Application of deep learning on mammographies to discriminate between low and high-risk DCIS for patient participation in active surveillance trials.

BACKGROUND: Ductal Carcinoma In Situ (DCIS) can progress to invasive breast cancer, but most DCIS lesions never will. Therefore, four clinical trials (COMET, LORIS, LORETTA, AND LORD) test whether active surveillance for women with low-risk Ductal carcinoma In Situ is safe (E. S. Hwang et al., BMJ Open, 9: e026797, 2019, A. Francis et al., Eur J Cancer. 51: 2296-2303, 2015, Chizuko Kanbayashi et al. The international collaboration of active surveillance trials for low-risk DCIS (LORIS, LORD, COMET, LORETTA),  L. E. Elshof et al., Eur J Cancer, 51, 1497-510, 2015). Low-risk is defined as grade I or II DCIS. Because DCIS grade is a major eligibility criteria in these trials, it would be very helpful to assess DCIS grade on mammography, informed by grade assessed on DCIS histopathology in pre-surgery biopsies, since surgery will not be performed on a significant number of patients participating in these trials. OBJECTIVE: To assess the performance and clinical utility of a convolutional neural network (CNN) in discriminating high-risk (grade III) DCIS and/or Invasive Breast Cancer (IBC) from low-risk (grade I/II) DCIS based on mammographic features. We explored whether the CNN could be used as a decision support tool, from excluding high-risk patients for active surveillance. METHODS: In this single centre retrospective study, 464 patients diagnosed with DCIS based on pre-surgery biopsy between 2000 and 2014 were included. The collection of mammography images was partitioned on a patient-level into two subsets, one for training containing 80% of cases (371 cases, 681 images) and 20% (93 cases, 173 images) for testing. A deep learning model based on the U-Net CNN was trained and validated on 681 two-dimensional mammograms. Classification performance was assessed with the Area Under the Curve (AUC) receiver operating characteristic and predictive values on the test set for predicting high risk DCIS-and high-risk DCIS and/ or IBC from low-risk DCIS. RESULTS: When classifying DCIS as high-risk, the deep learning network achieved a Positive Predictive Value (PPV) of 0.40, Negative Predictive Value (NPV) of 0.91 and an AUC of 0.72 on the test dataset. For distinguishing high-risk and/or upstaged DCIS (occult invasive breast cancer) from low-risk DCIS a PPV of 0.80, a NPV of 0.84 and an AUC of 0.76 were achieved. CONCLUSION: For both scenarios (DCIS grade I/II vs. III, DCIS grade I/II vs. III and/or IBC) AUCs were high, 0.72 and 0.76, respectively, concluding that our convolutional neural network can discriminate low-grade from high-grade DCIS.

Locoregional Breast Cancer Recurrence in the European Organisation for Research and Treatment of Cancer 10041/BIG 03-04 MINDACT Trial: Analysis of Risk Factors Including the 70-Gene Signature.

PURPOSE: A number of studies are currently investigating de-escalation of radiation therapy in patients with a low risk of in-breast relapses on the basis of clinicopathologic factors and molecular tests. We evaluated whether 70-gene risk score is associated with risk of locoregional recurrence (LRR) and estimated 8-year cumulative incidences for LRR in patients with early-stage breast cancer treated with breast conservation. METHODS: In this exploratory substudy of European Organisation for Research and Treatment of Cancer 10041/BIG 03-04 MINDACT trial, we evaluated women with a known clinical and genomic 70-gene risk score test result and who had breast-conserving surgery (BCS). The primary end point was LRR at 8 years, estimated by cumulative incidences. Distant metastasis and death were considered competing risks. RESULTS: Among 6,693 enrolled patients, 5,470 (81.7%) underwent BCS, of whom 98% received radiotherapy. At 8-year follow-up, 189 patients experienced a LRR, resulting in an 8-year cumulative incidence of 3.2% (95% CI, 2.7 to 3.7). In patients with a low-risk 70-gene signature, the 8-year LRR incidence was 2.7% (95% CI, 2.1 to 3.3). In univariable analysis, adjusted for chemotherapy, five of 12 variables were associated with LRR, including the 70-gene signature. In multivariable modeling, adjuvant endocrine therapy and to a lesser extent tumor size and grade remained significantly associated with LRR. CONCLUSION: This exploratory analysis of the MINDACT trial estimated an 8-year low LRR rate of 3.2% after BCS. The 70-gene signature was not independently predictive of LRR perhaps because of the low number of events observed and currently cannot be used in clinical decision making regarding LRR. The overall low number of events does provide an opportunity to design trials toward de-escalation of local therapy.

