Evaluating semi-supervision methods for medical image segmentation: applications in cardiac magnetic resonance imaging.

Purpose: Neural networks have potential to automate medical image segmentation but require expensive labeling efforts. While methods have been proposed to reduce the labeling burden, most have not been thoroughly evaluated on large, clinical datasets or clinical tasks. We propose a method to train segmentation networks with limited labeled data and focus on thorough network evaluation. Approach: We propose a semi-supervised method that leverages data augmentation, consistency regularization, and pseudolabeling and train four cardiac magnetic resonance (MR) segmentation networks. We evaluate the models on multiinstitutional, multiscanner, multidisease cardiac MR datasets using five cardiac functional biomarkers, which are compared to an expert's measurements using Lin's concordance correlation coefficient (CCC), the within-subject coefficient of variation (CV), and the Dice coefficient. Results: The semi-supervised networks achieve strong agreement using Lin's CCC ( > 0.8 ), CV similar to an expert, and strong generalization performance. We compare the error modes of the semi-supervised networks against fully supervised networks. We evaluate semi-supervised model performance as a function of labeled training data and with different types of model supervision, showing that a model trained with 100 labeled image slices can achieve a Dice coefficient within 1.10% of a network trained with 16,000+ labeled image slices. Conclusion: We evaluate semi-supervision for medical image segmentation using heterogeneous datasets and clinical metrics. As methods for training models with little labeled data become more common, knowledge about how they perform on clinical tasks, how they fail, and how they perform with different amounts of labeled data is useful to model developers and users.

Use of Telephone- and Internet-Based Support to Elicit and Address Financial Abuse and Mismanagement in Dementia: Experiences from the Care Ecosystem Study.

BACKGROUND: Financial mismanagement and abuse in dementia have serious consequences for patients and their families. Vulnerability to these outcomes reflects both patient and contextual factors. OBJECTIVE: Our study aimed to assess how multidisciplinary care coordination programs assist families in addressing psychosocial vulnerabilities and accessing needed resources. METHODS: Our study was embedded in a clinical trial of the Care Ecosystem, a telephone- and internet-based supportive care intervention for patients with dementia and caregivers. This program is built around the role of the Care Team Navigator (CTN), an unlicensed dementia care guide who serves as the patient and caregiver's primary point of contact, screening for common problems and providing support. We conducted a qualitative analysis of case summaries from a subset of 19 patient/caregiver dyads identified as having increased risk for financial mismanagement and abuse, to examine how Care Ecosystem staff identified vulnerabilities and provided support to patients and families. RESULTS: CTNs elicited patient and caregiver needs using templated conversations to address common financial and legal planning issues in dementia. Sources of financial vulnerability included changes in patients' behavior, caregiver burden, intrafamily tension, and confusion about resources to facilitate end-of-life planning. The Care Ecosystem staff's rapport with their dyads helped them address these issues by providing emotional support, information on how to access financial, medical, and legal resources, and improving intra-familial communication. CONCLUSION: The Care Ecosystem offers a scalable way to address vulnerabilities to financial mismanagement and abuse in patients and caregivers through coordinated care by unlicensed care guides supported by a multidisciplinary team.

Goals of Care Conversations and Subsequent Advance Care Planning Outcomes for People with Dementia.

BACKGROUND: Advance care planning has been shown to improve end of life decision-making for people with dementia. However, the impact of goals of care conversations between people with dementia and their caregivers has not been characterized. OBJECTIVE: In this study, we evaluate the association between goals of care conversations and advance care planning outcomes. METHODS: Retrospective advance care planning measures were collected via a questionnaire administered to 166 caregivers after the death of the person with dementia for whom they provided care. RESULTS: At time of death, the majority of decedents with dementia had advance directives, health care agents, and previous goals of care conversations with their caregiver. Goals of care conversations were significantly associated with the perceived usefulness of advance directives, the perceived adherence to advance directives, and decedent dying at their desired place of death, but not with disagreements around end-of-life care. CONCLUSION: Our findings suggest that goals of care conversations are an important component of advance care planning. These findings support the development of interventions that facilitate such conversations between people with dementia and their caregivers.

Impact of Upstream Medical Image Processing on Downstream Performance of a Head CT Triage Neural Network.

