Big changes in AI and medical treatment in 2018Posted by Cathy Miller on January 6th, 2020 With the increasing application of AI in medical treatment, a few major breakthroughs have been made in 2008.
A research team of leukemia experts from a number of scientific institutions around the world uses blood tests and machine learning techniques to predict whether a healthy individual is at risk for acute myeloid leukemia (AML). The study was published in Nature. It is expected that leukemia will be predicted up to five years in advance. Acute myeloid leukemia (AML) is a rapidly progressing, life-threatening blood tumor that affects people of all ages. This study means that we can detect and monitor high-risk populations of AML early, so as to look for ways to reduce the risk of the disease.
A number of digital medical products were approved in 2018, and some of them are of great significance. In April last year, the US FDA approved the first medical device IDx-DR to use AI to detect retinopathy of diabetic patients. Developed by IDx, IDx-DR is the first medical device to receive marketing authorization to provide screening decisions without the need for clinicians to interpret images or results, making it available to physicians who are not normally involved in eye care. In August last year, the US FDA approved Aidoc's AI-based workflow optimization portfolio, which works with radiologists to mark acute intracranial hemorrhage in head CT images. This is also the first FDA-approved product that uses deep learning techniques to assist radiologists in their work.
If you talk about the breakthrough of AI in the field of medical health, you can't help but mention Google and DeepMind. In the year of 2018, the AI developed by Google and DeepMind achieved a number of achievements. First, in February last year, the Google Brain team discovered that the retina images of the human eye can very accurately predict risk factors affecting cardiovascular health through AI technology, including age, gender, smoking status, systolic blood pressure, and adverse cardiovascular events. In September, researchers from New York University School of Medicine developed a new machine learning program using Google's deep convolutional neural network Inception v3, which can not only identify patients' lung cancer types with 97% accuracy, but also recognizes these variant genes that cause abnormal cell growth. DeepMind, as a company focused on AI research by Google, has made major breakthroughs in this area. In August last year, DeepMind published a research report in Nature Medicine. The system developed by it can identify about 50 kinds of eye diseases according to optical coherence tomography (OCT) data. More importantly, this study also solved the problem of AI black box. In the future, this result will surely enable patients to get better treatment, whether it is cancer, neurological diseases or vision problems. In addition, DeepMind's biosphere "AlphaGo", which was launched in early December last year, is even more impressive: the new system called AlphaFold is capable of predicting and generating 3D structures of proteins, and it defeated all human participants at the Structure Prediction Competition (CASP). In the competition, AlphaFold topped the list of 98 contestants, predicting the 25 most accurate structures of the 43 proteins, while the second team in the same category predicted only 3 of the 43 species.
In November, scientists from the Ecole polytechnique fédérale de Lausanne (EPFL) in Switzerland set up a machine learning program called ShiftML to predict the movement of atoms in a molecule in a magnetic field. Even for relatively simple molecules, ShiftML is nearly 10,000 times faster than existing methods. This research shows that AI can help chemists break the molecular structure of crystals in a faster way than traditional modeling methods.
In March last year, Professor Mark Waller from Shanghai University applied a deep neural network and AI algorithms to successfully plan a new chemical synthesis route. The results show that this new algorithm can predict 80% of the molecular synthesis route in the test set with a single molecule time limit of 5 seconds. When the single molecule was extended to 60 seconds, the new algorithm predicted a molecular synthesis route ratio of 92%. This result is nearly 30 times faster than traditional computer-aided synthetic route design. Even an authoritative synthetic chemist can't tell the difference between this software and human chemists. This is a major breakthrough in the field of chemical synthesis of AI. Professor Mark Waller has also been hailed as a pioneer of “Chemical AlphaGo” by media.
The gene editing tool CRISPR-Cas9, which has appeared in recent years, provides researchers with the ability to edit the genome at a fixed point. CRISPR can find precise location in the genome to induce DNA double-strand breaks based on user-designed guide RNA (gRNA). It is more convenient and cheaper than previous genetic editing tools and is therefore widely used. In November last year, researchers at the Massachusetts Institute of Technology and Harvard University developed a computational model that can improve the performance of CRISPR-Cas9 through machine learning and achieve accurate and predictable editing of disease-causing gene mutations. This achievement offers new possibilities for genetic disease research and potential therapy. About the author As a peptide supplier and related service provider, Creative Peptides keeps close watch on the recent advance of AI in medical field. To facilitate scientists’ research, it offers a comprehensive range of GMP grade peptide synthesis and peptide analysis services tailored to clients’ needs such as custom conjugation, Identification of protein gel sample, peptide drug bioconjugation, Peptide Mass Fingerprinting, Cross linkers, SAR analysis, epitope mapping, etc. Staffed with strong research teams, Creative Peptides could be your exclusive choice in solving scientific and clinical challenges. Like it? Share it! More by this author |