Using automated image analysis to identify CRC patients with low-risk in disease-specific death

This is the “Behind the Paper” blog post on our article “Spatial immune profiling of the colorectal tumor microenvironment predicts good outcome in stage II patients”, which was published on May 15, 2020, in npj Digital Medicine.
Published in Healthcare & Nursing
Using automated image analysis to identify CRC patients with low-risk in disease-specific death
Like

A method that integrates lymphocytic infiltration, the spatial relationship between lymphocytes and tumor buds, and the CD68+/CD163+ macrophage ratio could identify a subpopulation of stage 2 colorectal cancer (CRC) patients who exhibited 100% survival over a 5-year follow-up period. This model was found to be superior to the current prognostic staging systems in identifying patients who are at low-risk in disease-specific death.

CRC is the third most commonly diagnosed cancer worldwide. However, CRC is ranked second in mortality rates. The mortality rates have decreased significantly over the years. This is attributed to the improvements in CRC screening programs and treatments, as well as the identification of new risk factors. The most significant determinant of CRC patient survival used clinically, and the best current guide for any therapeutic decision is the tumor stage.

The currently used histological method for CRC staging is the tumor-node-metastasis (TNM), which takes into account tumor size, the involvement of regional lymph nodes, and distant metastasis. Although the TNM staging system is accurate at a population level, it is less so at a personal level as same stage patients can have distinct prognostic outcomes. For instance, even though the majority of stage 2 CRC patients experience good prognostic outcome, approximately 20-30% of these patients experience disease recurrence and poor survival outcome. The TNM system focuses solely on tumor features. However, it is now well established that the tumor microenvironment (TME) plays a key role in tumor progression and patients’ survival outcome. The complex and dynamic TME, consists of various heterogeneous cell subpopulations which interact with and influence each other. Two of the most promising prognostic factors in CRC are the presence of small isolated cancer clusters mainly at the tumor front, termed tumor buds, and lymphocytic infiltration. Another main component of the TME are the macrophages, though depending on their phenotype, they may have pro- or anti-tumor qualities.

In our study, by applying an automated image analysis methodology, we quantified the densities as well as the spatial inter-relationships of these features. Through a machine learning approach, we assessed their prognostic significance as well as that of the features from the clinicopathological report (such as age and gender) and developed a combinatorial prognostic index, termed the Spatial Immuno-Oncology Index (SIOI). The SIOI was able to identify a subpopulation of patients who exhibited 100% survival over a 5-year follow-up period. The combinatorial prognostic model integrated lymphocytic infiltration, the spatial interaction of lymphocytes with tumor buds, and the CD68+/CD163+ macrophage ratio. 

In order to develop this model, data from a training cohort of 113 patients with stage 2 CRC who underwent surgical resection in Edinburgh, UK, between 2002-2003 were used. Our findings were validated in an independent validation cohort, which consisted of 56 patients treated in Edinburgh, UK in 2004, and 61 patients treated in the National Defense Medical College Hospital, Japan, between 2006-2011.

Manual quantification of specific cell subpopulations is very time-consuming and labor-intensive. In addition, the manual assessment of such features can be associated with high inter-observer variability. The methodology applied in this study ensures the standardized and reproducible reporting of these features by employing an automated image analysis approach.

Once validated in large and international cohorts, the key application of our combinatorial prognostic model is in its potential to effectively identify CRC patients with low-risk in disease-specific death and who might therefore not need any further treatment. This would not only have a positive impact on the patients’ quality of life by preventing them from any unnecessary toxic treatment but also to the healthcare providers by reducing their costs.


Acknowledgments: I thank Dr. Peter D. Caie, and our co-authors of Spatial immune profiling of the colorectal tumor microenvironment predicts good outcome in stage II patients.

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Health Care
Life Sciences > Health Sciences > Health Care
  • npj Digital Medicine npj Digital Medicine

    An online open-access journal dedicated to publishing research in all aspects of digital medicine, including the clinical application and implementation of digital and mobile technologies, virtual healthcare, and novel applications of artificial intelligence and informatics.

Related Collections

With collections, you can get published faster and increase your visibility.

Harnessing digital health technologies to tackle climate change and promote human health

This collection invites research on the use of digital health technologies that innovate solutions to improve sustainable health care practice and delivery.

Publishing Model: Open Access

Deadline: Apr 30, 2024