Using AI to Predict Complications Before and After Surgery

by Madhu Reddiboina | Aug 26, 2021 | Health Care

Using AI to Predict Complications Before and After Surgery

In most cases of localized kidney cancer, surgery is the first-line treatment. Kidney tumors can be removed using bigger open incisions or by less invasive techniques such as laparoscopic or robotic surgery. These minimally invasive methods provide the same benefits of control as open surgery while allowing for an easier recovery.

Although surgery has grown in popularity among surgeons, it is associated with a complication risk of up to 30%. This is also true for robot-assisted and minimally invasive surgeries such as Robotic Partial Nephrectomy. The morbidity of partial nephrectomy is caused by a mix of tumor complexity, patient-related complexity, tumor surroundings, and surgeon-related complications.

However, when accurate Machine Learning (ML) models are used in clinical practice, they can play a key role in predicting adverse events and improving patient outcomes. Narrow-focused AI models can be utilized to estimate the likelihood of intraoperative (during surgery) and post-operative (after surgery) events associated with a patient.

How is this Beneficial?

With the application of ML models to surgery and establishment and implementation of situation-specific intervention procedures, the surgeon can take advantage of this window of opportunity to manage patient risks. Predictive models developed in this manner can assist surgeons in identifying high-risk groups, anticipating significant occurrences, and planning treatment measures to enhance patient outcomes. A great case of this in action was performed by RediMinds.

RediMinds worked with the Vattikuti Foundation to create models for predicting intraoperative complications, or in other words, complications before and after surgery, and 30-day morbidity. The models were built using a multi-institutional dataset that was focused on a specific cohort of kidney cancer patients who underwent Robotic Partial Nephrectomy.

Machine learning and AI models performed well in this study in predicting problems during surgery and 30-day complications following surgery. These tools could be applied to real life clinical settings, which would allow the centers to utilize the models to identify patients who are likely to experience a problem before or after surgery, as well as to innovate intervention methods and offer feedback on model performance. This would also allow researchers to track the models’ performance in a clinical context.

Model Development

Here’s a brief rundown of RediMinds’ model deployment process for this study. First, historical data from patients who underwent robot-assisted partial nephrectomy is obtained, including demographic, pre-operative, intra-operative, and follow-up information. The next step, the data engineering process, then includes pre-processing historical data and defining features that will allow machine-learning algorithms to be trained on the data.

Next, model training and model validation models are trained, and the model with the greatest performance on unseen test data is chosen. The model with the highest performance is put in a clinical setting for further evaluation. The final step is the model monitoring, which is the process of comparing model predictions against ground truth to see if the model is operating as expected.

Final Thoughts

Surgeons strive to make their actions and decisions to treat patients under their care as predictable as possible. With the successful application of ML, it is possible to predict complications before and after surgery, which could be consequential to the patient’s surgical outcome.

This approach and methodology may be used in any surgical circumstance across the healthcare spectrum. It is not restricted to Robotic Partial Nephrectomy patients. With adequate data and continued research, the next generation of assistive AI surgical tools can be built utilizing the technique that RediMinds used.

To learn more, check out the case study

This isn’t the only way that our research into AI in healthcare is improving patient outcomes.

Check out our guide to 6 exciting applications of AI in healthcare .

Get in Touch

AI in ICU using Natural Language Processing on EHR

by Madhu Reddiboina | Aug 26, 2021 | Health Care

AI in ICU using Natural Language Processing on EHR

Intensive care doctors require dependable clinical practice tools to anticipate important occurrences, schedule timely treatments, and keep patients’ families informed. Unfortunately, standard statistical models, like the Early Warning Score, only employ a few important variables, leaving a large unknown amount of potentially valuable data untapped. With AI and machine learning (ML) tools, doctors can potentially have access to more valuable data, such as occurrences that may happen in the first 24 hours of being in the ICU.

However, patient EHRs (Electronic Health Records) contain a large amount of information for every patient. In an ICU, around 9,000 to 10,000 different types of data can be collected for each patient. This poses a challenge for the development of AI and ML tools, which we will discuss further in this article.

The Challenge of ICU data for ML models

AI and machine learning (ML) tools models are the future of medical care. These tools allow hospitals and medical facilities to provide better care for patients and research new surgical techniques and medications. However, researching how well tools and models work in the medical environment and training ML models with past data is a vital first step.

