By now, it's almost old news that artificial intelligence (AI) will have a transformative role in medicine. Algorithms have the potential to work tirelessly, at faster rates and now with potentially greater accuracy than clinicians. In 2016, it was predicted that'machine learning will displace much of the work of radiologists and anatomical pathologists'. In the same year, a University of Toronto professor controversially announced that'we should stop training radiologists now'. But is it really the beginning of the end for some medical specialties?
The robot-assisted radical prostatectomy was segmented into 12 steps, and for each step, 41 validated automated performance metrics were reported. The predictive models were trained with three data sets: 1) 492 automated performance metrics; 2) 16 clinicopathological data (for example prostate volume, Gleason score); 3) automated performance metrics plus clinicopathological data. The authors utilized a random forest model (800 trees) to predict continence recovery (no pads or one safety pad) at three and six months after surgery. The prediction accuracy was estimated through a 10-fold cross-validation process. The area under the curve (AUC) and standard error (SE) was used to estimate prediction accuracy. Finally, the out-of-bag Gini index was used to rank the variables of importance.
Deep learning (DL) models are known for tackling the nonlinearities associated with data, which the traditional estimators such as logistic regression couldn't. However, there is still a cloud of doubt with regards to the increased use of computationally intensive DL for simple classification tasks. To find out if DL really outperforms shallow models significantly, the researchers from the University of Pennsylvania experiment with three ML pipelines that involve traditional methods, AutoML and DL in a paper titled, 'Is Deep Learning Necessary For Simple Classification Tasks.' The UPenn researchers stated that a support-vector machine (SVM) model might predict more accurately susceptibility to a certain complex genetic disease than a gradient boosting model trained on the same dataset. Moreover, choosing different hyperparameters within that SVM model can vary performances.
Small startups and big companies alike are recognizing that modern biotech R&D is as much a data ... [ ] problem as a science problem. Cloud technologies offer a way to bring together massive amounts of complex data to improve the way we feed, fuel, heal, and build our world with biology. These days, biotech R&D is as much a data problem as a science problem. Here's why: in the past decade, the exploding field of synthetic biology has done an incredible job solving the scientific challenges of making biology easier to engineer. I have written about how tools like gene editing, synthesis, sequencing, and automation are changing for the better the way we feed, fuel, heal, and build our world with biology.
The world of genomics has made abrupt strides in the past several years, with the first CRISPR-edited babies being born just a few weeks ago. Using advanced CRISPR technology, Scientist Jiankui He'announced that twin girls with an edited gene that reduces the risk of contracting HIV "came crying into this world as healthy as any other babies a few weeks ago."' The announcement was met with great backlash, sparking'outrage from many researchers and ethicists who say implanting edited embryos to create babies is premature and exposes the children to unnecessary health risks. Opponents also fear the creation of "designer babies," children edited to enhance their intelligence, athleticism or other traits.' CRISPR technology is used in editing human genomes.
A newly developed artificial intelligence (AI) system could help expedite the diagnosis of epileptic conditions such as Dravet syndrome. The AI system was described in a study, titled "A propositional AI system for supporting epilepsy diagnosis based on the 2017 epilepsy classification: Illustrated by Dravet syndrome," in the journal Epilepsy & Behavior. Epilepsy is a broad disease category for many different conditions that involve seizures. Properly diagnosing epileptic conditions can be a challenge, especially given their different causes and symptoms. For example, mutations in the SCN1A gene are the most common cause of Dravet syndrome, but not all people with Dravet syndrome have such mutations, and SCN1A mutations can also be associated with other conditions, such as febrile seizures plus.
Some -omics tools can be more accurate, sensitive or efficient than others. Yet benchmarking is no tell-all. Inflammatory bowel disease (IBD) is a complex genetic disease that is instigated and amplified by the confluence of multiple genetic and environmental variables that perturb the immune–microbiome axis. Here the authors describe IBD as a model disease in the context of leveraging human genetics to dissect interactions in cellular and molecular pathways that regulate homeostasis of the mucosal immune system. Machine learning can tell different types of knot apart just by'looking' at them.
We know OpenTable as the restaurant reservation system, but OpenTable has revolutionized the entire restaurant industry. By compiling a comprehensive database of dates, names, places, check size and so on, OpenTable creates operational advantages for its food and beverage customers. It provides the infrastructure to manage those reservations, assign tables, recognize repeat diners and remember diner preferences. It also allows restaurants to better manage costs by staffing correctly and minimizing food waste.
Deep neural networks (DNNs) show promise in breast cancer screening, but their robustness to input perturbations must be better understood before they can be clinically implemented. There exists extensive literature on this subject in the context of natural images that can potentially be built upon. However, it cannot be assumed that conclusions about robustness will transfer from natural images to mammogram images, due to significant differences between the two image modalities. In order to determine whether conclusions will transfer, we measure the sensitivity of a radiologist-level screening mammogram image classifier to four commonly studied input perturbations that natural image classifiers are sensitive to. We find that mammogram image classifiers are also sensitive to these perturbations, which suggests that we can build on the existing literature. We also perform a detailed analysis on the effects of low-pass filtering, and find that it degrades the visibility of clinically meaningful features called microcalcifications. Since low-pass filtering removes semantically meaningful information that is predictive of breast cancer, we argue that it is undesirable for mammogram image classifiers to be invariant to it. This is in contrast to natural images, where we do not want DNNs to be sensitive to low-pass filtering due to its tendency to remove information that is human-incomprehensible.
The system can be deployed on a smartphone, achieves 91 percent top-10-accuracy in identifying over 215 different genetic syndromes, and has outperformed clinical experts in three separate experiments. The FDNA team's research paper, Identifying facial phenotypes of genetic disorders using deep learning, has been published in Nature Medicine. Using deep learning algorithms and brain-like neural networks, the Face3Gene app can predict congenital and neural developmental disorders in people through the detection of distinctive facial features in photos. Face2Gene builds on a technique the FDNA team introduced last January in the paper DeepGestalt -- Identifying Rare Genetic Syndromes Using Deep Learning. Researchers started by training an AI system to distinguish two conditions which cause distinct facial features -- Cornelia de Langs syndrome and Angelman syndrome -- from other, similar conditions.