1. Home
  2. News
  3. News Archive
  4. Improving Kidney Stone Measurements With Automated Systems

Improving Kidney Stone Measurements With Automated Systems

Urinary stones, also called kidney stones, are pebble-like structures that form due to buildup of minerals from urine in the kidneys. About 11 percent of men and 6 percent of women in the United States have kidney stones at least once during their lifetime. If small enough, kidney stones can pass through the urinary tract freely, while larger stones can cause symptoms such as severe pain, bleeding, and urinary blockage. Treatment courses for kidney stones are informed by the size and location of stones and can involve simply monitoring for their passage, using methods to break them into smaller pieces, or surgically removing the stones. Reliable and precise measurement methods are therefore crucial for accurate diagnosis and treatment of kidney stones.

Multiple tests are used collectively to diagnose kidney stones, including lab tests like blood and urinary analyses to measure levels of certain minerals, and imaging tests like x-rays and computed tomography (CT) scans to visualize the location of the stones. Some stones can be missed by x-rays depending on their size and location. Therefore, CT scans, which consist of x-rays taken at multiple angles that are computer processed to produce detailed images, are more commonly used to visualize kidney stones. While CT scans are better able to detect stones, they require manual input to determine the size and other characteristics of kidney stones and to measure the anatomy surrounding them. This need for human input introduces variability and less precise measurements, as the same stone may be characterized differently from one person to the next. This variability can make it difficult to assess factors such as the likelihood of a stone passing and the measurement of changes in stone size over time. A more automated measurement process could help to ensure reliable kidney stone measurements and reduce the need for the tedious, more variable manual process.

NIDDK supports research to develop automated, more reliable measurement tools for kidney stone detection. For example, the Center for Machine Learning in Urology (CMLU), which is a joint venture of Children’s Hospital of Pennsylvania and University of Pennsylvania, aims to use machine learning with CT imaging to improve prediction of kidney stone passage. In a recent advance from the CMLU, scientists described the ability of their machine- learning algorithm to accurately measure kidney stones and compared it to the manual measurements of three different researchers. Of the 94 children and adults included in the study, both manual measurers and the algorithm were able to detect that 42 of the patients had kidney stones. However, the algorithm was shown to provide more reliable measurements of stones and regions of the urinary tract than manual input. The algorithm also had a quicker average measuring time of 12 seconds— regardless of the number of stones—compared to increased manual input times with the presence of more stones. Another NIDDK-funded program, the O’Brien Urology Research Center at Mayo Clinic, has supported the development of a semi-automated software system called qSAS. The system, which is currently in use for CT research at the Mayo Clinic and is freely available to other research groups, provides standardized stone characterization with minimal manual input.

Automated CT scan image measurement processes can lessen, and in some cases remove, the variability that comes from manual measuring, while providing more accurate assessments. Machine-learning CT imaging tools like the one developed by CMLU researchers have an added advantage of automatic continued refinement as more data points are entered. With further validation and optimization, the quicker and more reliable measurements offered by these new technologies have the potential to ensure more precise research and diagnosis of kidney stones.

Share this page
Facebook X Email WhatsApp LinkedIn Reddit Pinterest