The best way to analyze data and perform research like this is by analyzing the data of past patients. As stated earlier, EHRs have tons of patient data, so much that it’s complicated to train a ML model using this data. There are many elements and information collected on patients in the ICU. From the moment they arrive at the hospital, to follow-ups, to surgeries, there are so many pieces of information. This information isn’t even fully structured data. Most of it’s just doctors’ notes, random information, and pictures, and it’s all in different formats.

Training a ML model is a huge challenge when trying to analyze historical data from the ICU. To train models, the data has to be formatted in a specific way. For instance, one doctor at one hospital may label something as one variable, while another doctor may label it as another variable. The variables are usually the same, although the names differ amongst EMRs.

In certain EMRs, Heart Rate may be coded as ‘Heart Rate’, whereas in others it is coded as ‘HR’. This does not prevent model training, but it makes transferring a model from one facility to another difficult. To accomplish the model transfer in this situation, it would be necessary to map new data variables to old data variables names, then conduct data processing and model training processes. Trying to apply a learned model or procedure to several additional hospitals increases the complexity. Hence why interoperability standards like HLv7 and FHIR are created to make data storage more standardized.

Now, multiply that situation by thousands, and now there is a vast amount of data that needs to be cleaned before researchers can use it to train a ML model. Data cleaning is a huge process for data scientists. Researchers spend a lot of their effort gathering and preparing chaotic digital data before using it to build algorithms. Data comes in all different forms. Data scientists attempt to automate as many procedures as possible, but human language is ambiguous and requires data to be cleaned manually.

In fact, some research has indicated that preparing and maintaining data for analysis takes up around 80% of a data scientist’s work. For ICU data, this would mean going through thousands of variables to ensure they are all formatted correctly. However, it’s vital to have clean data because it provides better outcomes and more accurate ML models.

Natural Language Processing as a Solution

A potential solution to this issue that RediMinds have researched is the application of natural language processing (NLP). NLP analyzes language in text and speech using computer-based approaches. Think of it like this. When we talk to Google Home or Alexa, those devices break down our sentences into tokens. For instance, if you say, “Alexa, light switch on,” the device breaks down the sentence to “light,” “switch,” and “on” and performs an action that it has been trained to do based on these tokens.

Regarding using NLP in the ICU for historical data analysis, NLP models can look for words in EHR notes most commonly used in patients with certain conditions. For example, a patient with a heart issue probably has several words like “cardiac” or “arteries” in their notes. These words can be collected by a ML model, which then can train models on what happens next to this patient.

RediMinds investigated this technique because handpicking or curation of variables takes a long time and may result in the omission of potentially significant data items. The RediMinds team was able to analyze and use data for model training without rejecting large quantities of material by utilizing NLP methods.

Wrapping Up – Why is this Important?

Using NLP techniques to process both structured (tabular data) and unstructured (Physician’s Notes) cuts down on the time it takes to develop new models for patients. This technique also makes it easier to develop subsequent models because researchers can use the same data. Finally, this technique leverages entire datasets instead of hand-selected variables, which helps eliminate biases of what researchers think is important for the model. With this technique, researchers can tap into valuable insights that may have gone previously unknown due to hand-selecting variables to include in models.

Using NLP also leads to interoperability. With so many different ways to record data, it is nearly impossible to build ML models to make ICU predictions efficiently. However, this technique can help with gathering clean data for building future models.

Based on the first 24 hours in the ICU, ML models can answer questions about patient survival chances, how long patients will be hospitalized, and make several other predictions. If a ML model predicts that a patient will not survive more than a certain amount of hours, then this indicates to doctors and ICU staff that this patient needs to be checked on frequently. This type of AI tool in the ICU has the potential to save lives.

To learn more, check out the case study.

Get in Touch

Ground Truth Factory: A Collaborative Platform to Develop AI Enabled Surgical Safety Tools

by Madhu Reddiboina | Aug 24, 2021 | Health Care

Ground Truth Factory: A Collaborative Platform to Develop AI- Enabled Surgical Safety Tools

Artificial intelligence (AI) is on the rise and making its way into several industries. Healthcare is not excluded from that list.

Introducing AI into the operating room can save lives, increase safety, and enhance patient outcomes.

And the creation of intelligent , AI-enabled surgical tools combines the expertise of skilled surgeons with the precision of advanced technology.

But it’s not a simple task to integrate AI into surgery. Medical and AI professionals face many hurdles in gathering all the information needed to build these tools.

That’s why RediMinds created a platform called Ground Truth Factory (GTF) to facilitate data sharing and collaboration between the two professional groups, solving issues faced when developing AI surgical tools.

This article will discuss the details of GTF, how it’s solving major issues, and the impact that it has on the medical field.

