OncoTREAT: a software assistant for cancer therapy monitoring

The Liver Tumor Segmentation Benchmark (LiTS)

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in http://medicaldecathlon.com/. In addition, both data and online evaluation are accessible via https://competitions.codalab.org/competitions/17094.

Threshold for Enhancement in Treated Hepatocellular Carcinoma on MDCT: Effect on Necrosis Quantification

OBJECTIVE

The objective of our study was to determine whether the conventionally used enhancement threshold of 10 HU for assessing tumor viability in treated hepatocellular carcinoma (HCC) lesions is valid.

MATERIALS AND METHODS

To distinguish pseudoenhancement from enhancement in a tumor, we used an in vivo model: The attenuation of 54 hepatic cysts during the unenhanced and portal venous phases of MDCT, similar to what may be observed in HCC with central necrosis, was used to determine the threshold for pseudoenhancement. To validate this model, we compared the attenuation value of liver parenchyma in this cohort with that of 22 HCCs during the late arterial phase of enhancement. We tested the effect of this pseudoenhancement on quantifying necrosis in HCC compared with the conventionally used threshold of 10 HU.

RESULTS

Values of enhancing HCC tissue on arterial phase MDCT (mean, 121.3 HU) were comparable with normal liver parenchyma on venous phase MDCT (117.3 HU) (p = 0.27). The threshold of 17.1 HU was the best threshold for the detection of pseudoenhancement in cysts (99% accuracy, 100% sensitivity, and 98% specificity). When this threshold was used instead of the conventional threshold of 10 HU, the mean necrosis proportion of treated HCC increased from 34.0% to 42.6% and the mean viable tumor proportion decreased from 66.0% to 57.4%. The quantification of viable HCC tissue based on 10 HU and the quantification of viable HCC tissue based on 17.1 HU were found to be significantly different (p < 0.0001).

CONCLUSION

The threshold of 17.1 HU may be the appropriate cutoff for nonenhancement in a necrotic HCC. Use of this threshold may potentially affect how response to therapy is quantified and categorized.

Interactive Volumetry Of Liver Ablation Zones

Percutaneous radiofrequency ablation (RFA) is a minimally invasive technique that destroys cancer cells by heat. The heat results from focusing energy in the radiofrequency spectrum through a needle. Amongst others, this can enable the treatment of patients who are not eligible for an open surgery. However, the possibility of recurrent liver cancer due to incomplete ablation of the tumor makes post-interventional monitoring via regular follow-up scans mandatory. These scans have to be carefully inspected for any conspicuousness. Within this study, the RF ablation zones from twelve post-interventional CT acquisitions have been segmented semi-automatically to support the visual inspection. An interactive, graph-based contouring approach, which prefers spherically shaped regions, has been applied. For the quantitative and qualitative analysis of the algorithm's results, manual slice-by-slice segmentations produced by clinical experts have been used as the gold standard (which have also been compared among each other). As evaluation metric for the statistical validation, the Dice Similarity Coefficient (DSC) has been calculated. The results show that the proposed tool provides lesion segmentation with sufficient accuracy much faster than manual segmentation. The visual feedback and interactivity make the proposed tool well suitable for the clinical workflow.

Segmentation-based partial volume correction for volume estimation of solid lesions in CT

In oncological chemotherapy monitoring, the change of a tumor's size is an important criterion for assessing cancer therapeutics. Measuring the volume of a tumor requires its delineation in 3-D. This is called segmentation, which is an intensively studied problem in medical image processing. However, simply counting the voxels within a binary segmentation result can lead to significant differences in the volume, if the lesion has been segmented slightly differently by various segmentation procedures or in different scans, for example due to the limited spatial resolution of computed tomography (CT) or partial volume effects. This variability limits the sensitivity of size measurements and thus of therapy response assessments and it can even lead to misclassifications. We present a fast, generic algorithm for measuring the volume of solid, compact tumors in CT that considers partial volume effects at the border of a given segmentation result. The algorithm is an extension of the segmentation-based partial volume analysis proposed by Kuhnigk for the volumetry of solid lung lesions , such that it can be applied to inhomogeneous lesions and lesions with inhomogeneous surroundings. Our generalized segmentation-based partial volume correction is based on a spatial subdivision of the segmentation result, from which the fraction of tumor for each voxel is computed. It has been evaluated on phantom data, 1516 lesion segmentation pairs (lung nodules, liver metastases and lymph nodes) as well as 1851 lung nodules from the LIDC-IDRI database. The evaluations of our algorithm show a more accurate estimation of the real volume and its ability to reduce inter- and intra-observer variability significantly for each entity. Overall, the variability (interquartile range) for phantom data is reduced by 49% ( p ≪ 0.001) and the variability between different readers is reduced by 28% ( p ≪ 0.001). The average computation time is 0.2 s.

