AUCs were 0.997, 0.986 for distinguishing between benign, malignant biopsy cores, respectively
WEDNESDAY, Jan. 15, 2020 (HealthDay News) — An artificial intelligence (AI) system can be used to detect and grade prostate cancer in prostate needle biopsy samples, according to a study published online Jan. 8 in The Lancet Oncology.
Peter Ström, from the Karolinska Institutet in Stockholm, and colleagues used 6,682 slides from needle core biopsies from 976 participants in a Swedish population-based study (STHLM3) and 271 slides from 93 men from outside the study to train deep neural networks for prostate biopsy assessment. The networks were evaluated using an independent data set with 1,631 biopsies from 246 men from STHLM3 and an external validation set of 330 biopsies from 73 men. Grading performance was assessed on 87 biopsies graded by 23 experienced urological pathologists.
The researchers found that on the independent and external validation data sets, AI achieved an area under the receiver operating characteristics curve (AUC) of 0.997 and 0.986, respectively, for distinguishing between benign and malignant biopsy cores. The association between cancer length predicted by the AI and assigned by the reporting pathologist was 0.96 and 0.87 for the independent data set and the external validation data set, respectively. The AI system achieved a mean pairwise kappa of 0.62 for assigning Gleason grades; this value was within the range of the corresponding values for the expert pathologists (0.60 to 0.73).
“Our results show that it is possible to train an AI system to detect and grade prostate cancer on the same level as leading experts,” a coauthor said in a statement.
Several authors disclosed having patents pending related to cancer diagnostics.
Copyright © 2020 HealthDay. All rights reserved.
