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.

The role of tumour biological factors in technical anatomical resectability assessment of colorectal liver metastases following induction systemic treatment: An analysis of the Dutch CAIRO5 trial.

BACKGROUND: Large inter-surgeon variability exists in technical anatomical resectability assessment of colorectal cancer liver-only metastases (CRLM) following induction systemic therapy. We evaluated the role of tumour biological factors in predicting resectability and (early) recurrence after surgery for initially unresectable CRLM. METHODS: 482 patients with initially unresectable CRLM from the phase 3 CAIRO5 trial were selected, with two-monthly resectability assessments by a liver expert panel. If no consensus existed among panel surgeons (i.e. same vote for (un)resectability of CRLM), conclusion was based on majority. The association of tumour biological (sidedness, synchronous CRLM, carcinoembryonic antigen and RAS/BRAFV600E mutation status) and technical anatomical factors with consensus among panel surgeons, secondary resectability and early recurrence (<6 months) without curative-intent repeat local treatment was analysed by uni- and pre-specified multivariable logistic regression. RESULTS: After systemic treatment, 240 (50%) patients received complete local treatment of CRLM of which 75 (31%) patients experienced early recurrence without repeat local treatment. Higher number of CRLM (odds ratio 1.09 [95% confidence interval 1.03-1.15]) and age (odds ratio 1.03 [95% confidence interval 1.00-1.07]) were independently associated with early recurrence without repeat local treatment. In 138 (52%) patients, no consensus among panel surgeons was present prior to local treatment. Postoperative outcomes in patients with and without consensus were comparable. CONCLUSIONS: Almost a third of patients selected by an expert panel for secondary CRLM surgery following induction systemic treatment experience an early recurrence only amenable to palliative treatment. Number of CRLM and age, but no tumour biological factors are predictive, suggesting that until there are better biomarkers; resectability assessment remains primarily a technical anatomical decision.

A nationwide assessment of hepatocellular adenoma resection: Indications and pathological discordance.

Hepatocellular adenomas (HCAs) are benign liver tumors associated with bleeding or malignant transformation. Data on the indication for surgery are scarce. We analyzed indications and outcome of patients operated for HCAs < 50 mm compared to HCAs ≥ 50 mm. Changes in final postoperative diagnosis were assessed. We performed a retrospective study that included patients who underwent resection for (suspected) HCAs in the Netherlands from 2014 to 2019. Indication for resection was analyzed and stratified for small (<50 mm) and large (≥50 mm) tumors. Logistic regression analysis was performed on factors influencing change in tumor diagnosis. Out of 222 patients who underwent surgery, 44 (20%) patients had a tumor <50 mm. Median age was 46 (interquartile range [IQR], 33-56) years in patients with small tumors and 37 (IQR, 31-46) years in patients with large tumors ( p  = 0.016). Patients with small tumors were more frequently men (21% vs. 5%, p  = 0.002). Main indications for resection in patients with small tumors were suspicion of (pre)malignancy (55%), (previous) bleeding (14%), and male sex (11%). Patients with large tumors received operations because of tumor size >50 mm (52%), suspicion of (pre)malignancy (28%), and (previous) bleeding (5.1%). No difference was observed in HCA-subtype distribution between small and large tumors. Ninety-six (43%) patients had a postoperative change in diagnosis. Independent risk factors for change in diagnosis were tumor size <50 mm (adjusted odds ratio [aOR], 3.4; p  < 0.01), male sex (aOR, 3.7; p  = 0.03), and lack of hepatobiliary contrast-enhanced magnetic resonance imaging (CE-MRI) (aOR, 1.8; p  = 0.04). Resection for small (suspected) HCAs was mainly indicated by suspicion of (pre)malignancy, whereas for large (suspected) HCAs, tumor size was the most prevalent indication. Male sex, tumor size <50 mm, and lack of hepatobiliary CE-MRI were independent risk factors for postoperative change in tumor diagnosis.

Impact of cumulative complications on 1-year treatment-related healthcare costs in patients with colorectal peritoneal metastases undergoing cytoreductive surgery with hyperthermic intraperitoneal chemotherapy.

