Artificial active

Study finds artificial intelligence can improve diabetes diagnosis

A new study has found that a fully automated artificial intelligence (AI) deep learning model can identify early signs of type 2 diabetes on abdominal CT scans.

The results of the study have been published in the journal “Radiology”. Type 2 diabetes affects approximately 13% of all American adults, and an additional 34.5% adults meet the criteria for prediabetes. Due to the slow onset of symptoms, it is important to diagnose the disease early. Some cases of prediabetes can last up to 8 years and earlier diagnosis will allow patients to make lifestyle changes to alter the progression of the disease.

Abdominal computed tomography may be a promising tool for diagnosing type 2 diabetes. CT imaging is already widely used in clinical practice and can provide a significant amount of information about the pancreas.

Previous studies have shown that diabetic patients tend to accumulate more visceral fat and fat in the pancreas than non-diabetic patients. However, little work has been done to study the liver, muscles and blood vessels around the pancreas, said study co-senior author Ronald M. Summers, MD, PhD, principal investigator and radiologist at National Institutes of Health Clinical Center in Bethesda, Maryland.

“The analysis of pancreatic and extra-pancreatic features is a new approach and has not been demonstrated in previous work to our knowledge,” said first author Hima Tallam, BSE, MD/Ph.D. student.

Manual analysis of pancreatic low-dose contrastless CT images by a qualified radiologist or specialist is a time-consuming and difficult process. To address these clinical challenges, there is a need for improved automated pancreatic image analysis, the authors said.

For this retrospective study, Dr. Summers and colleagues, working closely with co-lead author Perry J. Pickhardt, MD, professor of radiology at the University of Wisconsin School of Medicine and Public Health, used a data set of patients who underwent routine colorectal surgery. CT scan cancer screening at the University of Wisconsin Hospital and Clinics.

Of the 8,992 patients who were screened between 2004 and 2016, 572 were diagnosed with type 2 diabetes and 1,880 with dysglycemia, a term that refers to blood sugar levels that are too low or too high. There was no overlap between diabetes and dysglycemia diagnosis.

To build the deep learning model, the researchers used a total of 471 images obtained from various datasets, including the Medical Data Decathlon, The Cancer Imaging Archive and the Beyond Cranial Vault challenge. The 471 images were then divided into three subsets: 424 for training, 8 for validation, and 39 for test sets.

The researchers also included data from four active learning cycles.

The deep learning model showed excellent results, demonstrating almost no difference compared to manual analysis. In addition to various pancreatic characteristics, the model also analyzed visceral fat, density and volumes of surrounding abdominal muscles and organs.

The results showed that diabetic patients had lower pancreas density and higher amounts of visceral fat than non-diabetic patients.

“We found that diabetes was associated with the amount of fat in the pancreas and in the abdomen of patients,” Dr. Summers said.

“The fatter the patients were in these two places, the more likely the patients were to have diabetes for a long time,” she added.

The best predictor of type 2 diabetes in the final model included intrapancreatic fat percentage, pancreatic fractal dimension, severity of plaque between the level of L1-L4 vertebrae, mean liver CT attenuation, and BMI .

The deep learning model used these predictors to accurately discern patients with and without diabetes.

“This study is a step towards wider use of automated methods to address clinical challenges,” the authors said.

“It could also inform future work on the reason for the pancreatic changes that occur in diabetic patients,” they concluded. (ANI)