PURPOSE: To develop a convolutional neural network (CNN) to triage head CT (HCT) studies and investigate the effect of upstream medical image processing on the CNN's performance. MATERIALS AND METHODS: A total of 9776 HCT studies were retrospectively collected from 2001 through 2014, and a CNN was trained to triage them as normal or abnormal. CNN performance was evaluated on a held-out test set, assessing triage performance and sensitivity to 20 disorders to assess differential model performance, with 7856 CT studies in the training set, 936 in the validation set, and 984 in the test set. This CNN was used to understand how the upstream imaging chain affects CNN performance by evaluating performance after altering three variables: image acquisition by reducing the number of x-ray projections, image reconstruction by inputting sinogram data into the CNN, and image preprocessing. To evaluate performance, the DeLong test was used to assess differences in the area under the receiver operating characteristic curve (AUROC), and the McNemar test was used to compare sensitivities. RESULTS: The CNN achieved a mean AUROC of 0.84 (95% CI: 0.83, 0.84) in discriminating normal and abnormal HCT studies. The number of x-ray projections could be reduced by 16 times and the raw sensor data could be input into the CNN with no statistically significant difference in classification performance. Additionally, CT windowing consistently improved CNN performance, increasing the mean triage AUROC by 0.07 points. CONCLUSION: A CNN was developed to triage HCT studies, which may help streamline image evaluation, and the means by which upstream image acquisition, reconstruction, and preprocessing affect downstream CNN performance was investigated, bringing focus to this important part of the imaging chain.Keywords Head CT, Automated Triage, Deep Learning, Sinogram, DatasetSupplemental material is available for this article.© RSNA, 2021.

Multiparametric Photoacoustic Analysis of Human Thyroid Cancers In Vivo.

Thyroid cancer is one of the most common cancers, with a global increase in incidence rate for both genders. Ultrasound-guided fine-needle aspiration is the current gold standard to diagnose thyroid cancers, but the results are inaccurate, leading to repeated biopsies and unnecessary surgeries. To reduce the number of unnecessary biopsies, we explored the use of multiparametric photoacoustic (PA) analysis in combination with the American Thyroid Association (ATA) Guideline (ATAP). In this study, we performed in vivo multispectral PA imaging on thyroid nodules from 52 patients, comprising 23 papillary thyroid cancer (PTC) and 29 benign cases. From the multispectral PA data, we calculated hemoglobin oxygen saturation level in the nodule area, then classified the PTC and benign nodules with multiparametric analysis. Statistical analyses showed that this multiparametric analysis of multispectral PA responses could classify PTC nodules. Combining the photoacoustically indicated probability of PTC and the ATAP led to a new scoring method that achieved a sensitivity of 83% and a specificity of 93%. This study is the first multiparametric analysis of multispectral PA data of thyroid nodules with statistical significance. As a proof of concept, the results show that the proposed new ATAP scoring can help physicians examine thyroid nodules for fine-needle aspiration biopsy, thus reducing unnecessary biopsies. SIGNIFICANCE: This report highlights a novel photoacoustic scoring method for risk stratification of thyroid nodules, where malignancy of the nodules can be diagnosed with 83% sensitivity and 93% specificity.

Effect of Collaborative Dementia Care via Telephone and Internet on Quality of Life, Caregiver Well-being, and Health Care Use: The Care Ecosystem Randomized Clinical Trial.

