5 Algorithms That Are Transforming The Healthcare Industry

by November 7, 2016 0 comments

By Vivek Prakash, CTO & Co-founder, HackerEarth

Algorithms have revolutionized our world as we know it.

Today, a lot of things come to us a lot more easily than before. We have moved forward from the age of waiting to hear our favorite song on the FM radio and recording it on a cassette. Now, we visit YouTube to play your favorite song. It miraculously recommends all your favorite songs in the auto playlist. When you visit IMDB’s website, it suggests the most relevant movies every time. When you Google something, you see the most relevant information on the first page.

vivek-1

Vivek Prakash, CTO & Co-founder, HackerEarth

Have you ever wondered how each website knows exactly what to show you?Algorithms

This post focuses on the impact that algorithms have in the field of medicines where you must be a 100% right always. There is no room for errors because even the trivial errors can create a major impact. However, even the smartest and best-trained professionals are prone to errors. Tragedies due to human error are common in the medical industry.

Today, by using algorithms, doctors and care providers can determine exactly where to point the lasers for maximum impact with minimum collateral damage. Algorithms have made the way we treat patients more effective.

Algorithms play a major role in medicines, from large medical equipment to simple microcontrollers. Let’s look at the top algorithms that are used in the healthcare industry.

untitled-1

Sampling

The healthcare industry generates large amounts of data, which must be mined and sorted. Some facts about the industry include:

  •         Every year almost a million medical studies are published.
  •         Additionally, 150,000 cancer-related studies are published annually.

The human brain is brilliant but there is a limit to the amount of information that it can process and store. While computers may not be able to outdo doctors, they can be used to increase the number of lives that are saved.

untitled-3

Sampling is a method of studying few selected items, instead of a big number of units. This small selected item is called sample, and many units is termed as population. For example, at a fruit shop you check one or two apples to decide the concentration of good apples. Thus, with the sample, we infer about the population. With such a huge density of data available in the medical field, we use various sampling techniques: –

  • Simple random sampling This is a basic technique where we select a sample for studying a larger population. Everyone or sample is chosen randomly and every member of the population has an equal chance of being selected.
  • Systematic sampling: This is a technique where sample members from a larger group are selected based, on a random starting point and a fixed periodic interval. The interval is known as a sampling interval. It is calculated by dividing the population size by using the desired sample.
  • Stratified sampling: The researcher divides the population into separate groups, called strata. Then, a probability sample (often simple random sample) is drawn from each group.
  • Clustering sampling: is used when many ‘natural’ but heterogeneous groups are present in the statistical population. These groups (or population) are divided into small classes and any one of the sampling methods is applied to them (preferably simple random).

Using sampling technique, computers provide information about individuals leading to concrete results without examining them.

Fourier Transform in MRI

The Fourier transform is often called one of the most important algorithms of our time. This algorithm applies to almost all aspects of our everyday life.

A Fourier transform is a mathematical transformation applied to transform a signal or image from the time domain to frequency domain which has many applications. A Fourier image converts a real value to complex value. And then for use in a different domain, the information is converted from complex value to a real value.

Magnetic Resonance Imaging (MRI) uses the concept that the human body is made up of 70% of water. It is due to these water molecules that our body consists of many micro magnets.These magnets all align themselves in one direction when an external magnetic field is applied— an MRI machine in this case.

Now since the human body is 3D, so the X & Y gradients are also determined for an image.The signals that we measure in an MRI is a combination of the signals that are captured while the human body is being traced. A signal is composed of a series of sine waves where each sine wave has an individual frequency and amplitude.

The Fourier transform allows us to work out, what that frequency and amplitude are. Since the signal is encoded with magnetic-field, ingredients which make the frequency and phase relate to the position of the object. Once the frequencies are separated, then the amplitude of the image can be plotted.

Without the Fourier Transform, medical imaging would not have been possible. Ultrasounds, MRIs, and other medical-imaging techniques use the Fourier Transform algorithm to convert the images that are then converted into a readable format.

Probabilistic data-matching

If you are a doctor who is treating Vito Corleone, you might consider the electronic records that are related to Vito Corleone’s medical history (You do know that he is a Don, right?).So how does probabilistic data matching come into the picture?

Probabilistic data will be used to look for all the possible information regarding all available medical data. It will sort the data by giving preference to that data which has a likelihood of matching with Don’s medical data.

Probabilistic matching uses a likelihood ratio theory to assign comparison outcomes to the correct or “more likely” decision.

A patient whose symptoms are like a specific disease may have the relevant data analyzed against existing information. Based on the available matches, a physician will be able to determine which disease is the best possible match.

This allows the physician to create an accurate treatment plan thus giving the patient, a better chance of receiving the right treatment on time.Probabilistic algorithms such as Naive Bayes Classifier and PAIRS (Physician assistant Artificial Intelligence System) are being used for efficient inference in large models to provide additional evidence for research and medical cases.

Proportional Integral Derivative (PID)

In the Cardiac Unit of Alabama Hospital, the Mean Arterial Pressure of a patient is managed by a computer. This computer controls the infusion of vasodilating agents and it has helped around 1100 hypertensive patients after heart surgery.

This computer uses the digital version of a PID controller algorithm to perform the intensive task. PID is a control feedback mechanism, which controls the computer that eventually calculates an error value as the difference between the desired set point and the measured values.In simple terms, this algorithm reduces the difference between the desired output and the expected output.

The P in the symbol accounts for the present value of the error, I for the previous value, and D is the possible future value of the error. What the controller does is it tries to reduce the error by manipulating the various factors that are associated with the mechanism. Based on the combination of P, I, and D the best result with the least errors is processed. Probably every electronic circuit that has a mechanical or thermal system attached to it will use a PID circuit.

Predictive Algorithm

Some predictive medical algorithms claim that they can use real-time data from an ICU to predict events like cardiac arrests 24 hours before they happen. It is difficult for the human mind to memorize all the information and data that it has learned over a period. The predictive algorithm matches the data and information over a period. It also analyzes and judges the data over the current information, and then predicts the outcome based on the analysis.

Various predictive algorithms include:

  • Time Series algorithm
  • Regressions algorithm
  • Association algorithm
  • Clustering algorithm
  • Decision Tree algorithm

With all the changes coming in the field of medicines and healthcare worldwide, algorithms are the next step. In a few years, physicians will be consultants and algorithms will work together to create a greater impact on human lives.

No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.

Your data will be safe!Your e-mail address will not be published. Also other data will not be shared with third person.