Deep learning models for automatic tumor segmentation and total tumor volume assessment in patients with colorectal liver metastases.

BACKGROUND: We developed models for tumor segmentation to automate the assessment of total tumor volume (TTV) in patients with colorectal liver metastases (CRLM). METHODS: In this prospective cohort study, pre- and post-systemic treatment computed tomography (CT) scans of 259 patients with initially unresectable CRLM of the CAIRO5 trial (NCT02162563) were included. In total, 595 CT scans comprising 8,959 CRLM were divided into training (73%), validation (6.5%), and test sets (21%). Deep learning models were trained with ground truth segmentations of the liver and CRLM. TTV was calculated based on the CRLM segmentations. An external validation cohort was included, comprising 72 preoperative CT scans of patients with 112 resectable CRLM. Image segmentation evaluation metrics and intraclass correlation coefficient (ICC) were calculated. RESULTS: In the test set (122 CT scans), the autosegmentation models showed a global Dice similarity coefficient (DSC) of 0.96 (liver) and 0.86 (CRLM). The corresponding median per-case DSC was 0.96 (interquartile range [IQR] 0.95-0.96) and 0.80 (IQR 0.67-0.87). For tumor segmentation, the intersection-over-union, precision, and recall were 0.75, 0.89, and 0.84, respectively. An excellent agreement was observed between the reference and automatically computed TTV for the test set (ICC 0.98) and external validation cohort (ICC 0.98). In the external validation, the global DSC was 0.82 and the median per-case DSC was 0.60 (IQR 0.29-0.76) for tumor segmentation. CONCLUSIONS: Deep learning autosegmentation models were able to segment the liver and CRLM automatically and accurately in patients with initially unresectable CRLM, enabling automatic TTV assessment in such patients. RELEVANCE STATEMENT: Automatic segmentation enables the assessment of total tumor volume in patients with colorectal liver metastases, with a high potential of decreasing radiologist's workload and increasing accuracy and consistency. KEY POINTS: • Tumor response evaluation is time-consuming, manually performed, and ignores total tumor volume. • Automatic models can accurately segment tumors in patients with colorectal liver metastases. • Total tumor volume can be accurately calculated based on automatic segmentations.

First-line systemic treatment strategies in patients with initially unresectable colorectal cancer liver metastases (CAIRO5): an open-label, multicentre, randomised, controlled, phase 3 study from the Dutch Colorectal Cancer Group.

