Escherichia coli O157:H7 is a food-borne pathogen that causes bloody diarrhea, hemorrhagic colitis, and hemolytic uremic syndrome (HUS). We compared three selective media and evaluated the performance of immunomagnetic separation (IMS) for the detection of low levels of E. coli O157:H7 in ground beef and radish sprouts with different levels of background flora. Bulk food samples (500 g for each trial) were artificially inoculated with nalidixic acid-resistant E. coli O157:H7 at the lowest dose that would generate 20 partial-positive samples of 25 g each. All samples were homogenized in mTSB (225 mL) and incubated overnight at <TEX>$37^{\circ}C$</TEX>. IMS was performed using the enriched mTSB samples (1 mL) along with conventional spreads plated onto three different selective media: Sorbitol MacConkey agar (SMAC), Sorbitol MacConkey agar with cefixime and tellulite (CT-SMAC), and Sorbitol MacConkey agar with nalidixic acid (NAL-SMAC) as the gold standard. Two suspicious colonies from each medium were selected and confirmed usinga serological test after transfer to tryptic soy broth with yeast extract (TSAYE). CT-SMAC was better than SMAC for detecting E. coli O157:H7 in all food types. Although there was no statistical difference in the number of positive samples when using IMS vs. non-IMS techniques, more positive samples were detected when IMS was used in both ground beef and radish sprouts. It appears that the improvement was more significant in radish sprouts, which had a higher level of background flora than ground beef. The results also suggest that the combination of CT-SMAC and IMS is sufficient to recover low levels of E. coli O157:H7 in high background flora food samples.