Obstacles to Collaboration 

Before going into what GTF is and how it works, it’s important to understand the exact challenges that need to be overcome.

When creating AI surgical tools, collaboration is essential. But there are two issues that come with collaborating for research:

1. Lack of Data

Many AI companies do not have access to the data or medical knowledge needed to analyze surgical pictures or identify items of interest such as tumors, blood vessels, and organs – which are both essential to train the AI model.

2. Lack of Time

Many healthcare facilities have data and access to medical knowledge, but not the time nor the ability to actually create the AI models.

The Main Challenge

The gap in skills and knowledge between the two main players: AI companies and medical professionals.

Despite the clear need for collaboration between these two groups, barriers to data sharing, data ownership questions, and annotation tools make collaboration difficult.

So What is Ground Truth Factory?

Purpose-built by RediMinds, the Ground Truth Factory platform aims to bridge the gap between medical and AI professionals and provide a secure platform for the development of AI enabled surgical tools. 

This platform allows users to upload data, annotate it, manage it, and construct models. It is intended to provide secure and efficient cooperation while shortening the time required for annotation.

Here are some ways that GTF tackles the usual collaborative issues:

GTF is HIPAA Compliant 

There is a process in place to ensure that the data uploaded to the platform does not reveal private information. Also, data storage, access, and processing services are HIPPA compliant.

GTF is Secure 

The entire platform is firewall protected, ensuring that data is safe from cyber-attacks. Furthermore, data can be stored in the country that it was created, making it easier to follow data localization laws.

GTF Gives Data Owners Control and Keeps Them Informed

When data is uploaded, the owner of the data can choose how it will be used and can assign annotation jobs to experts of their choosing who participate on the platform.

GTF Saves Time 

To protect the valuable time of the surgeons who annotate the images, an AI assistive tool can filter out images with specific anatomical objects such as a spleen, renal arteries, and renal veins.

The Future of AI Surgical Tools with Ground Truth Factory 

With the collaborative environment, thousands of experts can work together in the creation of AI surgical tools as GTF has the potential to handle terabytes of data and thousands of images.

The platform enables AI developers to create models that can identify thousands of anatomical objects from hundreds of different surgeries. And as the number of annotators, researchers, and the amount of data processed on the GTF platform continues to grow, the system will be able to handle the increased workload with ease.

Ground Truth Factory connects and combines the skill sets of artificial intelligence and medical research professionals. The information gathered can be used to develop, test, and deploy intelligent surgical tools, such as robotics.

By supporting these collaborative efforts, GTF has the potential to fuel the next generation of artificial intelligence surgical tools – having a significant impact by helping individuals to work together for the greater good.

The full benefits of healthcare providers taking on digital transformations and using AI to create the next generations of surgical tools are yet to be discovered.

AI is certainly helping us take massive strides in healthcare – but it isn’t the only industry seeing exponential change. Check out our article on how AI can make anyone a pro dancer.

Get in Touch

Google Taking on Healthcare and Clinical Artificial Intelligence

by Madhu Reddiboina | Aug 20, 2021 | Health Care

Google Taking on Healthcare and Clinical Artificial Intelligence

Google recently published a paper in NPJ Digital Medicine, wherein they worked with the electronic medical records (EMR) data from two large hospital systems to make a whole range of predictions.

In-hospital mortality
Length of stay
Readmissions
And diagnosis at discharge.

They achieved all this using deep learning and fast healthcare interoperability resources (FHIR).

So what are EMR (Electronic Medical Records)?

Digital transformation of the healthcare industry is well underway. As part of that, EMR have allowed healthcare providers to document everything you’d find on a paper chart electronically, providing all the fundamental patient data medical staff need to make informed decisions about their patients.

Making patient records more accessible and theoretically more sharable between healthcare providers.

Which brings us to FHIR.

What is FHIR (Fast Healthcare Interoperability Resources)?

FHIR is the standard defining how healthcare records can be exchanged between different computer systems.

The aim is to allow developers to build standardized applications no matter how the original EMR data is stored, connecting healthcare providers to share data in a more effective way and removing the need to share patient data via PDF or paper copies.

Limitations of Traditional Machine Learning Models

Traditional machine learning and statistical models are built using a limited number of known variables from patient records ignoring most of the data that exists on every admitted patient.

Technically speaking, this work is very different.

They utilized all the patient data including encounters, labs, medications, notes, procedures, observations, and more to build deep learning models superior to work that predated it.