BACKGROUND: Patients with initially unresectable colorectal cancer liver metastases might qualify for local treatment with curative intent after reducing the tumour size by induction systemic treatment. We aimed to compare the currently most active induction regimens. METHODS: In this open-label, multicentre, randomised, phase 3 study (CAIRO5), patients aged 18 years or older with histologically confirmed colorectal cancer, known RAS/BRAFV600E mutation status, WHO performance status of 0-1, and initially unresectable colorectal cancer liver metastases were enrolled at 46 Dutch and one Belgian secondary and tertiary centres. Resectability or unresectability of colorectal cancer liver metastases was assessed centrally by an expert panel of liver surgeons and radiologists, at baseline and every 2 months thereafter by predefined criteria. Randomisation was done centrally with the minimisation technique via a masked web-based allocation procedure. Patients with right-sided primary tumour site or RAS or BRAFV600E mutated tumours were randomly assigned (1:1) to receive FOLFOX or FOLFIRI plus bevacizumab (group A) or FOLFOXIRI plus bevacizumab (group B). Patients with left-sided and RAS and BRAFV600E wild-type tumours were randomly assigned (1:1) to receive FOLFOX or FOLFIRI plus bevacizumab (group C) or FOLFOX or FOLFIRI plus panitumumab (group D), every 14 days for up to 12 cycles. Patients were stratified by resectability of colorectal cancer liver metastases, serum lactate dehydrogenase concentration, choice of irinotecan versus oxaliplatin, and BRAFV600E mutation status (for groups A and B). Bevacizumab was administered intravenously at 5 mg/kg. Panitumumab was administered intravenously at 6 mg/kg. FOLFIRI consisted of intravenous infusion of irinotecan at 180 mg/m2 with folinic acid at 400 mg/m2, followed by bolus fluorouracil at 400 mg/m2 intravenously, followed by continuous infusion of fluorouracil at 2400 mg/m2. FOLFOX consisted of oxaliplatin at 85 mg/m2 intravenously together with the same schedule of folinic acid and fluorouracil as in FOLFIRI. FOLFOXIRI consisted of irinotecan at 165 mg/m2 intravenously, followed by intravenous infusion of oxaliplatin at 85 mg/m2 with folinic acid at 400 mg/m2, followed by continuous infusion of fluorouracil at 3200 mg/m2. Patients and investigators were not masked to treatment allocation. The primary outcome was progression-free survival, analysed on a modified intention-to-treat basis, excluding patients who withdrew consent before starting study treatment or violated major entry criteria (no metastatic colorectal cancer, or previous liver surgery for colorectal cancer liver metastases). The study is registered with ClinicalTrials.gov, NCT02162563, and accrual is complete. FINDINGS: Between Nov 13, 2014, and Jan 31, 2022, 530 patients (327 [62%] male and 203 [38%] female; median age 62 years [IQR 54-69]) were randomly assigned: 148 (28%) patients to group A, 146 (28%) patients to group B, 118 (22%) patients to group C, and 118 (22%) patients to group D. Groups C and D were prematurely closed for futility. 521 patients were included in the modified intention-to-treat population (147 in group A, 144 in group B, 114 in group C, and 116 in group D). The median follow-up at the time of this analysis was 51·1 months (95% CI 47·7-53·1) in groups A and B and 49·9 months (44·5-52·5) in in groups C and D. Median progression-free survival was 9·0 months (95% CI 7·7-10·5) in group A versus 10·6 months (9·9-12·1) in group B (stratified hazard ratio [HR] 0·76 [95% CI 0·60-0·98]; p=0·032), and 10·8 months (95% CI 9·9-12·6) in group C versus 10·4 months (9·8-13·0) in group D (stratified HR 1·11 [95% CI 0·84-1·48]; p=0·46). The most frequent grade 3-4 events in groups A and B were neutropenia (19 [13%] patients in group A vs 57 [40%] in group B; p<0·0001), hypertension (21 [14%] vs 20 [14%]; p=1·00), and diarrhoea (five [3%] vs 28 [19%]; p<0·0001), and in groups C and D were neutropenia (29 [25%] vs 24 [21%]; p=0·44), skin toxicity (one [1%] vs 29 [25%]; p<0·0001), hypertension (20 [18%] vs eight [7%]; p=0·016), and diarrhoea (five [4%] vs 18 [16%]; p=0·0072). Serious adverse events occurred in 46 (31%) patients in group A, 75 (52%) patients in group B, 41 (36%) patients in group C, and 49 (42%) patients in group D. Seven treatment-related deaths were reported in group B (two due to multiorgan failure, and one each due to sepsis, pneumonia, portal vein thrombosis, septic shock and liver failure, and sudden death), one in group C (multiorgan failure), and three in group D (cardiac arrest, pulmonary embolism, and abdominal sepsis). INTERPRETATION: In patients with initially unresectable colorectal cancer liver metastases, FOLFOXIRI-bevacizumab was the preferred treatment in patients with a right-sided or RAS or BRAFV600E mutated primary tumour. In patients with a left-sided and RAS and BRAFV600E wild-type tumour, the addition of panitumumab to FOLFOX or FOLFIRI showed no clinical benefit over bevacizumab, but was associated with more toxicity. FUNDING: Roche and Amgen.

Intersurgeon Variability in Local Treatment Planning for Patients with Initially Unresectable Colorectal Cancer Liver Metastases: Analysis of the Liver Expert Panel of the Dutch Colorectal Cancer Group.

BACKGROUND: Consensus on resectability criteria for colorectal cancer liver metastases (CRLM) is lacking, resulting in differences in therapeutic strategies. This study evaluated variability of resectability assessments and local treatment plans for patients with initially unresectable CRLM by the liver expert panel from the randomised phase III CAIRO5 study. METHODS: The liver panel, comprising surgeons and radiologists, evaluated resectability by predefined criteria at baseline and 2-monthly thereafter. If surgeons judged CRLM as resectable, detailed local treatment plans were provided. The panel chair determined the conclusion of resectability status and local treatment advice, and forwarded it to local surgeons. RESULTS: A total of 1149 panel evaluations of 496 patients were included. Intersurgeon disagreement was observed in 50% of evaluations and was lower at baseline than follow-up (36% vs. 60%, p < 0.001). Among surgeons in general, votes for resectable CRLM at baseline and follow-up ranged between 0-12% and 27-62%, and for permanently unresectable CRLM between 3-40% and 6-47%, respectively. Surgeons proposed different local treatment plans in 77% of patients. The most pronounced intersurgeon differences concerned the advice to proceed with hemihepatectomy versus parenchymal-preserving approaches. Eighty-four percent of patients judged by the panel as having resectable CRLM indeed received local treatment. Local surgeons followed the technical plan proposed by the panel in 40% of patients. CONCLUSION: Considerable variability exists among expert liver surgeons in assessing resectability and local treatment planning of initially unresectable CRLM. This stresses the value of panel-based decisions, and the need for consensus guidelines on resectability criteria and technical approach to prevent unwarranted variability in clinical practice.