Hepatic tumors: region-of-interest versus volumetric analysis for quantification of attenuation at CT

PURPOSE

To evaluate the reproducibility of liver tumor attenuation measurement performed by using the routinely used manual region-of-interest (ROI) method and that of measurement performed by using a semiautomated volumetric approach at computed tomography (CT).

MATERIALS AND METHODS

This HIPAA-compliant retrospective study had institutional review board approval. The requirement for patient informed consent was waived. Attenuation of colon cancer liver metastases in 208 patients was measured on portal venous phase multidetector CT images by using a single ROI, the average measurement in three ROIs on a single section, and with semiautomated segmentation of the entire tumor volume (volumetric attenuation) to evaluate intermethod agreement. Intraobserver and interobserver reproducibility were evaluated in the first 70 patients. Measurements were repeated after 30 days to assess intraobserver reproducibility. Differences between methods were tested by using repeated-measures analysis of variance. Intermethod, intraobserver, and interobserver agreements were tested by using Bland-Altman analysis and the Lin concordance correlation coefficient (ρc). P < .05 was considered to indicate a significant difference.

RESULTS

A total of 208 pathologically proven colon cancer hepatic metastases larger than 20 mm in diameter in 100 women and 108 men (mean age, 61.6 years ± 11.6 [standard deviation]; range, 28-87 years) were evaluated. Attenuation was significantly different between the three methods of measurement (P < .001 for all). Volumetric measurements had better intraobserver agreement (precision = 3.3%, ρc = 0.996, P < .001) than single-ROI measurements (precision = 12.0%, ρc = 0.947, P < .001) and measurements averaged over three ROIs (precision = 9.3%, ρc = 0.965, P < .001). Volumetric measurements also had better interobserver agreement (precision = 3.6%, ρc = 0.993, P < .001) than single-ROI measurements (precision = 11.3%, ρc = 0.957, P < .001) and the average measurement in three ROIs (precision = 8.5%, ρc = 0.976, P < .001).

CONCLUSION

Measurements of hepatic tumor attenuation at multidetector CT are reproducible. An approach based on the evaluation of whole-lesion attenuation demonstrated better reproducibility than ROI measurements.

MDCT necrosis quantification in the assessment of hepatocellular carcinoma response to yttrium 90 radioembolization therapy: comparison of two-dimensional and volumetric techniques

RATIONALE AND OBJECTIVES

The purpose of this study is to evaluate the reproducibility and agreement of tumor necrosis quantification performed by two-dimensional and volumetric methods in a cohort of patients with hepatocellular carcinoma (HCC) treated with yttrium-90 ((90)Y) radioembolization.

MATERIALS AND METHODS

Twenty-nine consecutive patients (21 men, 8 women; mean age 66.6 years; age range, 44-90 years) with HCC treated with (90)Y radioembolization that underwent liver multidetector computed tomography (MDCT) were included. Two independent radiologists evaluated the necrosis proportion of the lesions with two-dimensional (2D) measurements according to the European Association for the Study of the Liver guidelines, and with a volumetric method using a voxel-by-voxel analysis. Interobserver reproducibility for each method was assessed by using within-subject coefficients of variation (WSCV), intraclass correlation coefficients (ICC), and Lin's concordance correlation coefficients (LCC). Agreement between both methods was assessed by using the Bland-Altman plot and the paired t-test.

RESULTS

The volumetric method was more reproducible (WSCV = 27.8%; ICC = 0.914; LCC = 0.909) than the 2D (WSCV = 43.8%; ICC = 0.723; LCC = 0.841). There was a significant difference in the mean calculated necrosis proportions based on 2D and volumetric methods (P = .0129).

