>Business >ML in Healthcare – Twelve real-world use cases Part2

ML in Healthcare – Twelve real-world use cases Part2

Machine Learning, or ML for short, is a variant of artificial intelligence where computers are designed to go about learning data with no human action or intervention. Within machine learning, the production of the underlying ML algorithms is dependent on computational stats. Computers are furnished data and then the computers go about learning from that information. The data literally “trains” the computer through revelation of its advanced and complex patterns and the ML algorithms that underlie it. The bigger the data sample size the machine has furnished, the more accurate the machine’s output becomes.

ML in healthcare is witnessing increased penetration and is assisting patients and practitioners alike in a plethora of innovative ways. The most typical healthcare use cases with regards to ML are automation of medical billing, clinical decision support, and the generation of clinical care best-practices and guidelines. There are several noteworthy instances of ML and healthcare ideas that are beyond deployed within the healthcare and medical space. Meanwhile, at MD Anderson researchers have produced the first ML algorithm to forecast acute toxicities in patients receiving radiation therapy, for cancers in relation to the head and neck areas.

Within the domain of radiology, deep learning within healthcare detects advanced patterns on an automatic basis, and assists radiologists in engaging in smarter decision making during review of imagery like traditional radiographs, MRI, CT, PET images and radiology exports. The performance levels of learning-based automatic detection and diagnosis systems has demonstrated its equivalence to the out of a seasoned radiologist. Google’s ML applications within the healthcare space underwent training to identify and spot breast cancer and accomplished a remarkable 89% precision, on par, or at times improved, in comparison to radiologists. These are only some of the instances of the several uses of ML within healthcare.

Unstructured healthcare data for Machine Learning signifies approximately 80% of the data contained or “locked” in EHR systems. These are not considered data elements but as documents or textual files which historically could not have analysis undertaken with no human reading through the material. Human language, or “natural language” as it is often called, is very advanced and complicated, it lacks any kind of uniformity and integrates a massive amount of ambiguity, regional jargon, and vagueness. In order to translate these docs into more beneficial and interpretable data, ML in healthcare is usually dependent on natural language processing (NLP) programs. A majority of deep learning within healthcare applications that leverage NLP need some type of medical machine learning.

Healthcare use cases in machine learning are several. For instance, the NLP tech that is harnessed to decide creditworthiness for a client of sentiment analysis of an individual’s social media uploads can currently be harnessed to read a patient’s chart to gather critical data elements such as the patient’s medications, treatment schedules, and medical data.

Here are a few tasks that ML in healthcare can do:

ML strategies can be applied to identify solutions to a plethora of tasks. With regards to applications of ML in healthcare, these tasks are as follows:

 

Some instances of ML in Healthcare

Hurdles encountered with regards to ML in Healthcare

  1. Lacking quality data to develop accurate ML algorithms

The outcomes you obtain from ML algorithms are reliant on the quality of information placed into them. Unluckily, medical information is not always as accurate and standardized as it is usually required to be. There are gaps in records, imprecision in profiles, and other complications.

In total, EHR were not developed to be leveraged as a data source for an algorithm. So, prior to application of a machine learning tool, you’d be required to spend time collecting, cleansing, authenticating, and structuring your EHR data for its purposes.

  1. Developing ML Tools Conducive to Medical Workflow

There are several very particular ML use cases that can assist with patient diagnostics and treatment. Even if an ML utility functions well on paper, it does not necessarily imply that it will be taken up by physicians. As a result, it’s important to release ML tools that would be accessible and simple to harness in the day-to-day medical workflow. Without the required feedback from individuals who will work with the utility, it will not be as efficient, and the pros will not trust it.

  1. Gather Big Teams with Broader Skill Sets in One Place

Regardless of hands-on healthcare specialists, an efficient ML development team should consist of such roles:

  • Business analyst
  • Data scientist
  • Data engineer
  • Data architect
  • Machine learning expert

Aside from this, it’s critical to enable effective cooperation proceedings in the team so that it’s feasible to provide additional value and validate the product’s viability at the earlier possible time.

Conclusion

In the second part of this blog post, we looked at a few tasks that machine learning deployments are capable of executing. In addition, we were introduced to instances and use-cases of ML in healthcare, which will be further explored in a deep dive on the third part of this blog series. Finally, we also observed some of the challenges encountered by Machine Learning deployments within healthcare.

Add Comment