Interobserver Variability in CT-based Morphologic Tumor Response Assessment of Colorectal Liver Metastases.

Purpose To evaluate interobserver variability in the morphologic tumor response assessment of colorectal liver metastases (CRLM) managed with systemic therapy and to assess the relation of morphologic response with gene mutation status, targeted therapy, and Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 measurements. Materials and Methods Participants with initially unresectable CRLM receiving different systemic therapy regimens from the randomized, controlled CAIRO5 trial (NCT02162563) were included in this prospective imaging study. Three radiologists independently assessed morphologic tumor response on baseline and first follow-up CT scans according to previously published criteria. Two additional radiologists evaluated disagreement cases. Interobserver agreement was calculated by using Fleiss κ. On the basis of the majority of individual radiologic assessments, the final morphologic tumor response was determined. Finally, the relation of morphologic tumor response and clinical prognostic parameters was assessed. Results In total, 153 participants (median age, 63 years [IQR, 56-71]; 101 men) with 306 CT scans comprising 2192 CRLM were included. Morphologic assessment performed by the three radiologists yielded 86 (56%) agreement cases and 67 (44%) disagreement cases (including four major disagreement cases). Overall interobserver agreement between the panel radiologists on morphology groups and morphologic response categories was moderate (κ = 0.53, 95% CI: 0.48, 0.58 and κ = 0.54, 95% CI: 0.47, 0.60). Optimal morphologic response was particularly observed in patients treated with bevacizumab (P = .001) and in patients with RAS/BRAF mutation (P = .04). No evidence of a relationship between RECIST 1.1 and morphologic response was found (P = .61). Conclusion Morphologic tumor response assessment following systemic therapy in participants with CRLM demonstrated considerable interobserver variability. Keywords: Tumor Response, Observer Performance, CT, Liver, Metastases, Oncology, Abdomen/Gastrointestinal Clinical trial registration no. NCT02162563 Supplemental material is available for this article. © RSNA, 2022.

The Disease-Free Interval Between Resection of Primary Colorectal Malignancy and the Detection of Hepatic Metastases Predicts Disease Recurrence But Not Overall Survival.

INTRODUCTION: The disease-free interval (DFI) between resection of primary colorectal cancer (CRC) and diagnosis of liver metastases is considered an important prognostic indicator; however, recent analyses in metastatic CRC found limited evidence to support this notion. OBJECTIVE: The current study aims to determine the prognostic value of the DFI in patients with resectable colorectal liver metastases (CRLM). METHODS: Patients undergoing first surgical treatment of CRLM at three academic centers in The Netherlands were eligible for inclusion. The DFI was defined as the time between resection of CRC and detection of CRLM. Baseline characteristics and Kaplan-Meier survival estimates were stratified by DFI. Cox regression analyses were performed for overall (OS) and disease-free survival (DFS), with the DFI entered as a continuous measure using a restricted cubic spline function with three knots. RESULTS: In total, 1374 patients were included. Patients with a shorter DFI more often had lymph node involvement of the primary, more frequently received neoadjuvant chemotherapy for CRLM, and had higher number of CRLM at diagnosis. The DFI significantly contributed to DFS prediction (p =0.002), but not for predicting OS (p =0.169). Point estimates of the hazard ratio (95% confidence interval) for a DFI of 0 versus 12 months and 0 versus 24 months were 1.284 (1.114-1.480) and 1.444 (1.180-1.766), respectively, for DFS, and 1.111 (0.928-1.330) and 1.202 (0.933-1.550), respectively, for OS. CONCLUSION: The DFI is of prognostic value for predicting disease recurrence following surgical treatment of CRLM, but not for predicting OS outcomes.

Radiofrequency ablation compared to surgical resection for curative treatment of patients with colorectal liver metastases - a meta-analysis.