The effects of contemporary treatment of DCIS on the risk of developing an ipsilateral invasive Breast cancer (iIBC) in the Dutch population.

PURPOSE: To assess the effects of contemporary treatment of ductal carcinoma in situ (DCIS) on the risk of developing an ipsilateral invasive breast cancer (iIBC) in the Dutch female population. METHODS: Clinical data was obtained from the Netherlands Cancer Registry (NCR), a nationwide registry of all primary malignancies in the Netherlands integrated with the data from PALGA, the Dutch nationwide network and registry of histo- and cytopathology in the Netherlands, on all women in the Netherlands treated for primary DCIS from 2005 to 2015, resulting in a population-based cohort of 14.419 women. Cumulative iIBC incidence was assessed and associations of DCIS treatment type with subsequent iIBC risk were evaluated by multivariable Cox regression analyses. RESULTS: Ten years after DCIS diagnosis, the cumulative incidence of iIBC was 3.1% (95% CI: 2.6-3.5%) in patients treated by breast conserving surgery (BCS) plus radiotherapy (RT), 7.1% (95% CI: 5.5-9.1) in patients treated by BCS alone, and 1.6% (95% CI: 1.3-2.1) in patients treated by mastectomy. BCS was associated with a significantly higher risk for iIBC compared to BCS + RT during the first 5 years after treatment (HR 2.80, 95% CI: 1.91-4.10%). After 5 years of follow-up, the iIBC risk declined in the BCS alone group but remained higher than the iIBC risk in the BCS + RT group (HR 1.73, 95% CI: 1.15-2.61). CONCLUSIONS: Although absolute risks of iIBC were low in patients treated for DCIS with either BCS or BCS + RT, risks remained higher in the BCS alone group compared to patients treated with BCS + RT for at least 10 years after DCIS diagnosis.

The impact of patient characteristics and lifestyle factors on the risk of an ipsilateral event after a primary DCIS: A systematic review.

OBJECTIVE: The majority of 'low-risk' (grade I/II) Ductal Carcinoma In Situ (DCIS) may not progress to invasive breast cancer during a women's lifetime. Therefore, the safety of active surveillance versus standard surgical treatment for DCIS is prospectively being evaluated in clinical trials. If proven safe and selectively implemented in clinical practice, a significant group of women with low-risk DCIS may forego surgery and radiotherapy in the future. Identification of modifiable and non-modifiable risk factors associated with prognosis after a primary DCIS would also enhance our care of women with low-risk DCIS. METHODS: To identify modifiable and non-modifiable risk factors for subsequent breast events after DCIS, we performed a systematic literature search in PUBMED, EMBASE and Scopus. RESULTS: Six out of the 3870 articles retrieved were included for final data extraction. These six studies included a total of 4950 patients with primary DCIS and 640 recorded subsequent breast events. There was moderate evidence for an association of a family history of breast cancer, premenopausal status, high BMI, and high breast density with a subsequent breast cancer or further DCIS. CONCLUSION: There is a limited number of recent studies published on the impact of modifiable and non-modifiable risk factors on subsequent events after DCIS. The available evidence is insufficient to identify potential targets for risk reduction strategies, reflecting the relatively small numbers and the lack of long-term follow-up in DCIS, a low-event condition.

Posterior circumflex humeral artery pathology and digital ischemia in elite volleyball: Symptoms, risk factors & suggestions for clinical management.

