Careful measurement of tumors on digital scan.

Precision Matters: How Accurate Tumor Measurement Impacts Cancer Treatment

Lena Kashyap monogram
Lena Kashyap in Health & Wellness October 2026 • 5 min read.

"A new study reveals the challenges in manually measuring tumors and how it can affect treatment decisions, especially for liver metastases."


In the fight against cancer, accurately gauging how tumors respond to treatment is critical. For years, doctors have relied on systems like RECIST (Response Evaluation Criteria in Solid Tumors) to measure tumors on medical images and determine if a treatment is working. However, RECIST isn't perfect. One major challenge lies in the variability and potential inaccuracies that can occur when tumors are manually measured on scans.

A recent study published in European Radiology Experimental dives deep into this issue, focusing specifically on the precision of manual two-dimensional (2D) segmentations – essentially, tracing the outline of tumors on CT scans – for lung and liver metastases. Metastases are cancer cells that have spread from the primary tumor to other parts of the body. The goal was to understand how precise these manual measurements truly are and what impact any uncertainty might have on assessing a patient's response to treatment.

The study aimed to evaluate the precision of manual segmentation of lung and liver metastases and to quantify the uncertainty in tumor response assessment. This has significant implications for how we interpret imaging results and make treatment decisions.

AI Search Multiple angles on this topic

How Cancer Data Informs Action

The CDC says cancer data are used to understand rates and trends, support research, measure progress in prevention and control, focus action on disparities, and improve outcomes. Its National Center for Health Statistics provides topic-specific cancer statistics through FastStats. The International Agency for Research on Cancer’s Global Cancer Observatory is an interactive platform presenting global cancer statistics to inform cancer control and research.

Cancer Research Through the Centuries

The National Cancer Institute’s timeline highlights selected milestones across 250 years of efforts to make progress against a disease known to humanity for thousands of years. Cancer research and treatment have moved from early written descriptions of tumors toward modern molecular therapies, with advances in surgery, radiation therapy, and chemotherapy along the way. A 2024 review describes genetic engineering research, monoclonal antibodies, and immune checkpoint inhibitors as part of more recent advances, particularly for advanced or metastatic tumors.

The Challenge of Manual Tumor Measurement

Careful measurement of tumors on digital scan.

The researchers had a group of physicians and scientists manually outline a large number of tumors (520 segmentations) from CT images of both livers and lungs. This was done to mimic the variability you'd see in a real-world clinical setting. The people outlining the tumors didn't consult with each other, each contouring the tumors independently. They followed RECIST guidelines, and then on a preselected slice.

After the segmentations were complete, the researchers extracted key measurements like diameters and areas of the tumors. Using these measurements, they calculated standard deviations and confidence intervals to understand the range of potential error in the manual segmentations. This allowed them to build regression models and determine confidence intervals (95% CI) for tumor size and thresholds for progressive disease (PD) and partial response (PR) according to RECIST 1.1.

  • In lung, uncertainty decreased as the diameter of the segmented tumor increased.
  • In liver, uncertainty increased as the diameter of segmented tumor increased.
  • The accuracy of manual segmentation is not by the experience of the operator
AI Search Multiple angles on this topic

Automated Segmentation for Better Assessment

Recent research explores automated segmentation of organs and tumors in medical images. A CT-based deep-learning and level-set approach notes that standardized automatic segmentation could make therapy-response classification more reliable. Work on dynamic contrast-enhanced MRI uses 4D information with 3D convolution and convolutional LSTM to segment liver tumors; the authors describe potential benefits of volume-based measurements for liver therapy planning, response classification, tumor classification, and survival prediction. Another study argues that accurate organ and lesion segmentation could support diagnosis, therapy planning, and prognosis.

Limits of RECIST Measurement

A study of manual two-dimensional measurements of lung and liver metastases examined their impact on tumor-response assessment using RECIST 1.1. It reports that RECIST has significant limitations related to variability and reproducibility. The source also notes that these limitations may not be independent, underscoring that response assessments based on measurements can be affected by inconsistency.

The study revealed that while inter-observer agreement was generally excellent (meaning the different observers tended to produce similar measurements), there were still notable uncertainties, especially in the liver. In the lungs, the precision of measurements tended to improve as the size of the tumor increased. But in the liver, the opposite was true: larger tumors had greater measurement variability. These findings suggest that experience doesn't affect segmentation, however, manual segmentation remains uncertain for large liver metastases.

What This Means for Cancer Patients and the Future of Tumor Measurement

This study highlights some important limitations of manual tumor measurement, especially when it comes to liver metastases. While RECIST remains a valuable tool, it's essential to recognize that there's inherent uncertainty in the process, and this uncertainty can influence treatment decisions.

