Deep Learning-Based Segmentation of Locally Advanced Breast Cancer on MRI in Relation to Residual Cancer Burden: A Multi-Institutional Cohort Study

BACKGROUND

While several methods have been proposed for automated assessment of breast-cancer response to neoadjuvant chemotherapy on breast MRI, limited information is available about their performance across multiple institutions.

PURPOSE

To assess the value and robustness of deep learning-derived volumes of locally advanced breast cancer (LABC) on MRI to infer the presence of residual disease after neoadjuvant chemotherapy.

STUDY TYPE

Retrospective.

SUBJECTS

Training cohort: 102 consecutive female patients with LABC scheduled for neoadjuvant chemotherapy (NAC) from a single institution (age: 25-73 years). Independent testing cohort: 55 consecutive female patients with LABC from four institutions (age: 25-72 years).

FIELD STRENGTH/SEQUENCE

Training cohort: single vendor 1.5 T or 3.0 T. Testing cohort: multivendor 3.0 T. Gradient echo dynamic contrast-enhanced sequences.

ASSESSMENT

A convolutional neural network (nnU-Net) was trained to segment LABC. Based on resulting tumor volumes, an extremely randomized tree model was trained to assess residual cancer burden (RCB)-0/I vs. RCB-II/III. An independent model was developed using functional tumor volume (FTV). Models were tested on an independent testing cohort and response assessment performance and robustness across multiple institutions were assessed.

STATISTICAL TESTS

The receiver operating characteristic (ROC) was used to calculate the area under the ROC curve (AUC). DeLong's method was used to compare AUCs. Correlations were calculated using Pearson's method. P values <0.05 were considered significant.

RESULTS

Automated segmentation resulted in a median (interquartile range [IQR]) Dice score of 0.87 (0.62-0.93), with similar volumetric measurements (R = 0.95, P < 0.05). Automated volumetric measurements were significantly correlated with FTV (R = 0.80). Tumor volume-derived from deep learning of DCE-MRI was associated with RCB, yielding an AUC of 0.76 to discriminate between RCB-0/I and RCB-II/III, performing similar to the FTV-based model (AUC = 0.77, P = 0.66). Performance was comparable across institutions (IQR AUC: 0.71-0.84).

DATA CONCLUSION

Deep learning-based segmentation estimates changes in tumor load on DCE-MRI that are associated with RCB after NAC and is robust against variations between institutions.

EVIDENCE LEVEL

2.

TECHNICAL EFFICACY

Stage 4.

Ten years follow-up of histologically benign calcifications in the breast after vacuum-assisted stereotactic biopsy (VASB): Is additional mammographic follow-up warranted?

Objective

This study assessed the short-term and the long-term breast cancer rate in patients with benign histopathologic results after a vacuum-assisted stereotactic biopsy (VASB) for calcifications.

Methods

In a retrospective cohort study, all consecutive patients who had a benign diagnosis after VASB to analyze breast calcifications. Data of breast cancer development at short-term (four years) and long-term follow-up was gathered. Breast cancer rates in our cohort were compared to the breast cancer incidence in the general population.

Results

Of 1376 patients who underwent VASB to analyze breast calcifications, 823 had a benign histopathologic diagnosis. During short-term follow-up, eight patients developed breast cancer. During the mean long-term follow-up period of 9.3 ± 3.1 years, 22 patients were diagnosed with ipsilateral breast cancer. The incidence rate of breast cancer after benign biopsy was comparable to the rate in the general population.

Conclusion

In patients with VASB-confirmed benign calcifications of the breast, we found no excess incidence of ipsilateral breast cancer during ten years follow-up. Therefore, in patients with an increased risk of breast cancer (due to a history of breast cancer or familial risk) annual mammography should be sufficient. Patients with a population-based risk may be monitored via biennial mammography by the national screening program. More frequent screening would provide no benefit.

•

1% developed ipsilateral breast cancer in four-year follow-up after biopsy.

•

3% developed ipsilateral breast cancer in ten-years follow-up after biopsy.

