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.

Opportunities for personalised follow-up in breast cancer: the gap between daily practice and recurrence risk.

PURPOSE: Follow-up guidelines barely diverge from a one-size-fits-all approach, even though the risk of recurrence differs per patient. However, the personalization of breast cancer care improves outcomes for patients. This study explores the variation in follow-up pathways in the Netherlands using real-world data to determine guideline adherence and the gap between daily practice and risk-based surveillance, to demonstrate the benefits of personalized risk-based surveillance compared with usual care. METHODS: Patients with stage I-III invasive breast cancer who received surgical treatment in a general hospital between 2005 and 2020 were selected from the Netherlands Cancer Registry and included all imaging activities during follow-up from hospital-based electronic health records. Process analysis techniques were used to map patients and activities to investigate the real-world utilisation of resources and identify the opportunities for improvement. The INFLUENCE 2.0 nomogram was used for risk prediction of recurrence. RESULTS: In the period between 2005 and 2020, 3478 patients were included with a mean follow-up of 4.9 years. In the first 12 months following treatment, patients visited the hospital between 1 and 5 times (mean 1.3, IQR 1-1) and received between 1 and 9 imaging activities (mean 1.7, IQR 1-2). Mammogram was the prevailing imaging modality, accounting for 70% of imaging activities. Patients with a low predicted risk of recurrence visited the hospital more often. CONCLUSIONS: Deviations from the guideline were not in line with the risk of recurrence and revealed a large gap, indicating that it is hard for clinicians to accurately estimate this risk and therefore objective risk predictions could bridge this gap.

Systematic Review of Health Economic Evaluations Focused on Artificial Intelligence in Healthcare: The Tortoise and the Cheetah.

OBJECTIVES: This study aimed to systematically review recent health economic evaluations (HEEs) of artificial intelligence (AI) applications in healthcare. The aim was to discuss pertinent methods, reporting quality and challenges for future implementation of AI in healthcare, and additionally advise future HEEs. METHODS: A systematic literature review was conducted in 2 databases (PubMed and Scopus) for articles published in the last 5 years. Two reviewers performed independent screening, full-text inclusion, data extraction, and appraisal. The Consolidated Health Economic Evaluation Reporting Standards and Philips checklist were used for the quality assessment of included studies. RESULTS: A total of 884 unique studies were identified; 20 were included for full-text review, covering a wide range of medical specialties and care pathway phases. The most commonly evaluated type of AI was automated medical image analysis models (n = 9, 45%). The prevailing health economic analysis was cost minimization (n = 8, 40%) with the costs saved per case as preferred outcome measure. A total of 9 studies (45%) reported model-based HEEs, 4 of which applied a time horizon >1 year. The evidence supporting the chosen analytical methods, assessment of uncertainty, and model structures was underreported. The reporting quality of the articles was moderate as on average studies reported on 66% of Consolidated Health Economic Evaluation Reporting Standards items. CONCLUSIONS: HEEs of AI in healthcare are limited and often focus on costs rather than health impact. Surprisingly, model-based long-term evaluations are just as uncommon as model-based short-term evaluations. Consequently, insight into the actual benefits offered by AI is lagging behind current technological developments.

Diagnostics in Patients Suspect for Breast Cancer in The Netherlands.

The goal of this study was to describe the variation in hospital-based diagnostic care activities for patients with symptomatology suspect for breast cancer in The Netherlands. Two cohorts were included: the 'benign' cohort (30,334 women suspected of, but without breast cancer) and the 'malignant' cohort (2236 breast cancer patients). Hospital-based financial data was combined with tumor data (malignant cohort) from The Netherlands Cancer Registry. Patterns within diagnostic pathways were analyzed. Factors influencing the number of visits and number of diagnostic care activities until diagnosis were identified in the malignant cohort with multivariable Cox and Poisson regression models. Compared to patients with benign diagnosis, patients with malignant disease received their diagnosis less frequently in one day, after an equal average number of hospital visits and higher average number of diagnostic activities. Factors increasing the number of diagnostic care activities were the following: lower age and higher cM-and cN-stages. Factors increasing the number of days until (malignant) diagnosis were as follows: higher BIRADS-score, screen-detected and higher cN-and cT-stages. Hospital of diagnosis influenced both number of activities and days to diagnosis. The diagnostic care pathway of patients with malignant disease required more time and diagnostic activities than benign disease and depends on hospital, tumor and patient characteristics.