[ad_1]
Researchers of the Kimmel Oncological Center of Johns Hopkins Universitynext to Ludwig Center and the Whiting Engineering School from the same university (USA), They have presented two studies that introduce a more reliable and precise AI for medical diagnosis.
The first job describes the design of Might (Multidimensional Reported Generalized Hypothesis Testing), a specifically created method for Offer the relevant reliability in clinical decisions. The authors applied it to tests based on the detection of circulating free DNA fragments (CCFDNA), and found that certain fragmentation patterns not only occur in people with cancer, but also in those who suffer from autoimmune or vascular diseases.
To better distinguish between these conditions and Avoid false positivesMight expanded and incorporated data on autoimmune and vascular diseases. This improved approach achieved remarkable results: in an analysis with 1,000 samples (352 of patients with advanced cancer and 648 without cancer), the model reached a sensitivity of 72 % and a specificity of 98 % – that is, detected most real cases and barely generated false positives.
Dr. Joshua Vogelstein, one of the researchers, emphasizes that Might is effective “to measure uncertainty and strengthen reliability”, even with limited samples but complex data.
Greater clinical precision
In addition, a complementary algorithm called was developed Comightwhich combines multiple biological signals. It was tested in samples of patients with breast cancer and pancreas in early stages – 125 cases of each type – together with 500 controls without cancer. Although pancreas cancers presented a more common detectionComight evidenced the potential to combine different biomarkers to optimize the detection of breast cancer in initial stages.
An unexpected finding of the study was that inflammation, rather than cancer itself, could generate certain DNA fragmentation patternswhich explains why these marked are also in other diseases. Incorporating data on inflammation allowed to reduce, although not to eradicate, the false ones starring non -cancerous conditions.
Together, these investigations show so much The immense potential as the challenges in the development of reliable clinic. They also point out the need to continue working for tools such as Might to adapt to the daily medical environment.
[ad_2]
Source link
