How Can Clinical Labs Benefit from Machine Learning and AI?

Using AI and machine learning can help clinical labs improve workflows and manage costs

Photo portrait of emily-newton
Emily Newton
Photo portrait of emily-newton

Emily Newton is the editor-in-chief of Revolutionized Magazine, an online publication that discusses the latest news in science and technology. She has more than five years of experience covering stories in the science, technology, and industrial sectors.

ViewFull Profile
Learn about ourEditorial Policies.
Published:Oct 10, 2022
|4 min read
Register for free to listen to this article
Listen with Speechify
0:00
4:00

Artificial intelligence (AI) and machine learning (ML) have impacted many industries, resulting in tremendous, ongoing growth. Statistics suggest that worldwide AI software market revenues will reach approximately $126 billion by 2025.

The technology’s growing availability is allowing clinical labs to explore using AI and ML, such as for accelerating workflows and managing costs. Here are some exciting ways to use AI and ML in your clinical lab.

Supporting clinicians to make faster diagnoses

More than 13 billion lab tests are conducted annually in the US, and these lab results directly or indirectly influence more than two-thirds of medical decisions. The turnaround time of these results also matters, where receiving correct treatment faster can help save lives. AI and ML can help patients quickly get the correct treatments by rapidly analyzing massive amounts of data.

“AI in clinical labs could standardize aspects of interpretative microbiology, supporting microbiologists who cannot reach a consensus.”

An Israeli company called Predicta Med built an AI tool that aids in diagnosing psoriatic arthritis up to four years sooner than conventional methods. The company’s AI checks various information, such as clinical lab results and patient medical records, to identify potential diagnostic clues. For example, many patients seek back or joint pain treatment from orthopedic specialists without linking those issues to their skin condition. Shortening psoriatic arthritis diagnostic timeframes, which can take more than two years, could reduce permanent joint damage in patients since it may occur years before any skin abnormalities become apparent. Eventually, the project team hopes to learn whether earlier diagnoses might reduce joint destruction.

Accelerating workflows with artificial intelligence in clinical labs

Academic researchers and technology providers have also brought AI and ML to analyzer software used in clinical labs. For example, some clinical labs use AI and ML for microbiology plate reading and blood culture cabinet automation. Clinical lab staff can then focus on any positive samples, reducing reporting times. Other opportunities exist to triage plates with and without growth. AI in clinical labs could standardize aspects of interpretative microbiology, supporting microbiologists who cannot reach a consensus.

Some AI-powered automated analyzers differentiate between multiple disease biomarkers in a single sample. These systems eliminate the need to conduct several tests when patient symptoms overlap with numerous ailments.

AI-based analyzers can also help clinical lab professionals verify results, saving time. Urinalysis verifications typically involve a sample analysis to identify specific particle types. A 2021 study conducted by CMI and published in Clinical Microbiology and Infection suggested an AI system that standardized and secured results increased lab technicians’ efficiency.

Planning for clinical labs’ needs with artificial intelligence

Clinical labs play a critical role in supporting population health. However, lab managers often face the daunting task of operating their facilities after budget cuts. AI can help confirm the necessity and size of clinical lab budgets.

Many hospitals use the diagnosis-related group (DGR) system, which codes individual patients and gives them condition-based classifications. Each group has an associated patient care estimate. However, the downside is that actual cost is determined months later, usually after discharge.

According to a study published in NPJ Digital Medicine, researchers developed a better way of using ML algorithms to process clinical notes. Specifically, the researchers used a deep learning-based natural language processing (NLP) model to predict DRG codes and associated costs upon admission from ICU patients at a major medical center in Boston.

“Smart tech may be able to help cut down on repeat testing and reduce clinical lab workloads.”

According to the authors, “DRG prediction using clinical text consistently performed better than clinical measurements alone, demonstrating the value of mining clinical text to support active care management.”

The study’s results suggest that the NLP model could promote better expense forecasting than manually reviewing previously used resources after a patient is discharged. By helping administrators anticipate the use of resources, this forecasting could lead to better decisions around distributing funding to hospital labs and other facilities for staffing and equipment.

Reducing excessive workloads caused by unnecessary tests

Lab data collected from nine clinical sites at Royal Bolton Hospital in the UK has suggested about 25 percent of lab tests were unnecessary repeats. It can be frustrating for everyone when health care providers primarily order lab tests to verify initial findings. Repeat testing raises clinical costs and increases lab workloads. Patients also often get upset because it takes longer to find the cause of their symptoms.

Smart tech may be able to help cut down on repeat testing and reduce clinical lab workloads.

Recent research on smartwatches and fitness trackers has indicated that ML could enable these types of devices to flag potential medical issues, some of which are usually detected via clinical lab tests. Suppose a smartwatch recorded elevated sustained body temperature paired with reduced activity levels, suggesting potential illness. Jessilyn Dunn, a co-lead and co-corresponding author of the study, said in a press release that algorithms and long-term health data could help a doctor narrow down ailments before getting lab test results.

However, as the team of biomedical engineers and genomics researchers at Duke University and the Stanford University School of Medicine explained, while patient vital sign checks and testing are valuable, they only provide health snapshots of one point in time. Wearables can collect long-term data, pinpointing variations from a patient’s baseline, which may be more valuable.

"If you are an organizational decision-maker, start by asking your staff or colleagues which obstacles most disrupt their workflows or time management."

The three-year smartwatch study had participants use wearables that tracked vital signs such as heart rate and skin temperature. The participants also had regular lab visits that monitored blood cell counts, iron levels, and glucose readings through traditional means. The results confirmed multiple links between the smartwatch data and lab test results. Such technology could potentially reduce clinical lab testing workloads by providing clinicians with other means to monitor their patients, minimizing unnecessary tests.

Could artificial intelligence benefit your clinical lab?

Determining the best ways to apply AI and ML in your clinical lab takes time and resources. However, such dedicated efforts are likely to pay off.

If you are an organizational decision-maker, start by asking your staff or colleagues which obstacles most disrupt their workflows or time management. Then, research which purpose-built solutions on the market might help. Service providers familiar with the challenges of the clinical lab can thoroughly explain which use cases best fit their offerings. Those details will help you discover AI and ML applications to improve your lab’s performance.

Overall, optimizing how you use AI and ML in your lab is an ongoing process. Often, the best results occur when people make incremental changes and improvements to systems and settings based on real-world learnings.

Top Image:
Worldwide AI software market revenues will reach approximately $126 billion by 2025.
iStock, filo