CONCLUSION

Voxel-by-voxel quantification of HCC necrosis is a more reproducible method than 2D analysis.

Evaluation of hepatocellular carcinoma size using two-dimensional and volumetric analysis: effect on liver transplantation eligibility

RATIONALE AND OBJECTIVES

Milan criteria recommends selection of candidates with hepatocellular carcinoma (HCC) for liver transplantation based on strict tumor size thresholds. The purpose of this study is to compare the effect of two-dimensional and three-dimensional tumor measurements on the selection of candidates for liver transplantation using Milan criteria.

MATERIALS AND METHODS

This retrospective Health Insurance Portability and Accountability Act-compliant study was approved by our institutional review board. Patient-informed consent was waived. Forty-five HCCs in 19 patients, evaluated with triphasic multidetector row computed tomography scans, were included in the analysis. The largest diameters in each two-dimensional orthogonal plane (Max2D) and within three-dimensional tumor boundaries (Max3D) were calculated for each lesion. Diameters were compared and the eligibility based on lesion size for liver transplantation was assessed.

RESULTS

The mean Max2D diameter of HCC was 3.2 ± 0.9 cm and the mean Max3D diameter was 3.5 ± 1.2 cm. There was a significant difference between the mean Max2D and Max3D diameters (P < .001). Among the 45 lesions, 22 of them (48.9%) were ineligible for transplantation according to Max2D diameter, whereas 29 of them (64.44%) were ineligible when Max3D diameter was applied (P < .001).

CONCLUSION

HCC diameter based on 3D measurements is significantly different than the conventional 2D measurements and may affect eligibility for liver transplantation.

Object-based analysis of CT images for automatic detection and segmentation of hypodense liver lesions

PURPOSE

Hypodense liver lesions are commonly detected in CT, so their segmentation and characterization are essential for diagnosis and treatment. Methods for automatic detection and segmentation of liver lesions were developed to support this task.

METHODS

The detection algorithm uses an object-based image analysis approach, allowing for effectively integrating domain knowledge and reasoning processes into the detection logic. The method is intended to succeed in cases typically difficult for computer-aided detection systems, especially low contrast of hypodense lesions relative to healthy tissue. The detection stage is followed by a dedicated segmentation algorithm needed to synthesize 3D segmentations for all true-positive findings.

RESULTS

The automated method provides an overall detection rate of 77.8% with a precision of 0.53 and performs better than other related methods. The final lesion segmentation delivers appropriate quality in 89% of the detected cases, as evaluated by two radiologists.

CONCLUSIONS

A new automated liver lesion detection algorithm employs the strengths of an object-based image analysis approach. The combination of automated detection and segmentation provides promising results with potential to improve diagnostic liver lesion evaluation.

Semi-automated volumetric analysis of artificial lymph nodes in a phantom study

PURPOSE

Quantification of tumour burden in oncology requires accurate and reproducible image evaluation. The current standard is one-dimensional measurement (e.g. RECIST) with inherent disadvantages. Volumetric analysis is discussed as an alternative for therapy monitoring of lung and liver metastases. The aim of this study was to investigate the accuracy of semi-automated volumetric analysis of artificial lymph node metastases in a phantom study.

MATERIALS AND METHODS

Fifty artificial lymph nodes were produced in a size range from 10 to 55mm; some of them enhanced using iodine contrast media. All nodules were placed in an artificial chest phantom (artiCHEST®) within different surrounding tissues. MDCT was performed using different collimations (1-5 mm) at varying reconstruction kernels (B20f, B40f, B60f). Volume and RECIST measurements were performed using Oncology Software (Siemens Healthcare, Forchheim, Germany) and were compared to reference volume and diameter by calculating absolute percentage errors.

RESULTS

The software performance allowed a robust volumetric analysis in a phantom setting. Unsatisfying segmentation results were frequently found for native nodules within surrounding muscle. The absolute percentage error (APE) for volumetric analysis varied between 0.01 and 225%. No significant differences were seen between different reconstruction kernels. The most unsatisfactory segmentation results occurred in higher slice thickness (4 and 5 mm). Contrast enhanced lymph nodes showed better segmentation results by trend.