OBJECTIVES: To assess the association between posterior circumflex humeral artery (PCHA) pathology (PCHAP), symptoms and associated risk factors, in elite volleyball players, and to suggest profiles for clinical management and monitoring. DESIGN: Cross-sectional study. METHODS: A questionnaire assessed symptoms of digital ischemia (DI) in the dominant hand and risk factors among 278 elite indoor and beach volleyball players of whom 6.1% (17/278) was diagnosed with PCHAP using ultrasound. Odds Ratios (OR) including 95% confidence intervals (95%CI) were calculated using binary logistic regression. RESULTS: All 278 players completed the questionnaire. Three participants with PCHAP were symptomatic (18%). Ninety-three of 96 symptomatic participants had no PCHAP (OR=0.39; 95% CI 0.13-1.13). Total years playing volleyball (OR 1.14; 95% CI 1.03-1.25) and age (OR 1.17; 95% CI 1.00-1.29) were dose-response related risk factors: a volleyball career of ≥17 years and age of ≥27 years were associated with a 9-fold and 14-fold increased risk of PCHAP, respectively. CONCLUSIONS: The volleyball career duration and age are dose-response related risk factors for PCHAP among elite indoor and beach volleyball players. DI symptoms are prevalent in a minority of athletes with PCHAP (3/17; 18%). To enable worldwide standardized care for these athletes at risk, four profiles for clinical management and monitoring have been suggested based on questionnaire and ultrasound outcomes.

Diagnostic properties of the SPIQuestionnaire to detect Posterior Circumflex Humeral Artery Disease in elite volleyball players: a cross-sectional study.

OBJECTIVES: Sports related aneurysmal degeneration and thrombosis of the Posterior Circumflex Humeral Artery (PCHA) has been known to cause symptoms of digital ischemia (DI) in elite volleyball players. Studies have reported symptoms of DI in as much as 28% of the elite indoor volleyball players. The purpose of this study was to determine the diagnostic value of the Shoulder PCHA Pathology and digital Ischemia - Questionnaire (SPI-Q) for detection of sports related PCHA disease using ultrasound data as the standard of reference. METHODS: The SPI-Q was completed by elite indoor volleyball players from the highest and single highest Dutch volleyball division and by elite beach volleyball players participating in the 2014 Grand Slam Beach Tournament The Hague (GSBTH). Ultrasound assessment of the dominant shoulder was performed on-site using the SPI-US protocol. The SPI-Q sensitivity, specificity, positive - and negative predictive value and positive - and negative likelihood ratios, and the diagnostic odds ratio were calculated for detection of sports related PCHA disease, using ultrasound as the standard of reference. RESULTS: Two hundred twenty-four elite male indoor volleyball players from the Dutch division were included in this study and 62 elite male and female beach volleyball players participating in the GSBTH: a total of 278 players. Thirty-five percent of the players reported symptoms of DI. The prevalence of PCHA disease was 6.1%. For the SPI-Q we found a sensitivity of 18% (95% CI 4-43), specificity of 64% (95% CI 58-70), positive predictive value of 3% (95% CI 0.7-8.9) and negative predictive value of 92% (95% CI 87-96), positive likelihood ratio of 0.50 (95% CI 0.18-1.40), negative likelihood ratio of 1.28 (95% CI 1.01-1.62) and a diagnostic odds ratio of 0.39 (95% CI 0.11-1.38). CONCLUSION: The diagnostic value of the SPI-Q to detect PCHA disease in elite volleyball players is poor, which makes it unsuitable as a diagnostic instrument for sports related PCHA disease specifically. However, it can be used to assess all-cause symptoms of DI and raise awareness within athletes and sports physicians, which is important for preventing ischemic complications.

Ultrasound assessment of the posterior circumflex humeral artery in elite volleyball players: Aneurysm prevalence, anatomy, branching pattern and vessel characteristics.

