Four proposed state-of-the-art image search engines for automating search and retrieval of digital histopathology slides were found to be of inadequate performance for routine clinical care, US research suggests.
Some of the artificial intelligence (AI) algorithms to power the histopathology image databases had less than 50% accuracy.
Study co-lead UCLA’s Dr Helen Shang said: “There are many AI algorithms being developed for medical tasks but there are fewer efforts directed on rigorous, external validations. The field has also yet to standardise how AI algorithms should be best tested prior to clinical adoption.”
The four engines examined were Yottixel, SISH, RetCCL and HSHR. Each takes a different approach toward indexing, database generation, ranking and retrieval of images.
Overall, the researchers found inconsistent results across the four algorithms and search engine results were of low to average quality, with several visible errors.
They are devising new guidelines to standardise the clinical validation of AI tools, and developing new algorithms that leverage a variety of different data types to develop more reliable and accurate predictions.
Image Credit | iStock