•

No statistically significant excess incidence of breast cancer after benign calcifications.

•

A benign histopathologic result after VASB can be considered a safe decision tool.

Changes in body composition and muscle attenuation during taxane-based chemotherapy in patients with metastatic breast cancer

PURPOSE

Body composition parameters including low muscle mass, muscle attenuation (which reflects muscle quality) and adipose tissue measurements have emerged as prognostic factors in cancer patients. However, knowledge regarding the possibility of excessive muscle loss during specific systemic therapies is unknown. We describe the changes in body composition and muscle attenuation (MA) during taxane- and anthracycline-based regimens and its association with overall survival (OS) in metastatic breast cancer patients.

METHODS

The lumbar skeletal muscle index (LSMI) was used as marker of muscle mass. LSMI, MA, subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT) and intramuscular adipose tissue (IMAT) were measured before and after first-line treatment with paclitaxel (n = 73) or 5-fluorouracil-doxorubicin-cyclophosphamide (FAC) (n = 25) using CT-images. Determinants of the change of LSMI and MA were analyzed using multiple linear regression. OS was assessed using Cox proportional hazard models.

RESULTS

MA significantly decreased during paclitaxel treatment (- 0.9 HU, p = 0.03). LSMI (p = 0.40), SAT (p = 0.75), VAT (p = 0.84) and IMAT (p = 0.10) remained stable. No significant alterations in body composition parameters during FAC-treatment were observed. Previous (neo-)adjuvant chemotherapy contributed to larger loss of MA during the current treatment. Body composition changes during chemotherapy were not associated with OS.

CONCLUSIONS

MA decreased during treatment with paclitaxel, while muscle mass was stable. Body composition changes are not associated with survival in the absence of progressive disease.

Low muscle attenuation is a prognostic factor for survival in metastatic breast cancer patients treated with first line palliative chemotherapy

BACKGROUND

Low muscle mass (LMM) and low muscle attenuation (LMA) reflect low muscle quantity and low muscle quality, respectively. Both are associated with a poor outcome in several types of solid malignancies. This study determined the association of skeletal muscle measures with overall survival (OS) and time to next treatment (TNT).

PATIENTS AND METHODS

A skeletal muscle index (SMI) in cm²/m² and muscle attenuation (MA) in Hounsfield units (HU) were measured using abdominal CT-images of 166 patients before start of first-line chemotherapy for metastatic breast cancer. Low muscle mass (SMI <41 cm²/m²), sarcopenic obesity (LMM and BMI ≥30 kg/m²) and low muscle attenuation (MA <41 HU and BMI <25 kg/m2 or MA <33 HU and BMI ≥25 kg/m2) were related to OS and TNT.

RESULTS

The prevalence of LMM, sarcopenic obesity and LMA were 66.9%, 7.2% and 59.6% respectively. LMM and sarcopenic obesity showed no significant association with OS and TNT, whereas LMA was associated with both lower OS (HR 2.04, 95% CI 1.34-3.12, p = 0.001) and shorter TNT (HR 1.72, 95% CI 1.14-2.62, p = 0.010). Patients with LMA had a median OS and TNT of 15 and 8 months respectively, compared to 23 and 10 months in patients with normal MA.

CONCLUSION

LMA is a prognostic factor for OS and TNT in metastatic breast cancer patients receiving first-line palliative chemotherapy, whereas LMM and sarcopenic obesity are not. Further research is needed to establish what impact LMA should have in daily clinical practice.

Is aortoiliac calcification linked to colorectal anastomotic leakage? A case-control study

BACKGROUND

Anastomotic leakage in bowel surgery remains a devastating complication. Various risk factors have been uncovered, however, high anastomotic leakage rates are still being reported. This study describes the use of calcification markers of the central abdominal arteries as a prognostic factor for colorectal anastomotic leakage.

METHODS

This case-control study includes clinical data from three different hospitals. Calcium volume and calcium score of the aortoiliac tract were determined by CT-scan analysis. Cases were all patients with anastomotic leakage after a left-sided anastomosis (n = 30). Three controls were randomly matched for each case. Only patients with a contrast-enhanced pre-operative CT-scan were included.