BACKGROUND: The aim of this study was to evaluate the impact of all minor and major complications on treatment-related healthcare costs in patients who undergo cytoreductive surgery (CRS) with hyperthermic intraperitoneal chemotherapy (HIPEC) for the treatment of colorectal peritoneal metastases (PMs). METHOD: Patients with histologically proven colorectal PMs who underwent CRS + HIPEC from March 2006 to October 2019 in a tertiary referral centre were retrospectively identified from a prospectively maintained database. Patients were divided into six subgroups according to the severity of the complications, which were scored using the comprehensive complication index (CCI) (CCI 0-9.9, CCI 10-19.9, CCI 20-29.9, CCI 30-39.9, CCI 40-49.9, and CCI 50 or higher). Treatment-related healthcare costs up to 1 year after CRS + HIPEC were obtained from the financial department. Differences in costs and survival outcomes were compared using the chi-squared test and Kruskal-Wallis H test. RESULTS: A total of 142 patients were included (CCI 0-9.9, 53 patients; CCI 10-19.9, 0 patients; CCI 20-29.9, 45 patients; CCI 30-39.9, 14 patients; CCI 40-49, 9 patients; and CCI 50 or higher, 21 patients). Median (interquartile range) treatment-related healthcare costs increased significantly and exponentially for the CCI 30-39, CCI 40-49, and CCI 50 or higher groups (€48 993 (€44 262-€84 805); €57 167 (€43 047-€67 591); and €82 219 (€55 487-€145 314) respectively) compared with those for the CCI 0-9.9 and CCI 20-29.9 groups (€33 856 (€24 433-€40 779) and €40 621 (€31 501-€58 761) respectively, P < 0.010). CONCLUSION: Treatment-related healthcare costs increase exponentially as more complications develop among patients who undergo CRS + HIPEC for the treatment of colorectal PMs. Anastomotic leakages after CRS + HIPEC lead to an increase of 295 per cent of treatment-related healthcare costs.

Defining Textbook Outcome in liver surgery and assessment of hospital variation: A nationwide population-based study.

INTRODUCTION: Textbook outcome (TO) is a composite outcome measure covering the surgical care process in a single outcome measure. TO has an advantage over single outcome parameters with low event rates, which have less discriminating impact to detect differences between hospitals. This study aimed to assess factors associated with TO, and evaluate hospital and network variation after case-mix correction in TO rates for liver surgery. METHODS: This was a population-based retrospective study of all patients who underwent liver resection for malignancy in the Netherlands in 2019 and 2020. TO was defined as absence of severe postoperative complications, mortality, prolonged length of hospital stay, and readmission, and obtaining adequate resection margins. Multivariable logistic regression was used for case-mix adjustment. RESULTS: 2376 patients were included. TO was accomplished in 1380 (80%) patients with colorectal liver metastases, in 192 (76%) patients with other liver metastases, in 183 (74%) patients with hepatocellular carcinoma and 86 (51%) patients with biliary cancers. Factors associated with lower TO rates for CRLM included ASA score ≥3 (aOR 0.70, CI 0.51-0.95 p = 0.02), extrahepatic disease (aOR 0.64, CI 0.44-0.95, p = 0.02), tumour size >55 mm on preoperative imaging (aOR 0.56, CI 0.34-0.94, p = 0.02), Charlson Comorbidity Index ≥2 (aOR 0.73, CI 0.54-0.98, p = 0.04), and major liver resection (aOR 0.50, CI 0.36-0.69, p < 0.001). After case-mix correction, no significant hospital or oncological network variation was observed. CONCLUSION: TO differs between indications for liver resection and can be used to assess between hospital and network differences.

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.

Nationwide oncological networks for resection of colorectal liver metastases in the Netherlands: Differences and postoperative outcomes.

INTRODUCTION: Widespread differences in patient demographics and disease burden between hospitals for resection of colorectal liver metastases (CRLM) have been described. In the Netherlands, networks consisting of at least one tertiary referral centre and several regional hospitals have been established to optimize treatment and outcomes. The aim of this study was to assess variation in case-mix, and outcomes between these networks. METHODS: This was a population-based study including all patients who underwent CRLM resection in the Netherlands between 2014 and 2019. Variation in case-mix and outcomes between seven networks covering the whole country was evaluated. Differences in case-mix, expected 30-day major morbidity (Clavien-Dindo ≥3a) and 30-day mortality between networks were assessed. RESULTS: In total 5383 patients were included. Thirty-day major morbidity was 5.7% and 30-day mortality was 1.5%. Significant differences between networks were observed for Charlson Comorbidity Index, ASA 3+, previous liver resection, liver disease, preoperative MRI, preoperative chemotherapy, ≥3 CRLM, diameter of largest CRLM ≥55 mm, major resection, combined resection and ablation, rectal primary tumour, bilobar and extrahepatic disease. Uncorrected 30-day major morbidity ranged between 3.3% and 13.1% for hospitals, 30-day mortality ranged between 0.0% and 4.5%. Uncorrected 30-day major morbidity ranged between 4.4% and 6.0% for networks, 30-day mortality ranged between 0.0% and 2.5%. No negative outliers were observed after case-mix correction. CONCLUSION: Variation in case-mix and outcomes are considerably smaller on a network level as compared to a hospital level. Therefore, auditing is more meaningful at a network level and collaboration of hospitals within networks should be pursued.

The Costs of Complications and Unplanned Readmissions after Pancreatoduodenectomy for Pancreatic and Periampullary Tumors: Results from a Single Academic Center.

