Deep learning drills down on hyperlipidemia

Researchers in China have developed a deep learning algorithm able to diagnose hyperlipidemia—elevated levels of cholesterol, fats and triglycerides in the bloodstream—in both blood and urine specimens, potentially giving clinicians more information with less expense to the patient.  

Lead author Quan Zhang of Tianjin University and colleagues published their findings online in the journal Diabetes, Metabolic Syndrome and Obesity: Targets and Therapy.

The team built a neural network with an algorithm that, according to their study report, goes beyond hyperlipidemia classification to identify the condition’s parameters at the level of subtle but telling particulars.

For example, it can tell if patients have high blood sugar or markers of diabetes in their hemoglobin.

In the study, the algorithm made an accurate diagnosis 91.5% of the time.

The authors note their key achievement is advancing a previously proven deep learning algorithm such that it can now automatically extract all available clinical information from raw lab data, without human involvement.

Additionally, because the algorithm does not lose raw data, it may have the potential to find more diagnostic markers of different diseases, the authors explain.

“The proposed diagnostic method has a highly robust and accurate performance and can be used for tentative diagnosis,” the authors conclude. “It can automatically diagnose diseases by using human physiological parameters, thereby reducing labor cost, which results in effective improvement of clinical diagnostic efficiency.”

They added that, because their model uses parameters of hematology and urinology to diagnose hyperlipidemia rather than only using blood, it can give a more comprehensive diagnosis to guide clinicians’ treatment decisions.

The full study is available for downloading in PDF format.