People will never trust an AI over a human doctor. Not if the AI is going to give a more accurate, reliable and efficient answer. But VoxelCloud is not going to substitute a doctor, the company just want to assistant doctors to handle the influx of medical images.
April 2017, Watson Health, a spinoff of IBM, introduced its first cognitive solution IBM Watson Imaging Clinical Review. The solution provides impactful imaging solutions that address complex medical problems and is generally available to healthcare providers. Technically, it is a cognitive data review tool which supports accurate and timely clinical and administrative decision-making.
A startup called VoxelCloud（体素科技） is joining the business by teaming up with hospital systems and clinical practitioners, providing them automated medical image analysis services, as well as clinical decision support services.
VoxelCloud was founded by one founder and two co-founders in early 2016, all are experts and scholars and deep related. DING Xiaowei（丁晓伟）is the founder and chief executive officer, a research assistant professor in the UCLA Computer Science Department. Demetri Terzopoulos, Chief Scientific Officer, was DING’s Ph.D. professor in UCLA, and Jianming Liang（梁建明） is Vice President of R&D Department, who is also a student of Pro. Terzopoulos.
Pro. Terzopoulos is Chancellor's Professor of Computer Science at UCLA and One of the most highly cited authors in engineering and computer science. 80% of VoxelCloud staff are technology workers. They know their stuff. But to do a great business, you always need to address two things, demand and supply.
Demand: the Medical Imaging Market
According to Mordor Intelligence, The global artificial intelligence in medicine market was valued at USD 1.603 billion in 2017 and is expected to reach a market value of USD 12.150 billion by 2023, at a CAGR of 40.15% (2018-2023). Moreover, the inspection of images and diagnostics for various tests, such as CT(computed tomography) scans, MRI(Magnetic resonance imaging), could also be delegated to AI systems. It has been identified that error rates for image labeling have decreased from 28.5% to 2.5%, which emphasizes the need for AI systems in medicine.
AI is not just helping, actually, it can do things that practitioners can't even come close to doing. Human bodies can be diagnosed in 12,000 different aspects or possibilities, and one type of disease founded in a single organ can have hundreds of subtypes, which cause a high misdiagnosis rate, even in America where practitioners are not overloaded compared to China.
Next is medical data explosion. IDC Health Insights indicates that health care data is expected to skyrocket from about 500 PB(1 PB=1024TB) today to more than 25,000 PB by 2020 a 50-fold increase.
That's why the digitization of data over the next decade has enormous implications for health care and research, with cloud computing and the aggregation of big data sets on the brink of enabling breakthroughs in personalized medicine according to Greg Slabodkin. No human could ever be expected to handle such a huge amount of data, or in our case, medical images, without getting mentally exhausted over time or making any mistakes.
Under this circumstances, AI could leverage an enormous impact in medical imaging industry. Imagine this: every X-ray machine in hospitals is connected to the cloud and one human doctor can get the output from the AI algorithm in natural language and without any human interventions after the patients get scanned.
Supply: the Business Model of VoxelCloud
People will never trust an AI over a human doctor. Not if the AI is going to give a more accurate, reliable and efficient answer. But VoxelCloud is not going to substitute a doctor, the company just want to assistant doctors to handle the influx of medical images.
To understand and break down the work process of medical imaging analysis, the team work closely with clinical practitioners and leading institutes such as UCLA, National Institutes of Health (NIH) in U.S and Peking Union Medical College Hospital and Zhong Shang Hospital in China.
VoxelCloud also has a distinct business model mainly in three channels. First, VoxelCloud provides service directly to healthcare providers like hospitals and medical centers through clinical workflow integrated cloud-computing solutions. Second, by partnering with established vendors to provide AI as a service. Third, by building a platform to enable third-party developers to develop their own applications through the VoxelCloud medical knowledge graph API.
Further extension of their project will leverage heterogeneous data sources in synergy with the existing imaging data, according to DING Xiaowei, reported by LA Tech Watch.
The technology products to date by VoxelCloud are three types, as well. Instead of specializing in one particular area, the company wants to be generalists across multiple areas, from lung disease to retina and coronary disease.
Lung Cancer Screening is a CT based early lung cancer screening system which offers deep-learning based lung nodule detection and quantitative feature extraction algorithms, etc. Retina Disease Screening is an AI-aided retinal image management and interpretation platform for diabetic retinopathy screening and beyond. Coronary Plaque Analysis is an intelligent, quantitative coronary plaque analysis tool for coronary artery disease risk assessment. It has been used in more than 75 single and multi-center clinical studies around the world.
In conclusion, VoxelCloud does meet the market demand with their supply. VoxelCloud is balanced with research and commercial applications, which gives the company a unique advantage in the market.
——Author: LinYan.Write to LinYan at LinYan@EqualOcean.com