RESULTS

The measurements of the calcium score and calcium volume of the different trajectories showed that there was one significant difference with regard to the right external iliac artery. Multiple regression analysis showed a significant different negative odds ratio of the presence of calcium in the right external iliac artery.

CONCLUSION

This study demonstrates that calcium volume and calcium score of the aortoiliac trajectory does not correlate with the risk of colorectal anastomotic leakage after a left-sided anastomosis.

Lower Extremity Arterial Disease: Multidetector CT Angiography—Meta-Analysis

PURPOSE

To obtain the best available estimates of the diagnostic performance of multidetector computed tomographic (CT) angiography compared with that of digital subtraction angiography (DSA) in the assessment of symptomatic lower extremity arterial disease and to identify the most important sources of variation in diagnostic performance between studies.

MATERIALS AND METHODS

Reports of studies published from January 2000 through April 2006 in English, German, French, or Spanish were searched for by using the MEDLINE, EMBASE, and Cochrane databases. Studies were included if they allowed construction of 2 x 2 contingency tables for the detection of stenosis of 50% or greater at multidetector CT angiography compared with that at DSA -- the reference standard -- in patients with claudication or critical ischemia. Two observers extracted data about study design, patient characteristics, arterial tracts, and technical protocols. Random-effects summary receiver operating characteristic analysis was performed to examine the influence of these data on diagnostic performance.

RESULTS

Of the 70 studies initially identified, 12 were included in which multidetector CT angiography was used to evaluate 9541 arterial segments in 436 patients. The pooled sensitivity and specificity for detecting a stenosis of at least 50% per segment were 92% (95% confidence interval: 89%, 95%) and 93% (95% confidence interval: 91%, 95%), respectively. Three studies provided data about the diagnostic performance of multidetector CT angiography in subdivisions of the arterial tract. The diagnostic performance of multidetector CT angiography in the infrapopliteal tract was lower than but not significantly different from that in the aortoiliac (P > .11) and femoropopliteal (P > .40) tracts. Regression analysis showed that diagnostic performance was not significantly influenced by differences in study characteristics.

CONCLUSION

Multidetector CT angiography is an accurate diagnostic test in the assessment of arterial disease (> or =50% stenosis) of the entire lower extremity.

Multi-detector row computed tomography angiography of peripheral arterial disease

With the introduction of multi-detector row computed tomography (MDCT), scan speed and image quality has improved considerably. Since the longitudinal coverage is no longer a limitation, multi-detector row computed tomography angiography (MDCTA) is increasingly used to depict the peripheral arterial runoff. Hence, it is important to know the advantages and limitations of this new non-invasive alternative for the reference test, digital subtraction angiography. Optimization of the acquisition parameters and the contrast delivery is important to achieve a reliable enhancement of the entire arterial runoff in patients with peripheral arterial disease (PAD) using fast CT scanners. The purpose of this review is to discuss the different scanning and injection protocols using 4-, 16-, and 64-detector row CT scanners, to propose effective methods to evaluate and to present large data sets, to discuss its clinical value and major limitations, and to review the literature on the validity, reliability, and cost-effectiveness of multi-detector row CT in the evaluation of PAD.

Vessel wall calcifications at multi-detector row CT angiography in patients with peripheral arterial disease: effect on clinical utility and clinical predictors

PURPOSE

To evaluate retrospectively the effect of vessel wall calcifications on the clinical utility of multi-detector row computed tomographic (CT) angiography performed in patients with peripheral arterial disease and to identify clinical predictors for the presence of vessel wall calcifications.

MATERIALS AND METHODS

The study was approved by the hospital institutional review board, and informed consent was obtained from all patients. For this study the authors included patients from two randomized controlled trials that measured the costs and effects of diagnostic imaging in patients with peripheral arterial disease. All patients underwent CT angiography and were followed up for 6 months. Clinical utility was measured on the basis of therapeutic confidence (rated on a 10-point scale) in the results of initial CT angiography and the need for additional vascular imaging. Univariable and multivariable logistic and linear regression analysis and the area under the receiver operating characteristic curve were used to evaluate the effect of vessel wall calcifications on the clinical utility of CT angiography and the use of patient characteristics to predict the number of calcified segments at CT angiography.

