Defining and Addressing Research Priorities in Cancer Cachexia through Transdisciplinary Collaboration.

For many patients, the cancer continuum includes a syndrome known as cancer-associated cachexia (CAC), which encompasses the unintended loss of body weight and muscle mass, and is often associated with fat loss, decreased appetite, lower tolerance and poorer response to treatment, poor quality of life, and reduced survival. Unfortunately, there are no effective therapeutic interventions to completely reverse cancer cachexia and no FDA-approved pharmacologic agents; hence, new approaches are urgently needed. In May of 2022, researchers and clinicians from Moffitt Cancer Center held an inaugural retreat on CAC that aimed to review the state of the science, identify knowledge gaps and research priorities, and foster transdisciplinary collaborative research projects. This review summarizes research priorities that emerged from the retreat, examples of ongoing collaborations, and opportunities to move science forward. The highest priorities identified include the need to (1) evaluate patient-reported outcome (PRO) measures obtained in clinical practice and assess their use in improving CAC-related outcomes; (2) identify biomarkers (imaging, molecular, and/or behavioral) and novel analytic approaches to accurately predict the early onset of CAC and its progression; and (3) develop and test interventions (pharmacologic, nutritional, exercise-based, and through mathematical modeling) to prevent CAC progression and improve associated symptoms and outcomes.

Failure Detection in Deep Neural Networks for Medical Imaging.

Deep neural networks (DNNs) have started to find their role in the modern healthcare system. DNNs are being developed for diagnosis, prognosis, treatment planning, and outcome prediction for various diseases. With the increasing number of applications of DNNs in modern healthcare, their trustworthiness and reliability are becoming increasingly important. An essential aspect of trustworthiness is detecting the performance degradation and failure of deployed DNNs in medical settings. The softmax output values produced by DNNs are not a calibrated measure of model confidence. Softmax probability numbers are generally higher than the actual model confidence. The model confidence-accuracy gap further increases for wrong predictions and noisy inputs. We employ recently proposed Bayesian deep neural networks (BDNNs) to learn uncertainty in the model parameters. These models simultaneously output the predictions and a measure of confidence in the predictions. By testing these models under various noisy conditions, we show that the (learned) predictive confidence is well calibrated. We use these reliable confidence values for monitoring performance degradation and failure detection in DNNs. We propose two different failure detection methods. In the first method, we define a fixed threshold value based on the behavior of the predictive confidence with changing signal-to-noise ratio (SNR) of the test dataset. The second method learns the threshold value with a neural network. The proposed failure detection mechanisms seamlessly abstain from making decisions when the confidence of the BDNN is below the defined threshold and hold the decision for manual review. Resultantly, the accuracy of the models improves on the unseen test samples. We tested our proposed approach on three medical imaging datasets: PathMNIST, DermaMNIST, and OrganAMNIST, under different levels and types of noise. An increase in the noise of the test images increases the number of abstained samples. BDNNs are inherently robust and show more than 10% accuracy improvement with the proposed failure detection methods. The increased number of abstained samples or an abrupt increase in the predictive variance indicates model performance degradation or possible failure. Our work has the potential to improve the trustworthiness of DNNs and enhance user confidence in the model predictions.

Bilingual and monolingual children attend to different cues when learning new words.

The way in which children learn language can vary depending on their language environment. Previous work suggests that bilingual children may be more sensitive to pragmatic cues from a speaker when learning new words than monolingual children are. On the other hand, monolingual children may rely more heavily on object properties than bilingual children do. In this study we manipulate these two sources of information within the same paradigm, using eye gaze as a pragmatic cue and similarity along different dimensions as an object cue. In the crucial condition, object and pragmatic cues were inconsistent with each other. Our results showed that in this ambiguous condition monolingual children attend more to object property cues whereas bilingual children attend more to pragmatic cues. Control conditions showed that monolingual children were sensitive to eye gaze and bilingual children were sensitive to similarity by shape; it was only when the cues were inconsistent that children's preference for one or the other cue was apparent. Our results suggest that children learn to weigh different cues depending on their relative informativeness in their environment.