Machine learning plays a role in many deployed decision systems, often in ways that are difficult or impossible to understand by human stakeholders. Explaining, in a human-understandable way, the relationship between the input and output of machine learning models is essential to the development of trustworthy machinelearning-based systems. A burgeoning body of research seeks to define the goals and methods of explainability in machine learning. In this paper, we seek to review and categorize research on counterfactual explanations, a specific class of explanation that provides a link between what could have happened had input to a model been changed in a particular way. Modern approaches to counterfactual explainability in machine learning draw connections to the established legal doctrine in many countries, making them appealing to fielded systems in high-impact areas such as finance and healthcare. Thus, we design a rubric with desirable properties of counterfactual explanation algorithms and comprehensively evaluate all currently-proposed algorithms against that rubric. Our rubric provides easy comparison and comprehension of the advantages and disadvantages of different approaches and serves as an introduction to major research themes in this field. We also identify gaps and discuss promising research directions in the space of counterfactual explainability.
Breast cancer‐related lymphedema (BCRL) has become an increasingly important clinical issue as noted by the recent update of the 2015 NCCN breast cancer guidelines which recommends to “educate, monitor, and refer for lymphedema management.” The purpose of this review was to examine the literature regarding early detection and management of BCRL in order to (1) better characterize the benefit of proactive surveillance and intervention, (2) clarify the optimal monitoring techniques, and (3) help better define patient groups most likely to benefit from surveillance programs. A Medline search was conducted for the years 1992–2015 to identify articles addressing early detection and management of BCRL. After an initial search, 127 articles were identified, with 13 of these studies focused on early intervention (three randomized (level of evidence 1), four prospective (level of evidence 2–3), six retrospective trials (level of evidence 4)). Data from two, small (n = 185 cases), randomized trials with limited follow‐up demonstrated a benefit to early intervention (physiotherapy, manual lymphatic drainage) with regard to reducing the rate of chronic BCRL (>50% reduction) with two additional studies underway (n = 1280). These findings were confirmed by larger prospective and retrospective series. Several studies were identified that demonstrate that newer diagnostic modalities (bioimpedance spectroscopy, perometry) have increased sensitivity allowing for the earlier detection of BCRL. Current data support the development of surveillance programs geared toward the early detection and management of BCRL in part due to newer, more sensitive diagnostic modalities.
BackgroundBreast cancer-related lymphedema (BCRL) represents a major source of morbidity among breast cancer survivors. Increasing data support early detection of subclinical BCRL followed by early intervention. A randomized controlled trial is being conducted comparing lymphedema progression rates using volume measurements calculated from the circumference using a tape measure (TM) or bioimpedance spectroscopy (BIS).MethodsPatients were enrolled and randomized to either TM or BIS surveillance. Patients requiring early intervention were prescribed a compression sleeve and gauntlet for 4 weeks and then re-evaluated. The primary endpoint of the trial was the rate of progression to clinical lymphedema requiring complex decongestive physiotherapy (CDP), with progression defined as a TM volume change in the at-risk arm ≥ 10% above the presurgical baseline. This prespecified interim analysis was performed when at least 500 trial participants had ≥ 12 months of follow-up.ResultsA total of 508 patients were included in this analysis, with 109 (21.9%) patients triggering prethreshold interventions. Compared with TM, BIS had a lower rate of trigger (15.8% vs. 28.5%, p < 0.001) and longer times to trigger (9.5 vs. 2.8 months, p = 0.002). Twelve triggering patients progressed to CDP (10 in the TM group [14.7%] and 2 in the BIS group [4.9%]), representing a 67% relative reduction and a 9.8% absolute reduction (p = 0.130).
ConclusionsInterim results demonstrated that post-treatment surveillance with BIS reduced the absolute rates of progression of BCRL requiring CDP by approximately 10%, a clinically meaningful improvement. These results support the concept of post-treatment surveillance with BIS to detect subclinical BCRL and initiate early intervention.
The final analysis of treatment efficacy, cosmesis, and toxicity from the American Society of Breast Surgeons MammoSite(®) breast brachytherapy registry trial confirms previously noted excellent results and compares favorably with other forms of APBI with similar follow-up and to outcomes seen in selected patients treated with whole breast irradiation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.