RESULTS

A total of 145 patients were included (mean age, 64 years; 70% men). The authors found that the number of calcified segments was a significant predictor of the need for additional imaging (P = .001) and of the confidence scores (P < .001). The number of calcified segments discriminated between patients who required additional imaging after CT angiography and those who did not (area under the receiver operating characteristic curve, 0.66; 95% confidence interval: 0.54, 0.77). Age, diabetes mellitus, and cardiac disease were significant predictors of the number of calcified segments in both the univariable and multivariable analyses (P < .05).

CONCLUSION

Vessel wall calcifications decrease the clinical utility of CT angiography in patients with peripheral arterial disease. Diabetes mellitus, cardiac disease, and elderly age (older than 84 years) are independently predictive for the presence of vessel wall calcifications.

Interobserver agreement for the interpretation of contrast-enhanced 3D MR angiography and MDCT angiography in peripheral arterial disease

OBJECTIVE

The objective of our study was to compare interobserver agreement for interpretations of contrast-enhanced 3D MR angiography and MDCT angiography in patients with peripheral arterial disease.

SUBJECTS AND METHODS

Of 226 eligible patients, 69 were excluded. The remaining 157 consecutive patients were prospectively randomized to either MR angiography (n = 78) or MDCT angiography (n = 79). Two observers independently evaluated for arterial stenosis or occlusion on MR angiography (2,157 segments) and MDCT angiography (2,419 segments) using a 5-point ordinal scale. Vessel wall calcifications were noted. Interobserver agreement for each technique was evaluated with a weighted kappa (kappa(w)) statistic.

RESULTS

Although interobserver agreement for both was excellent, the interobserver agreement for MR angiography (kappa(w) = 0.90; 95% confidence interval [CI], 0.89-0.92) was higher than that for MDCT angiography (kappa(w) = 0.85; 95% CI, 0.83-0.86) for reporting the degree of arterial stenosis or occlusion in all segments. For the different anatomic locations, the interobserver agreement for MR angiography versus MDCT angiography was as follows: aortoiliac (kappa(w) =0.91 vs 0.84, respectively), femoropopliteal (kappa(w) = 0.91 vs 0.87), and crural (kappa(w) = 0.90 vs 0.83) segments. The interobserver agreement of MDCT angiography significantly decreased in the presence of calcifications but was still good for all anatomic locations. The lowest agreement was found for crural segments in the presence of calcifications (kappa(w) = 0.67). With MR angiography, there were 12 times more nondiagnostic segments than with MDCT angiography (81 vs 7, respectively).

CONCLUSION

Interpretations of MR angiography and MDCT angiography for peripheral arterial disease have an excellent interobserver agreement. MR angiography has a higher interobserver agreement than MDCT angiography, and the presence of calcified segments significantly decreases interobserver agreement for MDCT angiography.

Contrast-enhanced MR angiography and digital subtraction angiography in living renal donors: diagnostic agreement, impact on decision making, and costs

OBJECTIVE

The objective of our study was to evaluate the diagnostic agreement, the impact on decision making, and the costs of contrast-enhanced MR angiography and digital subtraction angiography in the workup of living renal donors.

CONCLUSION

Contrast-enhanced MR angiography for the preoperative evaluation of renal donors is superior to digital subtraction angiography in revealing vascular anomalies and depicting parenchymal abnormalities and is less costly; furthermore, it does not lead to preoperative decisions that differ from those based on digital subtraction angiography. If contrast-enhanced MR angiography does not provide sufficient information to make a confident decision, an additional digital subtraction angiography examination should be performed.

Peripheral arterial disease: therapeutic confidence of CT versus digital subtraction angiography and effects on additional imaging recommendations

PURPOSE

To compare multi-detector row computed tomographic (CT) angiography and digital subtraction angiography (DSA) prior to revascularization in patients with symptomatic peripheral arterial disease for the purpose of assessing recommendations for additional imaging and physician confidence ratings for chosen therapy.