AI Search Multiple angles on this topic

Measurement Standards and Clinical Assessment

A multicenter retrospective study published in 2026 set out to evaluate the accuracy and reproducibility of tumor-size measurements from a commercially available deep-learning clinical decision support system. It compared the system with radiologist assessments and pathological reference in resected lung cancer cases. Separately, a task force formed in 1994 by the European Organization for Research and Treatment in Oncology, the U.S. National Cancer Institute, and the National Cancer Institute of Canada Clinical Trials Group reviewed guidelines, including tumor measurement technique.

Integrating Imaging and Molecular Data

A review of precision oncology identifies single-cell spatial multiomics, patient-derived tumor organoids, liquid biopsy, and non-invasive imaging among technologies shaping the field’s development and future outlook. Emerging computational analysis of medical images may extract quantitative imaging biomarkers beyond anatomical tumor delineation from conventional CT, MRI, and PET. Another review says expanding precision-oncology treatment options will require complementary molecular profiling platforms, including whole-transcriptome RNA sequencing, as well as stronger multimodal data integration and variant interpretation to personalize treatment selection and optimize combination therapies.

Cancer’s Wider Effects and Care Challenges

Cancer is described as systemic across its evolutionary course, its interactions with the microenvironment, and communication with the macro-environment. NCI think-tank materials report that innervated tumors can wire into the somatosensory system and affect aspects of systemic physiology through the central nervous system, including cachexia-associated symptoms. Cancer-care challenges also include drug resistance, minimizing side effects, and equitable access to advanced treatments; the cited review emphasizes interdisciplinary collaboration and patient-centered care as important responses.

So, what's the solution? The researchers suggest that these findings justify the continued development and integration of alternative, more quantitative methods for assessing tumor response. This could include things like volumetric assessment (measuring the entire 3D volume of the tumor) or the use of functional imaging techniques, which look at the biological activity of the tumor rather than just its size. Also, computer-aided detection software for tumor response assessment is helpful.

Ultimately, the goal is to move towards more precise and reliable ways to track how tumors are responding to treatment, ensuring that patients receive the most effective and appropriate care possible. These findings underscore the need for further research and development in the field of cancer imaging and response assessment. As technology advances, we can hopefully move towards more objective and accurate methods for measuring tumors and guiding treatment decisions.

Response thresholds and measurement uncertainty

  • 10 mm minimum lung lesion sizeThe paper links a 10 mm minimum lung lesion size in RECIST with improved reproducibility between WHO and RECIST. This gives context for the threshold, rather than measuring the precision of all lung lesions across sizes.[1]
  • 30%RECIST defines partial response as at least a 30% decrease in the sum of target-lesion diameters from baseline. The percentage is relative to that baseline sum, not an absolute change in diameter.[1]
  • 20%RECIST defines progressive disease as at least a 20% increase in the sum of target-lesion diameters from the smallest sum. The reference is the smallest recorded sum, rather than the baseline sum.[1]

Reading the measurement results

The confidence-interval overlaps mean that, for some larger liver lesions, contouring uncertainty can reach across a RECIST response boundary. A measured change near such a boundary may therefore be harder to classify with confidence.[2]

The two contouring approaches produced no significant measurement differences in the reported comparison. The authors also concluded that slight slice-selection variation did not appear to affect manual 2D segmentation; this does not establish that all imaging or measurement choices are interchangeable.[1]

Boundaries of the study's conclusions

  • The authors found that reader experience did not appear to limit precision, but manual response assessment remained uncertain for large liver metastases.[1]
  • The authors caution that variability and reproducibility limitations in RECIST may not be independent, complicating interpretation of measurement inconsistency.[1]
  • The paper notes that volumetric assessment and RECIST are not interchangeable, and neither has demonstrated clinical superiority.[1]
  • The authors describe computer-aided detection as promising, while noting that its generalizability to clinical practice remains uncertain.[1]

Who performed the CT segmentations and how

Study designObservational study[2]
PopulationLung and liver metastases segmented by physicians and scientists on CT images[2]
Sample size520 segmentations[2]
SettingCT images[2]
ComparisonManual segmentations according to RECIST versus on a preselected slice[2]
Main outcomePrecision of manual 2D segmentation and uncertainty in tumour response assessment[2]
p-valuep = 0.034 and p = 0.021[2]
Times cited15[2]

How this work relates to measurement and segmentation research

Later CT research cited standardized automatic segmentation as a possible route to more reliable therapy-response classification. This is a later study’s rationale, not a result demonstrated by the manual-segmentation paper.[3]