BACKGROUND/OBJECTIVES: Complications after pancreatoduodenectomy (PD) lead to unplanned readmissions (UR), with a two- to threefold increase in admission costs. In this study, we aimed to create an understanding of the costs of complications and UR in this patient group. Furthermore, we aimed to generate a detailed cost overview that can be used to build a theoretical model to calculate the cost efficacy for prehabilitation. METHODS: A retrospective cohort analysis was performed using the Dutch Pancreatic Cancer Audit (DPCA) database of patients who underwent a PD at our institute between 2013 and 2017. The total costs of the index hospital admission and UR related to the PD were collected. RESULTS: Of the 160 patients; 35 patients (22%) had an uncomplicated course; 87 patients (54%) had minor complications, and 38 patients (24%) had severe complications. Median costs for an uncomplicated course were EUR 25.682, and for a complicated course, EUR 32.958 (p = 0.001). The median costs for minor complications were EUR 30.316, and for major complications, EUR 42.664 (p = 0.001). Costs were related to the Comprehensive Complication Index (CCI). The median costs of patients with one or more UR were EUR 41.199. CONCLUSIONS: Complications after PD led to a EUR 4.634-EUR 16.982 (18-66%) increase in hospital costs. A UR led to a cost increase of EUR 12.567 (44%). Since hospital costs are directly related to the CCI, reduction in complications will lead to cost-effectiveness.

A prospective cohort study evaluating screening and assessment of six modifiable risk factors in HPB cancer patients and compliance to recommended prehabilitation interventions.

INTRODUCTION: Despite improvements in perioperative care, major abdominal surgery continues to be associated with significant perioperative morbidity. Accurate preoperative risk stratification and optimisation (prehabilitation) are necessary to reduce perioperative morbidity. This study evaluated the screening and assessment of modifiable risk factors amendable for prehabilitation interventions and measured the patient compliance rate with recommended interventions. METHOD: Between May 2019 and January 2020, patients referred to our hospital for HPB surgery were screened and assessed on six modifiable preoperative risk factors. The risk factors and screening tools used, with cutoff values, included (i) low physical fitness (a 6-min walk test < 82% of patient's calculated norm and/or patient's activity level not meeting the global recommendations on physical activity for health). Patients who were unfit based on the screening were assessed with a cardiopulmonary exercise test (anaerobic threshold ≤ 11 mL/kg/min); (ii) malnutrition (patient-generated subjective global assessment ≥ 4); (iii) iron-deficiency anaemia (haemoglobin < 12 g/dL for women, < 13 g/dL for men and transferrin saturation ≤ 20%); (iv) frailty (Groningen frailty indicator/Robinson frailty score ≥ 4); (v) substance use (smoking and alcohol use of > 5 units per week) and (vi) low psychological resilience (Hospital Anxiety and Depression Scale ≥ 8). Patients had a consultation with the surgeon on the same day as their screening. High-risk patients were referred for necessary interventions. RESULTS: One hundred consecutive patients were screened at our prehabilitation outpatient clinic. The prevalence of high-risk patients per risk factor was 64% for low physical fitness, 42% for malnutrition, 32% for anaemia (in 47% due to iron deficiency), 22% for frailty, 12% for smoking, 18% for alcohol use and 21% for low psychological resilience. Of the 77 patients who were eventually scheduled for surgery, 53 (68.8%) needed at least one intervention, of whom 28 (52.8%) complied with 100% of the necessary interventions. The median (IQR) number of interventions needed in the 77 patients was 1.0 (0-2). CONCLUSION: It is feasible to screen and assess all patients referred for HPB cancer surgery for six modifiable risk factors. Most of the patients had at least one risk factor that could be optimised. However, compliance with the suggested interventions remains challenging.

Appearance of breast cysts in planar geometry photoacoustic mammography using 1064-nm excitation.

In the search for improved imaging modalities for detection and diagnosis of breast cancer, a high negative prediction value is also important. Photoacoustic (optoacoustic) imaging is a relatively new technique that has high potential for visualizing breast malignancies, but little is known about the photoacoustic appearance of benign lesions. In this work, we investigate the visibility of benign breast cysts in forward-mode photoacoustic mammography using 1064-nm light, as currently applied in the Twente photoacoustic mammoscope. Results from (Monte Carlo and k-wave) simulations and phantom measurements were used to interpret results from patient measurements. There was a strong agreement among the results from simulations, phantom, and patient measurements. Depending on the absorption contrast between cyst and breast tissue, cysts were visible as either one or two confined high-contrast areas representing the front and the back of the cyst, respectively. This edge enhancement is most likely the consequence of the local sudden change in the absorbed energy density and Grüneisen coefficients. Although the current forward-mode single-wavelength photoacoustic mammoscope cannot always unambiguously discriminate cysts from malignancies, this study reveals specific features of cysts compared to malignancies, which can be exploited for discrimination of the two abnormalities in future modifications of the imager.