cells is uniform (high kurtosis) or dispersed (low kurtosis).

“We consider that high kurtosis is indicative of thick and uniform collagen encasing the majority of a tumor,” Akazawa stated. “In such instances, the tumor is distinctly separated from fibrosis, which is likely to result in elevated kurtosis values. On the other hand, when complex fibrosis patterns surround small tumor nests, there is a lower kurtosis value due to increased contact between the tumor and collagen (dispersed pattern), which raises the likelihood of collagen aligning perpendicularly to the tumor cells, which may allow the cancer cells to migrate through the fibrosis highway.”

Predicting Treatment Response

The research team also evaluated whether the patterns of fibrosis could predict treatment outcomes. Among the 19 patients with lymph node metastases who received oxaliplatin-based chemotherapy, the Ph-FCS predicted recurrence with 71.4% sensitivity and 100% specificity, achieving an area under the curve of 0.88.

The study also showed that patients who experienced recurrence within three years had different collagen morphology than those who experienced recurrence after three years of treatment, particularly in the number of fiber branches. “The number of collagen branches was the only individual factor for cancer recurrence,” the researchers noted. “Consequently, branched collagen may enhance the likelihood of such tumor-collagen interaction, potentially resulting in occult distant metastasis at the time of surgical intervention.”

Potential Clinical Implications

Among the seven patients who experienced recurrence, six developed hepatic metastasis, pulmonary metastasis, or both, and one developed peritoneal metastasis. Akazawa emphasized that the ability to identify the high-risk patients before recurrence occurs could inform patient monitoring and treatment planning.

In addition, the ability to identify patients at high risk for lymph node metastasis could inform surgical decision-making, potentially guiding more aggressive lymph node dissection strategies or influencing the choice of neoadjuvant therapy. However, Akazawa acknowledged that larger, prospective studies are needed to verify these findings before clinical integration of Ph-FCS.

Study Limitations and Future Directions

The authors acknowledge that the sample size of this proof-of-concept study was relatively small and that their results need to be validated in larger, independent cohorts. Moreover, the statistical robustness of the quantitative fibrosis traits derived from this cohort needs confirmation in larger and more diverse patient populations.

“Another key limitation is the lack of molecular analysis,” Akazawa noted, emphasizing the need for studies to assess the mechanisms linking fibrosis patterns to tumor behavior and patient outcomes. She added that the research team is planning to study the biological mechanisms underlying their observations before conducting larger validation studies.

The study received financial support from the Network-type Joint Usage/Research Center for Radiation Disaster Medical Science.

References

1. Mine A, Tabuchi M, Nakao Y, et al. Intratumoral Fibrotic Features Are Associated With Lymph Node Metastasis and Recurrence in Patients With Advanced Colon Cancer. Mod Pathol. Published online June 24, 2025. doi:10.1016/j.modpat.2025.100828