CONCLUSION

The semi-automated 3D-volumetric analysis software tool allows a reliable and convenient segmentation of artificial lymph nodes in a phantom setting. Lymph nodes adjacent to tissue of similar density cause segmentation problems. For volumetric analysis of lymph node metastases in clinical routine a slice thickness of ≤3mm and a medium soft reconstruction kernel (e.g. B40f for Siemens scan systems) may be a suitable compromise for semi-automated volumetric analysis.

Does multidetector CT attenuation change in colon cancer liver metastases treated with 90Y help predict metabolic activity at FDG PET?

PURPOSE

To evaluate the correlation between change in attenuation and tumor metabolic activity assessed by using fluorodeoxyglucose (FDG) positron emission tomography (PET) in colon cancer liver metastases treated with yttrium 90 ((90)Y) radioembolization.

MATERIALS AND METHODS

This Health Insurance Portability and Accountability Act-compliant retrospective study was approved by the institutional review board; patient informed consent was waived. Unresectable chemorefractory colon cancer liver metastases treated with (90)Y radioembolization in 28 patients were evaluated at pre- and posttreatment multidetector computed tomographic (CT) and FDG PET scans. Maximum cross-sectional diameter, volume, and overall attenuation of target lesions were calculated. The percentage change (%Delta) in these parameters after treatment was calculated and correlated with the standardized uptake value (SUV) analysis at FDG PET. The accuracy of the radiologic parameters in helping predict response to treatment at FDG PET was assessed. Data were analyzed by using the Student t, Wilcoxon matched pair, Mann-Whitney, Spearman rank correlation, and chi(2) tests. The significance level was set at .05.

RESULTS

Seventy-four metastatic lesions in 10 women and 18 men (mean age, 61.5 years +/- 14.3 [standard deviation]) were evaluated. Mean follow-up interval for multidetector CT after treatment was 30 days. A significant reduction in maximum cross-sectional diameter, volume, and attenuation was observed from pre- to posttreatment multidetector CT (P < .05). The %Delta in attenuation had higher correlation with %Delta in SUV (r = 0.61) than diameter (r = 0.39) or volume (r = 0.49) and also predicted the metabolic activity at FDG PET with higher sensitivity (P < .001). By using a threshold level of a reduction in attenuation of 15% or greater, attenuation showed 84.2% sensitivity and 83.3% specificity in predicting response at FDG PET evaluation.

CONCLUSION

Changes in attenuation of colon cancer liver metastases treated with (90)Y radioembolization correlate highly with metabolic activity at FDG PET and may be useful as an early surrogate marker for assessing treatment response.

Radiofrequency ablation of liver metastases-software-assisted evaluation of the ablation zone in MDCT: tumor-free follow-up versus local recurrent disease

The purpose of this study was to investigate differences in change of size and CT value between local recurrences and tumor-free areas after CT-guided radiofrequency ablation (RFA) of hepatic metastases during follow-up by means of dedicated software for automatic evaluation of hepatic lesions. Thirty-two patients with 54 liver metastases from breast or colorectal cancer underwent triphasic contrast-enhanced multidetector-row computed tomography (MDCT) to evaluate hepatic metastatic spread and localization before CT-guided RFA and for follow-up after intervention. Sixteen of these patients (65.1 + or - 10.3 years) with 30 metastases stayed tumor-free (group 1), while the other group (n = 16 with 24 metastases; 62.0 + or - 13.8 years) suffered from local recurrent disease (group 2). Applying an automated software tool (SyngoCT Oncology; Siemens Healthcare, Forchheim, Germany), size parameters (volume, RECIST, WHO) and attenuation were measured within the lesions before, 1 day after, and 28 days after RFA treatment. The natural logarithm (ln) of the quotient of the volume 1 day versus 28 days after RFA treament was computed: lnQ1//28/0(volume). Analogously, ln ratios of RECIST, WHO, and attenuation were computed and statistically evaluated by repeated-measures ANOVA. One lesion in group 2 was excluded from further evaluation due to automated missegmentation. Statistically significant differences between the two groups were observed with respect to initial volume, RECIST, and WHO (p < 0.05). Furthermore, ln ratios corresponding to volume, RECIST, and WHO differed significantly between the two groups. Attenuation evaluations showed no significant differences, but there was a trend toward attenuation assessment for the parameter lnQ28/0(attenuation) (p = 0.0527), showing higher values for group 1 (-0.4 + or - 0.3) compared to group 2 (-0.2 + or - 0.2). In conclusion, hepatic metastases and their zone of coagulation necrosis after RFA differed significantly between tumor-free and local-recurrent ablation zones with respect to the corresponding size parameters. A new parameter (lnQ1//28/0(volume/RECIST/WHO/attenuation)) was introduced, which appears to be of prognostic value at early follow-up CT.

