A Comparison of Precision in General Search Engines and Specialized Databases for Radiology Image Retrieval

The aim of this study was to compare the general search engines and specialized radiology databases in radiology image retrieval and ranking them in terms of accuracy, respectively. An analaythical survey was conducted in this applied research. For doing this, 5 disease groups including heart and brain strokes, coronary artery diseases, cancers, cerebral embolism and respiratory diseases that based on “Death Statistics” book had the highest mortality in Iran were selected. Eight keywords were selected. These keywords were searched in the selected general search engines such as Google,Yahoo, Bing and radiology-related search databases i.e. Goldminer, Yottalook, Searching Radiology and Learning Radiology. Then top 10 retrieved images in each search engine were stored and presented to 30 Radiologist in several hospitals by using researcher made tool. Searching Radiology with 113,533(74.3%) retrieved images is in 1st place, the next place belongs to Yottalook with 28,874(18.9%) retrieved images. Goldminer with 10185(6.64%) retrieved images stands in 3rd place and Learning Radiology with 197(0.12%) retrieved images stands in the last place. In terms of accuracy, due to significant differences between databases and search engines, Google with 95% Precision is in 1st place. The next one is Yahoo with 92%, Bing with 89.1% is allocated with the next rank. 4th place belongs to Searching Radiology whit Precision rate 87.8%, Goldminer with 83.3% stands in 5th place, Yottalook with 81.6% is in 6th place and Learning Radiology with 65% accuracy rate stands in the last position. Google, among general search engines and Searching Radiology among specialized radiology databases had the highest accuracy. Given the significant level to get 0.00 and that significance accuracy differences between general search engines and specialized radiology databases, it seems general search engines to be more accurately than specialized radiology databases.