A later MRI segmentation paper described potential uses of volume-based information, including liver therapy planning and response classification. Another later segmentation paper discussed possible support for diagnosis, planning, and prognosis; these are proposed applications, not outcomes tested in the source study.[4] [5]

What follow-up comparisons could clarify

  • Would the reported measurement uncertainty and response-boundary overlaps hold in a broader range of clinical images and readers?[2]
  • How would standardized automatic segmentation compare with manual contours for response classification in clinical practice?[1] [3]
  • When volumetric and RECIST assessments are not interchangeable, which differences matter for clinical response assessment?[1]

Terms used in the measurement study

Confidence interval
An interval calculated in the study to express uncertainty around measurements and response limits.[2]

Sources

  1. Precision of manual two-dimensional segmentations of lung and liver metastases and its impact on tumour response assessment using RECIST 1.1 (Europe PMC, 2017)
  2. Precision of manual two-dimensional segmentations of lung and liver metastases and its impact on tumour response assessment using RECIST 1.1 (Springer Science and Business Media LLC, 2017)
  3. Deep learning and level set approach for liver and tumor segmentation from CT scans (2020)
  4. Automatic Liver Tumor Segmentation on Dynamic Contrast Enhanced MRI Using 4D Information: Deep Learning Model Based on 3D Convolution and Convolutional LSTM (2022)
  5. Transformer based Generative Adversarial Network for Liver Segmentation (2022)

About this Article -

Written with AI assistance from published research, and reviewed by the Mystum team. See our About page for more information.

This article is based on research published under:

DOI-LINK: 10.1186/s41747-017-0015-4, Alternate LINK

Title: Precision Of Manual Two-Dimensional Segmentations Of Lung And Liver Metastases And Its Impact On Tumour Response Assessment Using Recist 1.1

Subject: Radiology, Nuclear Medicine and imaging

Journal: European Radiology Experimental

Publisher: Springer Science and Business Media LLC

Authors: F. H. Cornelis, M. Martin, O. Saut, X. Buy, M. Kind, J. Palussiere, T. Colin

Published: 2017-10-30

Everything You Need To Know

1

What is the main focus of the study regarding tumor measurement?

The study focuses on the accuracy of manual two-dimensional (2D) segmentations of tumors, specifically in the context of liver and lung metastases. It investigates the precision of measuring tumors on CT scans, highlighting the variability and potential inaccuracies in this process. The goal is to quantify how much uncertainty exists when assessing a patient's response to treatment based on these manual measurements, particularly using RECIST guidelines. This is crucial because precise tumor measurement directly influences treatment decisions.

2

What is RECIST, and why is it important in this context?

RECIST (Response Evaluation Criteria in Solid Tumors) is a system used by doctors to evaluate how tumors respond to cancer treatments. It involves measuring tumors on medical images, like CT scans, to determine if a treatment is effective. Although valuable, RECIST relies on manual measurements of tumors, which this study points out can have variability and inaccuracies. Recognizing these limitations is important because treatment decisions are often based on RECIST assessments. The study's focus on RECIST helps understand how measurement errors can affect the evaluation of treatment effectiveness, especially for liver and lung metastases.

3

What did the study reveal about the precision of manual tumor measurements?

The study found that manual segmentation, the process of physicians and scientists manually outlining tumors on CT scans, has inherent uncertainties. While there was good agreement among observers, the precision varied. In the lungs, larger tumors showed more precise measurements. However, in the liver, larger tumors had greater measurement variability, highlighting a significant challenge. The study suggests that even though there can be high levels of agreement, manual segmentation introduces uncertainty that can affect the assessment of treatment response. This variability can influence the interpretation of imaging results and treatment decisions.

4

Did the experience of the operator affect the accuracy of manual tumor measurements?

The study's findings suggest that the accuracy of manual segmentation is not affected by the experience of the operator. The study showed that the level of experience did not correlate with improved precision in the measurements, which indicates that even experienced professionals face challenges. This insight is important because it emphasizes the limitations of relying solely on manual measurements for assessing treatment response. The study's results suggest that despite the observer's experience, manual segmentation remains uncertain for large liver metastases.

5

Why are the study's findings important for cancer patients?

This study's findings on manual tumor measurement have implications for cancer patients because the accuracy of tumor measurements directly affects treatment decisions. The study emphasizes the need to recognize the limitations of current methods, especially for liver metastases. While RECIST is used to evaluate treatment effectiveness, the uncertainties identified in manual segmentation can lead to potential misinterpretations of imaging results. This may influence whether a patient is considered to have a partial response or progressive disease, highlighting the importance of improving measurement precision and potentially integrating more automated, precise techniques in the future.

Newsletter Subscribe

Subscribe to get the latest articles and insights directly in your inbox.