Semi-automated quantification of hepatic lesions in a phantom

PURPOSE

Accurate measurement is crucial for the assessment of tumor dimensions to allow accurate evaluation of tumor response. Thus, the purpose of our study was to assess the accuracy of semi-automated RECIST and volumetric measurements of liver lesions in a liver phantom with different CT acquisition parameters.

MATERIALS AND METHODS

A phantom of the upper abdomen with 14 hepatic lesions of different sizes (diameter: 12.0-40.0 mm), densities (45/180 HU at 120 kV), or alignment (vertical/transverse) was scanned with a 16-slice multidetector row computed tomography using varying tube currents (40/60/80/100/120/165mAs eff), reconstruction kernels (Siemens B20/30/40/50/70s), or slice thicknesses (1/2/3/4/5 mm). Longest axial diameter and volume of the 14 lesions were quantified using a semi-automated software tool (SyngoOncology, Siemens Medical Solutions, Forchheim, Germany) and compared with the known real longest axial diameter and volume values of the lesions. Absolute percentage errors (APE) were calculated. Degree of agreement in longest axial diameter and volume between software and real measurements was represented graphically in Bland-Altman plots and by corresponding concordance correlation coefficient.

RESULTS

At standard soft tissue reconstruction kernel (Siemens B30s) and slice thickness (3 mm) mean absolute percentage error APE (concordance correlation coefficients) ranged between 6.93 and 14.27 (0.96 and 0.99) for longest axial diameter and between 4.98 and 10.85 (0.99 and 1.00) for volume. At varying reconstruction kernels, APE values (concordance correlation coefficients) ranged between 7.92 and 8.31 (0.98 and 0.99) for longest axial diameter and between 4.95 and 6.93 (1.00) for volume. Applying different slice sections APE values (concordance correlation coefficients) differed from 6.54 to 11.82 (0.97 and 0.99) for longest axial diameter and from 6.93 to 9.17 (1.00) for volume.

CONCLUSIONS

Software quantification of longest axial diameter and volume of hepatic lesions in a phantom demonstrated a high correlation and accuracy under varying multidetector row computed tomography parameter.

Semi-automated volumetric analysis of lymph node metastases in patients with malignant melanoma stage III/IV--a feasibility study

Therapy monitoring in oncological patient care requires accurate and reliable imaging and post-processing methods. RECIST criteria are the current standard, with inherent disadvantages. The aim of this study was to investigate the feasibility of semi-automated volumetric analysis of lymph node metastases in patients with malignant melanoma compared to manual volumetric analysis and RECIST. Multislice CT was performed in 47 patients, covering the chest, abdomen and pelvis. In total, 227 suspicious, enlarged lymph nodes were evaluated retrospectively by two radiologists regarding diameters (RECIST), manually measured volume by placement of ROIs and semi-automated volumetric analysis. Volume (ml), quality of segmentation (++/--) and time effort (s) were evaluated in the study. The semi-automated volumetric analysis software tool was rated acceptable to excellent in 81% of all cases (reader 1) and 79% (reader 2). Median time for the entire segmentation process and necessary corrections was shorter with the semi-automated software than by manual segmentation. Bland-Altman plots showed a significantly lower interobserver variability for semi-automated volumetric than for RECIST measurements. The study demonstrated feasibility of volumetric analysis of lymph node metastases. The software allows a fast and robust segmentation in up to 80% of all cases. Ease of use and time needed are acceptable for application in the clinical routine. Variability and interuser bias were reduced to about one third of the values found for RECIST measurements.