MATERIALS AND METHODS

In a randomized controlled trial, 73 patients were assigned to CT angiography, and 72 were assigned to DSA. Physician confidence in the treatment decision was measured as a continuous outcome on a scale of 0-10 (uncertain to certain) and as a dichotomous outcome (further imaging recommended, yes or no). Mean confidence scores and additional imaging recommendations were compared between CT and DSA groups in an intention-to-diagnose-and-treat analysis. To detect trends in confidence, confidence scores were plotted over time, and multiple linear regression analysis was performed. To detect trends in additional imaging recommendations, logistic regression analysis was used. Data from eligible nonrandomized patients were analyzed separately.

RESULTS

No statistically significant difference in baseline characteristics between randomized groups was found. CT had a lower confidence score than did DSA (7.2 vs 8.2, P < .001). Further imaging was recommended more often after CT (25 of 71 patients, 35%) than after DSA (nine of 66 patients, 14%; P = .003). Analysis of trends demonstrated increasing (but not statistically significant) confidence in CT and stable confidence in DSA. No significant difference was found in baseline characteristics between randomized and nonrandomized patients. Among nonrandomized patients, no significant difference in mean confidence score (8.2 vs 8.3, P = .26) was found between CT (n = 24) and DSA (n = 26).

CONCLUSION

With CT angiography, physician confidence decreases with an associated increase in additional imaging prior to revascularization in patients with symptomatic peripheral arterial disease. Given that CT is less invasive than DSA, results suggest that CT may replace DSA in selected cases.

Cost-effectiveness targets for multi-detector row CT angiography in the work-up of patients with intermittent claudication

PURPOSE

To determine the costs, sensitivity for detection of significant stenoses, and proportion of equivocal multi-detector row computed tomographic (CT) angiography results in the work-up of patients with intermittent claudication that would make this imaging examination cost-effective compared with gadolinium-enhanced magnetic resonance (MR) angiography.

MATERIALS AND METHODS

A decision model was used to compare the societal cost-effectiveness of a new imaging modality with that of gadolinium-enhanced MR angiography. Main outcome measures were quality-adjusted life years (QALYs) and lifetime costs. By using threshold analysis of a given willingness to pay per QALY, target values for costs, sensitivity for detection of significant stenoses, and proportion of cases requiring additional work-up with intraarterial digital subtraction angiography owing to equivocal results of the new modality were determined. The base case evaluated was that of 60-year-old men with severe intermittent claudication and assumed an incremental cost-effectiveness threshold of 100,000 US dollars per QALY.

RESULTS

If treatment were limited to angioplasty, a new imaging modality would be cost-effective if the costs were 300 US dollars and the sensitivity was 85%, even if up to 35% of patients needed additional work-up. When both angioplasty and bypass surgery were considered as treatment options, a new imaging modality was cost-effective if the costs were 300 US dollars, the sensitivity was higher than 94%, and 20% of patients required additional work-up.

CONCLUSION

Multi-detector row CT angiography, as compared with currently used imaging modalities such as MR angiography, has the potential to be cost-effective in the evaluation of patients with intermittent claudication.

MR imaging: a "one-stop shop" modality for preoperative evaluation of potential living kidney donors

At many institutions, magnetic resonance (MR) angiography is the technique of choice for assessment of the renal arteries and renal parenchyma in potential living kidney donors. The renal arteries and renal veins have a varied anatomy and may consist of one or more vessels at several levels with variable calibers and levels of branching. These findings may play an important role in the surgeon's decision about which kidney to harvest, especially if laparoscopic nephrectomy is used. A comprehensive MR imaging protocol is used at one hospital to assess the arteries, veins, parenchyma, and collecting system of the kidneys. The protocol includes T2-weighted single-shot fast spin-echo imaging, fat-saturated T2-weighted fast spin-echo imaging, three-dimensional MR angiography and MR venography, and delayed fat-saturated three-dimensional T1-weighted gradient-echo imaging. Meticulous assessment of the source images as well as images produced with various postprocessing methods, such as full maximum intensity projection, targeted maximum intensity projection, and axial and oblique reformation, allows detailed description of the vascular anatomy and its relationship to the collecting system and parenchyma to facilitate the surgeon's decision making. The findings of MR imaging are comparable with those of other imaging modalities.