BACKGROUND: Hepatic resection and ablative treatments, such as RFA are available treatment options for liver tumors. Advantages and disadvantages of these treatment options in patients with colorectal liver metastases need further evaluation. The purpose of this study was to systematically evaluate the role of radiofrequency ablation (RFA) compared to surgery in the curative treatment of patients with colorectal liver metastases (CRLM). METHODS: A systematic search was performed from MEDLINE, EMBASE and the Cochrane Library for studies directly comparing RFA with resection for CRLM, after which variables were evaluated. RESULTS: RFA had significantly lower complication rates (OR = 0.44, 95% CI = 0.26-0.75, P = 0.002) compared to resection. However, RFA showed a higher rate of any recurrence (OR = 1.66, 95% CI = 1.15-2.40, P = 0.007), local recurrence (OR = 9.56, 95% CI = 6.85-13.35, P = 0.001), intrahepatic recurrence (OR = 1.96, 95% CI = 1.34-2.87, P = 0.001) and extrahepatic recurrence (OR = 1.21, 95% CI = 0.90-1.63, P = 0.22). Also, 5-year disease-free survival (OR = 2.20, 95% CI = 1.28-3.79, P = 0.005) and overall survival (OR = 2.35, 95% CI = 1.49-3.69, P = 0.001) were significantly lower in patients treated with RFA. CONCLUSIONS: RFA showed a significantly lower rate of complications, but also a lower survival and a higher rate of recurrence as compared to surgical resection. All the included studies were subject to possible patient selection bias and therefore randomized clinical trials are needed to accurately evaluate these treatment modalities.

"On the Spot": The Use of Four-Dimensional Computed Tomography Angiography to Differentiate a True Spot Sign From a Distal Intracranial Aneurysm.

BACKGROUND: Spontaneous intracerebral hemorrhage may arise from underlying abnormalities, including aneurysms. Computed tomography angiography (CTA) is widely used for the detection of possible underlying causes, which is important because it may have immediate therapeutic consequences. In addition, CTA is used to detect the so-called spot sign, indicating active hemorrhage, which carries a worse prognosis. However, CTA is a snapshot in time. Four-dimensional (4D) CTA is a dynamic type of imaging and has emerged as a valuable imaging technique for different neurovascular disorders. CASE DESCRIPTION: Two patients with intracerebral hemorrhage both showed an assumed spot sign on CTA, suggesting active hemorrhage. Additional 4D-CTA showed true active hemorrhage in one patient and a distal intracranial aneurysm in the other. This aneurysm was initially falsely interpreted as a spot sign on conventional CTA. CONCLUSIONS: Our case findings show how 4D-CTA can discern active bleeding from aneurysmal hemorrhage in cases with hemorrhagic stroke. This finding proves the additional value of this relatively new technique, because the detected underlying disorders have different therapeutic consequences in the acute setting.

Development of a high-field MR-guided HIFU setup for thermal and mechanical ablation methods in small animals.

BACKGROUND: Thermal and mechanical high intensity focused ultrasound (HIFU) ablation techniques are in development for non-invasive treatment of cancer. However, knowledge of in vivo histopathologic and immunologic reactions after HIFU ablation is still limited. This study aims to create a setup for evaluation of different HIFU ablation methods in mouse tumors using high-field magnetic resonance (MR) guidance. An optimized MR-guided-HIFU setup could be used to increase knowledge of the different pathologic and immunologic reactions to different HIFU ablation methods. METHODS: Three different HIFU treatment strategies were applied in mouse melanomas (B16): a thermal (continuous wave), a mechanical (5 ms pulsed wave), and an intermediate setting (20 ms pulsed wave) for HIFU ablation, all under MR guidance using a 7 tesla animal MR system. Histopathologic evaluation was performed 3 days after treatment. RESULTS: The focus of the ultrasound transducer could accurately be positioned within the tumor under MR image guidance, without substantial damage to the surrounding tissue and skin. All mice retained complete use of the treated leg after treatment. Temperatures of >60, <50, and <44 °C were reached during thermal, intermediate, and mechanical HIFU ablation, respectively. Thermal-treated tumors showed large regions of coagulative necrosis. Tumors of both the mechanical and intermediate groups showed fractionated tissue with islands of necrosis and some pseudocysts with hemorrhage. CONCLUSION: A stable small animal MR-guided HIFU setup was designed and evaluated for follow-up MR imaging and histopathologic responses of the treated tumors. This will facilitate further studies with a larger number of mice for detailed evaluation of the pathologic and immunologic response to different HIFU strategies.

Aneurysm diameter as a risk factor for pretreatment rebleeding: a meta-analysis.