Importance: Few health systems have adopted effective dementia care management programs. The Care Ecosystem is a model for delivering care from centralized hubs across broad geographic areas to caregivers and persons with dementia (PWDs) independently of their health system affiliations. Objective: To determine whether the Care Ecosystem is effective in improving outcomes important to PWDs, their caregivers, and payers beyond those achieved with usual care. Design, Setting, and Participants: A single-blind, randomized clinical trial with a pragmatic design was conducted among PWDs and their caregivers. Each PWD-caregiver dyad was enrolled for 12 months between March 20, 2015, and February 28, 2017. Data were collected until March 5, 2018. Study interventions and assessments were administered over the telephone and internet by clinical and research teams in San Francisco, California, and Omaha, Nebraska. Of 2585 referred or volunteer PWD-caregiver dyads in California, Iowa, or Nebraska, 780 met eligibility criteria and were enrolled. A total of 512 PWD-caregiver dyads were randomized to receive care through the Care Ecosystem and 268 dyads to receive usual care. All eligible PWDs had a dementia diagnosis; were enrolled or eligible for enrollment in Medicare or Medicaid; and spoke English, Spanish, or Cantonese. Analyses were intention-to-treat. Intervention: Telephone-based collaborative dementia care was delivered by a trained care team navigator, who provided education, support and care coordination with a team of dementia specialists (advanced practice nurse, social worker, and pharmacist). Main Outcomes and Measures: Primary outcome measure: Quality of Life in Alzheimer's Disease based on caregiver's rating of 13 aspects of PWD's well-being (including physical health, energy level, mood, living situation, memory, relationships, and finances) on a 4-point scale (poor to excellent). Secondary outcomes: frequencies of PWDs' use of emergency department, hospitalization, and ambulance services; caregiver depression (score on 9-Item Patient Health Questionnaire; higher scores indicate more severe depression); and caregiver burden (score on 12-Item Zarit Burden Interview; higher scores indicate more severe caregiver burden). Results: The 780 PWDs (56.3% female; mean [SD] age, 78.1 [9.9] years) and 780 caregivers (70.9% female; mean [SD] age, 64.7 [12.0] years) lived in California (n = 452), Nebraska (n = 284), or Iowa (n = 44). Of 780 dyads, 655 were still active at 12 months, and 571 completed the 12-month survey. Compared with usual care, the Care Ecosystem improved PWD quality of life (B, 0.53; 95% CI, 0.25-1.30; P = .04), reduced emergency department visits (B, -0.14; 95% CI, -0.29 to -0.01; P = .04), and decreased caregiver depression (B, -1.14; 95% CI, -2.15 to -0.13; P = .03) and caregiver burden (B, -1.90; 95% CI, -3.89 to -0.08; P = .046). Conclusions and Relevance: Effective care management for dementia can be delivered from centralized hubs to supplement usual care and mitigate the growing societal and economic burdens of dementia. Trial Registration: ClinicalTrials.gov identifier: NCT02213458.

Development of an adaptive, personalized, and scalable dementia care program: Early findings from the Care Ecosystem.

Katherine Possin and colleagues report on the implementation, development, and early findings of the Care Ecosystem, an adaptive, personalized, and scalable dementia care program.

Lost in translation: the unintended consequences of advance directive law on clinical care.

BACKGROUND: Advance directive law may compromise the clinical effectiveness of advance directives. PURPOSE: To identify unintended legal consequences of advance directive law that may prevent patients from communicating end-of-life preferences. DATA SOURCES: Advance directive legal statutes for all 50 U.S. states and the District of Columbia and English-language searches of LexisNexis, Westlaw, and MEDLINE from 1966 to August 2010. STUDY SELECTION: Two independent reviewers selected 51 advance directive statutes and 20 articles. Three independent legal reviewers selected 105 legal proceedings. DATA EXTRACTION: Two reviewers independently assessed data sources and used critical content analysis to determine legal barriers to the clinical effectiveness of advance directives. Disagreements were resolved by consensus. DATA SYNTHESIS: Legal and content-related barriers included poor readability (that is, laws in all states were written above a 12th-grade reading level), health care agent or surrogate restrictions (for example, 40 states did not include same-sex or domestic partners as default surrogates), and execution requirements needed to make forms legally valid (for example, 35 states did not allow oral advance directives, and 48 states required witness signatures, a notary public, or both). Vulnerable populations most likely to be affected by these barriers included patients with limited literacy, limited English proficiency, or both who cannot read or execute advance directives; same-sex or domestic partners who may be without legally valid and trusted surrogates; and unbefriended, institutionalized, or homeless patients who may be without witnesses and suitable surrogates. LIMITATION: Only appellate-level legal cases were available, which may have excluded relevant cases. CONCLUSION: Unintended negative consequences of advance directive legal restrictions may prevent all patients, and particularly vulnerable patients, from making and communicating their end-of-life wishes and having them honored. These restrictions have rendered advance directives less clinically useful. Recommendations include improving readability, allowing oral advance directives, and eliminating witness or notary requirements. PRIMARY FUNDING SOURCE: U.S. Department of Veterans Affairs and the Pfizer Foundation.

POT1-interacting protein PIP1: a telomere length regulator that recruits POT1 to the TIN2/TRF1 complex.

Human telomere length is controlled by a negative feedback loop based on the binding of TRF1 to double-stranded telomeric DNA. The TRF1 complex recruits POT1, a single-stranded telomeric DNA-binding protein necessary for cis-inhibition of telomerase. By mass spectrometry, we have identified a new telomeric protein, which we have named POT1-interacting protein 1 (PIP1). PIP1 bound both POT1 and the TRF1-interacting factor TIN2 and could tether POT1 to the TRF1 complex. Reduction of PIP1 or POT1 levels with shRNAs led to telomere elongation, indicating that PIP1 contributes to telomere length control through recruitment of POT1.