OBJECTIVES: To determine the prevalence of posterior circumflex humeral artery (PCHA) aneurysms and vessel characteristics of the PCHA and deep brachial artery (DBA) in elite volleyball players. METHODS: Two-hundred and eighty players underwent standardized ultrasound assessment of the dominant arm by a vascular technologist. Assessment included determination of PCHA aneurysms (defined as segmental vessel dilatation ≥150 %), PCHA and DBA anatomy, branching pattern, vessel course and diameter. RESULTS: The PCHA and DBA were identified in 100 % and 93 % (260/280) of cases, respectively. The prevalence of PCHA aneurysms was 4.6 % (13/280). All aneurysms were detected in proximal PCHA originating from the axillary artery (AA). The PCHA originated from the AA in 81 % of cases (228/280), and showed a curved course dorsally towards the humeral head in 93 % (211/228). The DBA originated from the AA in 73 % of cases (190/260), and showed a straight course parallel to the AA in 93 % (177/190). CONCLUSIONS: PCHA aneurysm prevalence in elite volleyball players is high and associated with a specific branching type: a PCHA that originates from the axillary artery. Radiologists should have a high index of suspicion for this vascular overuse injury. For the first time vessel characteristics and reference values are described to facilitate ultrasound assessment. KEY POINTS: • Prevalence of PCHA aneurysms is 4.6 % among elite volleyball players. • All aneurysms are in proximal PCHA that originates directly from AA. • Vessel characteristics and reference values are described to facilitate US assessment. • Mean PCHA and DBA diameters can be used as reference values. • Radiologists need a high index of suspicion for this vascular overuse injury.

Self-reported symptoms and risk factors for digital ischaemia among international world-class beach volleyball players.

The prevalence of ischaemia-related symptoms is remarkably high among elite indoor volleyball players. Since the exposure to sport-specific demands may be higher in beach volleyball compared to indoor volleyball, the aim of this study was to assess the prevalence of ischaemia-related symptoms and associated risk factors among world-class beach volleyball players. Therefore, a questionnaire survey was performed among beach volleyball players active during the 2013 Grand Slam Beach Volleyball in the Netherlands. In total, 60 of the 128 beach volleyball players (47%) participated: 26 males and 34 females from 17 countries. The self-reported prevalence of cold or blue or pale digits in the dominant hand during or immediately after practice or competition was 38% (n = 23). Two risk factors were independently associated with symptoms of blue or pale digits: more than 14 years playing volleyball (odds ratio (OR) 4.42, 90% confidence interval (90% CI) 1.30-15.07) and sex (female) (OR 4.62, 90% CI 1.15-18.57). In conclusion, the prevalence of symptoms associated with digital ischaemia is high among international world-class beach volleyball players. Female sex and the length of the volleyball career were independently associated with an increased risk of ischaemia-related symptoms. The high prevalence of these seemingly innocuous symptoms and possible associated risk factors warrant regular monitoring since early detection can potentially prevent thromboembolic complications and irreversible tissue damage.

Reproducibility of the SPI-US protocol for ultrasound diameter measurements of the Posterior Circumflex Humeral Artery and Deep Brachial Artery: an inter-rater reliability study.

OBJECTIVES: Elite overhead athletes are at risk of posterior circumflex humeral artery (PCHA) degeneration, aneurysm formation and thrombosis. Identification of the proximal PCHA and the nearby originating deep brachial artery (DBA) can be a challenge, even among experienced sonographers. The aim of this study was to assess the accuracy and precision of a newly designed standardized ultrasound (US) protocol (SPI-US) for assessment of the PCHA and DBA. METHODS: Two experienced sonographers determined diameters of the PCHA and DBA using the SPI-US protocol. Inter-observer agreement was evaluated using intra-class correlation coefficient (ICC), standard error of measurement (SEM), minimal detectable change (MDC), Bland-Altman (BA) analysis, and variance component (VARCOMP) analysis. RESULTS: Thirty-three healthy volunteers participated. The ICC for diameter measurement of the PCHA and DBA were 0.70 (95 %CI 0.50-0.83) and 0.60 (95 %CI 0.30-0.80), respectively. The SEM for the PCHA and DBA was 0.32 mm and 0.29 mm and MDC was 0.90 mm and 0.80 mm, respectively. The BA and VARCOMP analyses showed no systematic and only marginal sonographer bias. CONCLUSIONS: The SPI-US protocol is accurate and precise for PCHA and DBA diameter assessment in cases where they originate from the axillary artery. PCHA and DBA diameter measurements are sonographer-independent using the SPI-US-protocol. KEY POINTS: • PCHA & DBA diameter assessment is accurate and reliable using the SPI-US protocol • PCHA & DBA diameter measurements are sonographer-independent using the SPI-US protocol • The SPI-US protocol minimal detectable change is 0.90 mm for PCHA diameter measurement • This minimal detectable change enables detection of PCHA aneurysms • First step towards international periodic surveillance of athletes at risk of PCHA-injury.