OBJECT: Aneurysmal rerupture prior to treatment is a major cause of death and morbidity in aneurysmal subarachnoid hemorrhage. Recognizing risk factors for aneurysmal rebleeding is particularly relevant and might help to identify the aneurysms that benefit from acute treatment. It is uncertain if the size of the aneurysm is related to rebleeding. This meta-analysis was performed to evaluate whether an association could be determined between aneurysm diameter and the rebleeding rate before treatment. Potentially confounding factors such age, aneurysm location, and the presence of hypertension were also evaluated. METHODS: The authors systematically searched the PubMed, Embase, and Cochrane databases up to April 3, 2013, for studies of patients with aneurysmal subarachnoid hemorrhage that reported the association between aneurysm diameter and pretreatment aneurysmal rebleeding. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria were used to evaluate study quality. RESULTS: Seven studies, representing 2121 patients, were included in the quantitative analysis. The quality of the studies was low in 2 and very low in 5. Almost all of the studies used 10 mm as the cutoff point for size among other classes, and only one used 7 mm. An analysis was performed with this best unifiable cutoff point. Overall rebleeding occurred in 360 (17.0%) of 2121 patients (incidence range, from study to study, 8.7%-28.4%). The rate of rebleeding in small and large aneurysms was 14.0% and 23.6%, respectively. The meta-analysis of the 7 studies revealed that larger size aneurysms were at a higher risk for rebleeding (OR 2.56 [95% CI 1.62-4.06]; p = 0.00; I (2) = 60%). The sensitivity analysis did not alter the results. Five of the 7 studies reported data regarding age; 4 studies provided age-adjusted results and identified a persistent relationship between lesion size and the risk of rebleeding. The presence of hypertension was reported in two studies and was more prevalent in patients with rebleeding in one of these. Location (anterior vs posterior circulation) was reported in 5 studies, while in 4 there was no difference in the rebleeding rate. One study identified a lower risk of rebleeding associated with posterior location aneurysms. CONCLUSIONS: This meta-analysis showed that aneurysm size is an important risk factor for aneurysmal rebleeding and should be used in the clinical risk assessment of individual patients. The authors' results confirmed the current guidelines and underscored the importance of acute treatment for large ruptured aneurysms.

Caseload as a factor for outcome in aneurysmal subarachnoid hemorrhage: a systematic review and meta-analysis.

OBJECT: Increasing evidence exists that treatment of complex medical conditions in high-volume centers is found to improve outcome. Patients with subarachnoid hemorrhage (SAH), a complex disease, probably also benefit from treatment at a high-volume center. The authors aimed to determine, based on published literature, whether a higher hospital caseload is associated with improved outcomes of patients undergoing treatment after aneurysmal subarachnoid hemorrhage. METHODS: The authors identified studies from MEDLINE, Embase, and the Cochrane Library up to September 28, 2012, that evaluated outcome in high-volume versus low-volume centers in patients with SAH who were treated by either clipping or endovascular coiling. No language restrictions were set. The compared outcome measure was in-hospital mortality. Mortality in studies was pooled in a random effects meta-analysis. Study quality was reported according to the GRADE (Grading of Recommendations Assessment, Development and Evaluation) criteria. RESULTS: Four articles were included in this analysis, representing 36,600 patients. The quality of studies was graded low in 3 and very low in 1. Meta-analysis using a random effects model showed a decrease in hospital mortality (OR 0.77 [95% CI 0.60-0.97]; p = 0.00; I(2) = 91%) in high-volume hospitals treating SAH patients. Sensitivity analysis revealed the relative weight of the 1 low-quality study. Removal of the study with very low quality increased the effect size of the meta-analysis to an OR of 0.68 (95% CI 0.56-0.84; p = 0.00; I(2) = 86%). The definition of hospital volume differed among studies. Cutoffs and dichotomizations were used as well as division in quartiles. In 1 study, low volume was defined as 9 or fewer patients yearly, whereas in another it was defined as fewer than 30 patients yearly. Similarly, 1 study defined high volume as more than 20 patients annually, and another defined it as more than 50 patients a year. For comparability between studies, recalculation was done with dichotomized data if available. Cross et al., 2003 (low volume ≤ 18, high volume ≥ 19) and Johnston, 2000 (low volume ≤ 31, high volume ≥ 32) provided core data for recalculation. The overall results of this analysis revealed an OR of 0.85 (95% CI 0.72-0.99; p = 0.00; I(2) = 87%). CONCLUSIONS: Despite the shortcomings of this study, the mortality rate was lower in hospitals with a larger caseload. Limitations of the meta-analysis are the not uniform cutoff values and uncertainty about case mix.