This paper entitled, Scalable and Accurate Deep Learning with Electronic Health Records published in NPJ Digital Medicine, is a masterpiece with amazing results in predicting patient outcomes using personalized predictive models.

The Study

In the study, the authors assumed the position that by representing a patient’s raw EHR data according to the FHIR format, deep learning can accurately predict outcomes and medical events regardless of the location of the patient from one setting to the next.

According to the authors:

“We validated our approach using de-identified EHR data from two US academic medical centers with 216,221 adult patients hospitalized for at least 24 h. In the sequential format we propose, this volume of EHR data unrolled into a total of 46,864,534,945 data points, including clinical notes.”

Events Predicted and Accuracy Rates Include the Following:

In-hospital mortality – 93 to 94 percent accuracy
30-day unplanned readmission – 75 to 76 percent accuracy
Prolonged length of stay – 85 to 86 percent accuracy
All of a patient’s discharge diagnoses – 90 percent accuracy
Impact of Deep Learning on Healthcare

What This Means for Machine Learning in Healthcare

Deep learning techniques used to accurately predict medical events hold great potential.

Providing better care and monitoring of critical care patients while also optimizing hospital resources.

If put to work in a clinical setting, we can expect the machine and deep learning algorithms to improve the predictability of critical events and the care delivery workflow. Just like how AI algorithms are saving lives affected by hemodynamic instability.

Possibilities to use deep learning to improve the quality of care are endless – with the full benefits AI can have on the healthcare industry still yet to be discovered.

And healthcare isn’t the only area where AI is taking massive strides. Check out our article on how AI can make anyone a pro dancer.

Get in Touch

What the Future of AI and Hemodialysis Therapy Looks Like

by Madhu Reddiboina | Apr 3, 2020 | Health Care

What the Future of AI and Hemodialysis Therapy Looks Like

End-stage kidney disease (ESKD) is something that affects over a million Americans and puts a major burden on healthcare facilities all over the country. And as one of the most common courses of treatment for ESKD, hemodialysis is a current field of medicine being looked at for possible improvements that can be brought on by using artificial intelligence (AI).

This is a crucial step for the healthcare industry because the process of using hemodialysis has remained relatively unchanged in the last couple of decades. Plus, with more than 700,000 people receiving dialysis and only about 7,350 specialized dialysis centers across the country, this is putting a tremendous strain on these facilities.

Therefore, anything that can be done to help improve the efficiency of hemodialysis treatments is something that can lead to significant improvements in the healthcare industry as a whole. Here is what the current state of AI use in dialysis treatments is and what the future looks like for the relationship between AI and hemodialysis therapy.

Existing Problems With Hemodialysis

As previously mentioned, the technology and techniques surrounding hemodialysis therapy have remained largely unchanged so far in the 21st century. This has resulted in medical professionals still having a lack of dialysis devices that are capable of reacting to any changes that occur in the middle of the dialysis treatment. These devices are also unable to continually improve themselves through learning from past treatment sessions and using this information to personalize future sessions.

With the number of complications that occur during dialysis treatments, being able to make real-time changes is crucial to help further lower the amount of dialysis-related deaths and reduce the amount of resources required from healthcare professionals.

Current Use of AI in Hemodialysis Therapy

There are several ways that AI technology is currently being tested for its effectiveness in hemodialysis therapy. Two of the most promising areas are the anemia control model (ACM) and the Hemocontrol™ system.

With the ACM, the AI neural network model predicts future hemoglobin concentrations for patients based on their updated clinical data. It also uses an algorithm that provides the optimal exogenous erythropoiesis-stimulating agents (ESA) and iron dosage in an attempt to reach the target hemoglobin level.

Hemocontrol™ helps to prevent intradialytic hypotension (IDH) by analyzing three different variables that include total patient weight loss, blood volume, and average dialysate sodium level. It then takes this up-to-date information and precisely adjusts the sodium concentration of dialysate and UF rate in order to help reduce the rate of IDH.

Future of AI Use in Hemodialysis Therapy

While the current systems being developed and used are a step in the right direction, they still have many improvements to make before they will be able to help significantly reduce the $28 billion cost of hemodialysis care in the U.S.

The biggest areas currently being focused on are miniaturization, flexibility, portability, wearable tech, and “water use” efficiency. Once implantable bioartificial kidneys and wearable artificial kidney tech exist that use dialysate-regenerating sorbent techniques, the treatment of hemodialysis therapy will see significant advancements and result in lesser burdens on healthcare professionals.

In order to get other current news regarding AI and machine learning use in healthcare, make sure to follow our blog at RediMinds today.

Get in Touch