big data artificial intelligence

The healthcare sector is transforming through the internet of medical things, the convergence of technology, digitalisation, and 3-D modelling.


Healthcare technological revolution

In the past decade, new trends in diagnostic radiology through artificial intelligence, machine learning, virtual reality, wearable medical devices, and 3-D modelling have brought about a revolution in the healthcare sector.  Today, these technological platforms will have a more significant impact on the management of the patient.

Future healthcare innovations to medical devices will be more modular and flexible to improve the patient experience.  For example, 3-D printing and virtual simulations create new opportunities to improve quality and safety, especially optimising. However, 3-D printing can also offer faster prototypes and the creation of personalised prosthetics. Furthermore, 3-D printing is being used to adjust tablet porosity to produce and personalised medicines at the point of care.

The COVID-19 pandemic has accelerated virtual healthcare technology due to consumers and providers finding ways to deliver aspects of healthcare.  These digital interface systems allow the patient to connect remotely with healthcare professionals using video conferencing or mobile apps. Therefore, future general practices will decline due to the expansion and production of innovative medical devices.

Big Data is categorised in three ways

Structured Data: This refers to ordered data already stored in the traditional matrix databases and accounts for about 20% of the total data.  The data is useful in programming and computer-related applications.

Unstructured Data:  This type of data has no clear format in storage and accounts for about 80% of the unstructured data.  This data is more assessable and is just stored and analysed manually.

Semi-Structured Data: The type of unstructured data and semi-structured data is sometimes not apparent to the user because most semi-structured data appear to be unstructured when first encountered. For example, NoSQL documents are considered semi-structured because they contain keywords used to process the document.

Difference between Structured, Semi-structured and Unstructured data

ParameterStructured Data
Semi-structured DataUnstructured Data
Organisation
Structured data is organised with the highest level of organisation
Semi-structured data is partially organised; hence the level of organisation is less than
structured data but higher than unstructured data
Unstructured data is unorganised
Flexibility
Structured data is part of an interactive database. It is schema dependent resulting in a less flexible database and challenging to scale
Semi-structured data is more flexible and less complicated to scale than structured data
Unstructured data does not have a schema. Therefore, the data becomes more flexible
Versioning
Structured data is based on an interactive database. Versioning is performed over tuples, rows and tables
Semi-structured data, tuples or graphs are possible as only a partial database is supported
Unstructured data, versioning is likely as whole data with no database support
Transaction Management
Structured data, data concurrency is available and therefore, usually preferred for the multitasking process
Semi-structured data transactions get adapted from Database Management System, but still, data concurrency is not available
Unstructured data, neither transaction management nor data concurrency, is present

Internet of Medical Things

The healthcare sector is transforming through the convergence of technology, digitalisation, and 3-D modelling.  These changes have created technological development through innovation and market expansion by the robotic revolution.  Robotic process automation can carry out basic and repetitive tasks in the healthcare setting while allowing healthcare professionals to concentrate more on high-value projects.

Big Data is creating a digital revolution that needs more analytics, equating to 2.5 quintillion bytes of data every day.  Big Data is becoming part of the everyday industry today, which is the driving force behind the worldwide success of organisations.

The internet of medical things (IoMT) refers to the network that connects intelligent medical devices via the internet.

The next healthcare interaction will involve a medical device, for example, a blood pressure monitor, a continuous glucose monitor or a medical scanner.  Today there are over 500,000 available medical technologies. These medical devices provide internet-connected services that can improve efficiencies and improve patient treatment plans.  Furthermore, increased computer power and wireless capabilities will force healthcare organisations towards more IoMTs. 

IoMTs devices will have the ability to accumulate, analyse, and transmit healthcare data, especially for clinicians to evaluate the patient’s chronic illnesses and evolve the future of care.

Big Data, Artificial Intelligence and Algorithms all play a role in medical diagnostics towards personalised medicine. Big Data brings with it a vast amount of data generated from several sources. Also, Big Data needs to be automated, stored in the correct category to find correlations, hidden patterns, and other valuable insights.  The categorisation of mixed heterogeneous data is known as data classification based on predefined features.   

However, in the last decade, the healthcare sector has expanded and generates enormous amounts of data in terms of volume, velocity, variety, and veracity.  All these Big Data practices in healthcare can increase the business value and improve healthcare services.

Big Data is highly complex datasets characterised by the ‘V’ attributes

Big Data four Vs – Volume, Velocity, Variety and Veracity

Volume
The datasets generated by radiology procedures are high in volume due to the pixel image size from computed tomography, magnetic resonance imaging, computed tomography angiography, x-rays, PET and SPECT imaging and mammography.
Velocity
Data is processed at speed. Radiology procedures produce vast amounts of data at high speed. MRI and CT scanners provide continuous datasets into the PACS networks, and the images are then stored using the Vendor Neutral Archives (VNA). All the data is generated in real-time.
Variety
The images produced by the radiology data is generated from a range of modalities, for example, computed radiography, conventional radiography, interventional radiology, digital radiography, PET and SPECT imaging, MRI, and ultrasound.
Veracity
The integrity of the dataset is paramount in any project. However, a systematic analysis of data is required to make sure the input datasets are accurate results. Therefore, scans with motion artefacts and low quality can be eliminated from the study group. In addition, the checks maintain the uniformity of datasets.

Big Data analytics

Big Data analytics manage and analyse massive data volumes.  Another way to characterise Big Data more effectively is to apply the HACE (Huge, Autonomous, Complex, Evolving) approach.

Big data utilises large and heterogeneous data volume. This approach includes independent sources to enable distribution and reorganised controls to explore complex and evolving relationships with the data.

Big Data incorporates parametric and non-parametric measures, such as diagnosis, demographics, treatment, and disease prevention. All these attributes stem from a variety of sources by applying incongruent sampling.

This approach produces structured data focusing on genotype, proteomic or clinical scores compared to unstructured data, including clinical notes, medical imaging,  prescriptions, lifestyle, environmental and health economics data.

Healthcare professionals should strive to make progress toward harnessing Big Data in imaging, which has the potential to lead to advanced clinical support, personalised diagnostic and prognostic tools, and the ability to optimise individual patient outcomes in ways that were previously not possible.

Big Data analytics implies the evaluation of large, the identification of clusters and the correlation between datasets leading to the development of predictive models using data mining techniques.

Conclusion

The diagnostic imaging sector has undergone significant growth both in terms of technological development and market expansion.  This growth leads to the increased production of a vast amount of data that puts diagnostic imaging in Big Data storage in the context of healthcare.  Consequently, it is necessary to build digital platforms and medical devices that facilitate diagnostic images using Big Data.

Big Data users need new advancing procedures that incorporate cloud computing, teleradiology from a site far from the acquisition scanner or finalised second look activities to check the quality and advice for specialist reports.

All these technological procedures require the transmission of thousands of images to other locations. Moreover, at the terminals, it must facilitate download/upload velocity, data integrity and security.  Also, the handling of data must comply with privacy laws.

Furthermore, applying various algorithmic tools and converting raw data to large datasets will enable a better understanding of the radiological data to gain new knowledge and insights into a medical problem.

By Open Medscience

Open Medscience is a platform to discuss a range of imaging modalities including radiology, ultrasound, computed tomography, MRI, nuclear medicine (PET & SPECT) and radiation therapy.