Living renal donors: optimizing the imaging strategy--decision- and cost-effectiveness analysis

PURPOSE

To determine the most cost-effective strategy for preoperative imaging performed in potential living renal donors.

MATERIALS AND METHODS

In a decision-analytic model, the societal cost-effectiveness of digital subtraction angiography (DSA), gadolinium-enhanced magnetic resonance (MR) angiography, contrast material-enhanced spiral computed tomographic (CT) angiography, and combinations of these imaging techniques was evaluated. Outcome measures included lifetime cost, quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratios. A base-case analysis was performed with a 40-year-old female donor and a 40-year-old female recipient.

RESULTS

For the donor, MR angiography (24.05 QALYs and 9,000 dollars) dominated all strategies except for MR angiography with CT angiography, which had an incremental ratio of 245,000 dollars per QALY. For the recipient, DSA and DSA with MR angiography yielded similar results (10.46 QALYs and 179,000 dollars) and dominated all other strategies. When results for donor and recipient were combined, DSA dominated all other strategies (34.51 QALYs and 188,000 dollars). If DSA was associated with a 99% specificity or less for detection of renal disease, MR angiography with CT angiography was superior (34.47 QALYs and 190,000 dollars).

CONCLUSION

For preoperative imaging in a potential renal donor, DSA is the most cost-effective strategy if it has a specificity greater than 99% for detection of renal disease; otherwise, MR angiography with CT angiography is the most cost-effective strategy.

Purification and properties of liver fructose 1,6-bisphosphatase from C57BL/KsJ normal and diabetic mice

PURPOSE

To prospectively compare therapeutic confidence in, patient outcomes (in terms of quality of life) after, and the costs of digital subtraction angiography (DSA) with those of multi-detector row computed tomographic (CT) angiography as the initial diagnostic imaging test in patients with peripheral arterial disease (PAD).

MATERIALS AND METHODS

Institutional medical ethics committee approval and patient informed consent were obtained. Between April 2000 and August 2001, patients with PAD were randomly assigned to undergo either DSA or multi-detector row CT angiography as the initial diagnostic imaging test. Outcomes were the therapeutic confidence assessed by physicians (on a scale from 0 to 10), the need for additional imaging, the health-related quality of life at 6-month follow-up, diagnostic and therapeutic costs, and the costs for a hospital stay. Costs were computed from a hospital perspective according to Dutch guidelines for cost calculations in health care. Mean outcomes were compared between groups with unpaired t testing and were adjusted for predictive baseline characteristics with multivariable regression analysis.

RESULTS

Among the 145 patients, 72 were randomly allocated to the DSA group and 73 to the CT angiography group. One patient in the DSA group had to be excluded. Mean age was 63 years in the DSA group and 64 years in the CT angiography group. There were 47 men in the DSA group and 58 men in the CT angiography group. Physician confidence in making a correct therapeutic choice was significantly higher at DSA (mean confidence score, 8.2) than at CT angiography (mean score, 7.2; P < .001). During 6-month follow-up, 14% less additional imaging was performed in the DSA group than in the CT angiography group (P = .3). No significant quality-of-life differences were found between groups. The diagnostic cost associated with DSA (564 +/- 210 euro [standard deviation]) was significantly higher than that associated with CT angiography (363 +/- 273 euro), a difference of -201 euro (95% confidence interval: -281 euro, -120 euro; P < .001). Therapeutic and hospitalization costs were similar for both strategies.

CONCLUSION

These results suggest that use of noninvasive multi-detector row CT angiography instead of DSA as the initial diagnostic imaging test for PAD provides sufficient information for therapeutic decision making and reduces imaging costs.