What the Tech? Matrix AI Server Medical Diagnosis Services
A detailed look at Matrix’s AI Server Medical Diagnosis Services.
For the past few months, the Matrix community has been able to interact and experiment with a couple of the Matrix AI Server’s basic features — namely, Matrix AI Pose Detection and Object Detection. Indeed, these two functions were the basis for a couple community competitions. While important building blocks in their own right, these two AI services barely scratch the surface of what is possible with the Matrix AI Server and its many AI services.
Artificial Intelligence Meets Medical Diagnosis
Like with many industries, artificial intelligence is poised to revolutionize the medical industry. Medical diagnosis, in particular, is primed for a revolution. When working in concert with established diagnostic approaches, AI has the potential to improve the accuracy of medical diagnoses, detect diseases and medical conditions at an earlier stage, and limit the variability in the quality of care a patient receives; just to name a few obvious advantages. According to Matrix Chief AI Scientist Professor Steve Deng, the “integration of blockchain and AI is bridging several emerging issues for AI in the medical field. There is the growing demand for parallel processing and quick turnaround of different kinds of AI assisted diagnoses.” To this end, the Matrix AI Network is currently working with several hospitals including the Shanghai Pulmonary Hospital (affiliated to Tongji University) and the Huadong Hospital (affiliated to Fudan University).
Active projects with these universities are primarily focused on small cell lung cancer and rib fracture detection. These projects rely on powerful AI models to analyze a litany of past cases, treatment guidelines and academic literature. The ultimate goal is to propose complete, appropriate treatment plans while simultaneously reducing the likelihood of missed- and misdiagnosis. This “leads to better care for patients, cost savings for insurers and suppliers, and larger data sets to improve AI-assisted diagnosis models,” adds Professor Deng.
The Small Cell Lung Cancer Project
Small cell lung cancer (SCLC) is a highly aggressive type of lung cancer. It accounts for approximately 15–20% of all lung cancers and has a short doubling time, high growth fractions and metastasizes very quickly. Matrix uses AI-enhanced imaging protocols to detect SCLCs. These protocols include data collection, data preprocessing, image segmentation, lung nodule marketing, model training and classification prediction.
The Matrix AI Server takes these high-resolution CT scans (20000 x 20000) and enhances them by analyzing macro- and micro-data. Doing so enables Matrix to identify SCLC cells, measure the tumor volume and complete an outline of the lesion.
The Rib Fracture Project
Diagnosing rib fractures is a surprising difficult endeavor. Mistakes and missteps are commonplace as doctors need to manually conduct a huge amount of diagnostic work. This is largely due to the fact that people have a large number of ribs (usually 12 pairs, though some people have 11 or 13 pairs!) and that ribs have a complicated spatial relationship to one another.
Matrix enables the diagnosis of fracture types, greatly improving the accuracy and efficiency of diagnosis and treatment. To this end, Matrix reconstructs the original CT scan and generates a 3D image. First, the patient’s spine and sternum are distinguished from the ribs using advanced morphology. Then, as each bone is segmented, the sternum and the spine are removed. Finally, any other bones are detected, and all ribs are uniquely identified. After identification, each rib is then individually matched to existing specialist-generated labels for the purpose of confirmation and fracture data is produced using VOC & COCO standards. Finally, the generated data is used to continuously train the AI models to identify an increasing number of fracture types.
One thing that differentiates this approach from other methods is that, in order to solve the problems of unclear bone segmentation and sparse bone corrosion, Matrix uses advanced morphology and incremental learning throughout the analytical process. Simultaneously, in order to improve the efficiency in identifying individual ribs, Matrix uses code acceleration, interpolation algorithms and opencv.
The Future of AI-Medical Diagnosis
Matrix believes that AI-assisted medical diagnosis will improve early-stage disease screening, help patients reduce medical costs and reduce the occurrences of misdiagnosis. Therefore, in addition to pose and object detection, Matrix is launching a suite of Medical Diagnosis services on the Matrix AI Server.
The Medical Diagnosis tools are accessible via the Matrix web wallet, under the AI Transaction tab.
To use the Matrix AI Server Medical Diagnosis services, simply upload the relevant files. In the case of SCLC, these should comprise an industry-standard set of medical images (dcm format). Once uploaded, click Generate Transaction.
A full report will be produced and returned in about 15 minutes. Like all blockchain transactions, the results can be queried using the relevant Txhash. The Matrix AI Server analyzes the CT scans and generates a full report (downloadable as a pdf!). In this way, Medical Diagnosis transactions work in the same way as the AI Server Pose and Object Detection transactions.
It should be noted that, currently, only the AI-enhanced imaging protocols to detect SCLCs is open to the public. The rib fracture protocol requires some more testing before the Matrix team releases it publicly.
Artificial Intelligence helps Matrix expand and deliver on the promise of blockchain. The team is very proud of their work on AI Medical Diagnosis services. Moving forward, Matrix — alongside its partners — will continue to train AI models using the Matrix AI Network’s excess computing power. In return, access to these AI models will be given to everyone. “The future of a secure, high-performance, and interoperable distributed computing infrastructure for AI is an important foundation for democratizing cost and access to AI modeling, effective public/private system integration, and helping AI scale and improve,” says Professor Deng. Matrix is democratizing artificial intelligence.