Learning Individual Survival Models from PanCancer Whole Transcriptome Data.

PURPOSE: Personalized medicine attempts to predict survival time for each patient, based on their individual tumor molecular profile. We investigate whether our survival learner in combination with a dimension reduction method can produce useful survival estimates for a variety of patients with cancer. EXPERIMENTAL DESIGN: This article provides a method that learns a model for predicting the survival time for individual patients with cancer from the PanCancer Atlas: given the (16,335 dimensional) gene expression profiles from 10,173 patients, each having one of 33 cancers, this method uses unsupervised nonnegative matrix factorization (NMF) to reexpress the gene expression data for each patient in terms of 100 learned NMF factors. It then feeds these 100 factors into the Multi-Task Logistic Regression (MTLR) learner to produce cancer-specific models for each of 20 cancers (with >50 uncensored instances); this produces "individual survival distributions" (ISD), which provide survival probabilities at each future time for each individual patient, which provides a patient's risk score and estimated survival time. RESULTS: Our NMF-MTLR concordance indices outperformed the VAECox benchmark by 14.9% overall. We achieved optimal survival prediction using pan-cancer NMF in combination with cancer-specific MTLR models. We provide biological interpretation of the NMF model and clinical implications of ISDs for prognosis and therapeutic response prediction. CONCLUSIONS: NMF-MTLR provides many benefits over other models: superior model discrimination, superior calibration, meaningful survival time estimates, and accurate probabilistic estimates of survival over time for each individual patient. We advocate for the adoption of these cancer survival models in clinical and research settings.

Health Care Students' Perspectives on Artificial Intelligence: Countrywide Survey in Canada.

BACKGROUND: Artificial intelligence (AI) is no longer a futuristic concept; it is increasingly being integrated into health care. As studies on attitudes toward AI have primarily focused on physicians, there is a need to assess the perspectives of students across health care disciplines to inform future curriculum development. OBJECTIVE: This study aims to explore and identify gaps in the knowledge that Canadian health care students have regarding AI, capture how health care students in different fields differ in their knowledge and perspectives on AI, and present student-identified ways that AI literacy may be incorporated into the health care curriculum. METHODS: The survey was developed from a narrative literature review of topics in attitudinal surveys on AI. The final survey comprised 15 items, including multiple-choice questions, pick-group-rank questions, 11-point Likert scale items, slider scale questions, and narrative questions. We used snowball and convenience sampling methods by distributing an email with a description and a link to the web-based survey to representatives from 18 Canadian schools. RESULTS: A total of 2167 students across 10 different health professions from 18 universities across Canada responded to the survey. Overall, 78.77% (1707/2167) predicted that AI technology would affect their careers within the coming decade and 74.5% (1595/2167) reported a positive outlook toward the emerging role of AI in their respective fields. Attitudes toward AI varied by discipline. Students, even those opposed to AI, identified the need to incorporate a basic understanding of AI into their curricula. CONCLUSIONS: We performed a nationwide survey of health care students across 10 different health professions in Canada. The findings would inform student-identified topics within AI and their preferred delivery formats, which would advance education across different health care professions.

Reappraising the Luminescence Lifetime Distributions in Silicon Nanocrystals.

The luminescence dynamics in ensembles of nanocrystals are complicated by a variety of processes, including the size-dependence of the radiative and non-radiative rates in inhomogeneous broadened samples and interparticle interactions. This results in a non-exponential decay, which for the specific case of silicon nanocrystals (SiNCs) has been widely modeled with a Kohlrausch or "stretched exponential" (SE) function. We first derive the population decay function for a luminescence decay following exp[- (t/τ)ß]. We then compare the distributions and mean times calculated by assuming that either the luminescence decay or the population decay follows this function and show that the results are significantly different for ß much below 1. We then apply these two types of SE functions as well as other models to the luminescence decay data from two thermally grown SiNC samples with different mean sizes. The mean lifetimes are strongly dependent on the experimental setup and the chosen fitting model, none of which appears to adequately describe the ensemble decay dynamics. Frequency-resolved spectroscopy (FRS) techniques are then applied to SiNCs in order to extract the lifetime distribution directly. The rate distribution has a half width of ~ 0.5 decades and mainly resembles a somewhat high-frequency-skewed lognormal function. The combination of TRS and FRS methods appear best suited to uncovering the luminescence dynamics of NC materials having a broad emission spectrum.