Google AI predicts hospital inpatient death risks with 95% accuracy

#artificialintelligence

Using raw data from the entirety of a patient's electronic health record, Google researchers have developed an artificial intelligence network capable of predicting the course of their disease and risk of death during a hospital stay, with much more accuracy than previous methods. The deep learning models were trained on over 216,000 deidentified EHRs from more than 114,000 adult patients, who had been hospitalized for at least one day at either the University of California, San Francisco or the University of Chicago. For those two academic medical centers, the AI predicted the risks of mortality, readmission and prolonged stays, as well as discharge diagnoses, by ICD-9 code. The network was 95% accurate in predicting a patient's risk of dying while in the hospital--with a much lower rate of false alerts--than the traditional regressive model--the augmented Early Warning Score--which measures 28 factors and was about 85% accurate at the two centers. The researchers' findings were published last month in the Nature journal npj Digital Medicine.


Medical Concept Embedding with Time-Aware Attention

arXiv.org Artificial Intelligence

Embeddings of medical concepts such as medication, procedure and diagnosis codes in Electronic Medical Records (EMRs) are central to healthcare analytics. Previous work on medical concept embedding takes medical concepts and EMRs as words and documents respectively. Nevertheless, such models miss out the temporal nature of EMR data. On the one hand, two consecutive medical concepts do not indicate they are temporally close, but the correlations between them can be revealed by the time gap. On the other hand, the temporal scopes of medical concepts often vary greatly (e.g., \textit{common cold} and \textit{diabetes}). In this paper, we propose to incorporate the temporal information to embed medical codes. Based on the Continuous Bag-of-Words model, we employ the attention mechanism to learn a "soft" time-aware context window for each medical concept. Experiments on public and proprietary datasets through clustering and nearest neighbour search tasks demonstrate the effectiveness of our model, showing that it outperforms five state-of-the-art baselines.


Google: AI death predictor predicts life expectancy with 95 per cent accuracy

#artificialintelligence

With the enough data, the company thinks it can predict when a patient will die with up to 95 per cent accuracy. In May, Google scientists published the account of a woman who came to hospital with late stage breast cancer and fluid building in her lungs. After the hospital equipment and computers took the woman's vital signs, it estimated that she had a 9.3 per cent chance of dying during her stay at the hospital. Then it was Google's turn. Its neural network, a type of artificial intelligence that can analyse huge reams of data and automatically learn and improve, was fed 175,639 data points on the woman including past health records and her current vital signs.


Patient trajectory prediction in the Mimic-III dataset, challenges and pitfalls

arXiv.org Machine Learning

Automated medical prognosis has gained interest as artificial intelligence evolves and the potential for computer-aided medicine becomes evident. Nevertheless, it is challenging to design an effective system that, given a patient's medical history, is able to predict probable future conditions. Previous works, mostly carried out over private datasets, have tackled the problem by using artificial neural network architectures that cannot deal with low-cardinality datasets, or by means of non-generalizable inference approaches. We introduce a Deep Learning architecture whose design results from an intensive experimental process. The final architecture is based on two parallel Minimal Gated Recurrent Unit networks working in bi-directional manner, which was extensively tested with the open-access Mimic-III dataset. Our results demonstrate significant improvements of automated medical prognosis, as measured with Recall@k. We summarize our experience as a set of relevant insights for the design of Deep Learning architectures. Our work improves the performance of computer-aided medicine and can serve as a guide in designing artificial neural networks used in prediction tasks.


Few-shot Learning for Named Entity Recognition in Medical Text

arXiv.org Machine Learning

Deep neural network models have recently achieved state-of-the-art performance gains in a variety of natural language processing (NLP) tasks (Young, Hazarika, Poria, & Cambria, 2017). However, these gains rely on the availability of large amounts of annotated examples, without which state-of-the-art performance is rarely achievable. This is especially inconvenient for the many NLP fields where annotated examples are scarce, such as medical text. To improve NLP models in this situation, we evaluate five improvements on named entity recognition (NER) tasks when only ten annotated examples are available: (1) layer-wise initialization with pre-trained weights, (2) hyperparameter tuning, (3) combining pre-training data, (4) custom word embeddings, and (5) optimizing out-of-vocabulary (OOV) words. Experimental results show that the F1 score of 69.3% achievable by state-of-the-art models can be improved to 78.87%.