M. Hajmeer

A probabilistic neural network approach for modeling and classification of bacterial growth/no-growth data

In this paper, we propose to use probabilistic neural networks (PNNs) for classification of bacterial growth/no-growth data and modeling the probability of growth. The PNN approach combines both Bayes theorem of conditional probability and Parzens method for estimating the probability density functions of the random variables. Unlike other neural network training paradigms, PNNs are characterized by high training speed and their ability to produce confidence levels for their classification decision. As a practical application of the proposed approach, PNNs were investigated for their ability in classification of growth/no-growth state of a pathogenic Escherichia coli R31 in response to temperature and water activity. A comparison with the most frequently used traditional statistical method based on logistic regression and multilayer feedforward artificial neural network (MFANN) trained by error backpropagation was also carried out. The PNN-based models were found to outperform linear and nonlinear logistic regression and MFANN in both the classification accuracy and ease by which PNN-based models are developed.

Comparison of logistic regression and neural network-based classifiers for bacterial growth

Abstract In this paper, we introduce two fundamentally different approaches for designing classification models (classifiers); the traditional statistical method based on logistic regression and the emerging computationally powerful techniques based on artificial neural networks (ANNs). Linear and nonlinear logistic regression as well as feedforward error backpropagation artificial neural networks (FEBANN) and probabilistic neural networks (PNN) based classifiers were developed and compared in relation to their accuracy in classification of bacterial growth/no–growth data pertaining to pathogenic Escherichia coli R31 as affected by temperature and water activity. The comparisons between the four developed classifiers were primarily based on analysis of the receiver operating characteristic (ROC) curves as well as a number of scalar performance measures pertaining to the classification contingency matrices. The ANN-based classifiers outperformed the logistic regression based counterparts. Within the same group, the PNN-based classifier was more accurate than the FEBANN-based classifier, and the nonlinear logistic regression-based classifier was more accurate than the linear one. The optimal PNN-based classifier was a perfect classifier with 100% growth detection accuracy and zero false alarm rate. The advantages and limitations pertaining to the development of the various classifiers were discussed.

Spinal cord tissue detection in comminuted beef: comparison of two immunological methods.

Two commercial immunological kits for detection of central nervous system (CNS) tissue in beef were compared: ScheBo® Brainostic™, based on CNS-specific antigen (neuron specific enolase) detection, and Ridascreen® Risk Material 10/5 test, an enzyme immunoassay for glial fibrillary acidic protein. Spinal cord (SC) was added to batches of choice, select, and utility grades of ground fresh beef shoulder clod to yield 0.0, 0.0125, 0.025, 0.05, 0.1, 0.2, 0.4, 0.8, and 1.6% SC in meat. Sensitivity and specificity in detecting SC in fresh and frozen samples were determined. Both Brainostic™ and Ridascreen® kits detected SC at claimed levels: 0.25% and 0.11%, respectively. The Ridascreen® test consistently detected SC at 0.025%, below its claimed sensitivity level, expressed for brain and SC combined. The Ridascreen® test was ∼10× more sensitive, easier, faster to run and less expensive than the Brainostic™. Overall, quality grade had no influence on SC detection in fresh or frozen meat.

Water, sodium chloride and acidified sodium chlorite effects on Escherichia coli O157:H7 and Staphylococcus aureus on beef briskets.

Effectiveness of spray application of potable water wash (WW), 25% (w/v) sodium chloride (NaCl), and 0.1% (v/v) acidified sodium chlorite (ASC) was evaluated against Escherichia coli O157:H7 and Staphylococcus aureus inoculated onto beef briskets. The purpose was to identify antimicrobial treatments which may be applied to beef carcasses and more specifically in kosher meat facilities. Treatments were applied for 10-60 s at pressure of 419 kPa. Water wash, NaCl, and ASC significantly reduced E. coli O157:H7 as compared with the control, although, only ASC resulted in improved removal with increased exposure time. Water wash did not significantly reduce S. aureus counts throughout exposure and NaCl was only effective after 60 s of exposure, while ASC reduced counts throughout exposure. E. coli O157:H7 was twice as sensitive to WW and NaCl as S. aureus in terms of percent reduction in cell count.

Effects of preparation method, age, and plating technique of thin agar layer media on recovery of Escherichia coli O157:H7 injured by sodium chloride

The thin agar layer (TAL) method was experimentally tested to determine its ability to recover Escherichia coli O157:H7 injured by sodium chloride (NaCl). Cells grown in Brain Heart Infusion broth with 0%, 5%, or 7.5% (w/v) NaCl were spread and spiral plated onto Tryptic Soy agar (TSA), MacConkey Sorbitol agar (MSA), and TSA/MSA TAL combinations. Generally, TSA recovered more injured cells than TAL (p < or =0.05), and TAL recovered more cells than MSA (p < or =0.05). Preparation mode (two vs. three layers) and age (0, 1, or 7 days) of TAL had negligible effect on resuscitation of injured cells (p > 0.05). TAL, which is conventionally used to recover heat, cold, and acid-injured foodborne pathogens, may be used to recover NaCl-injured E. coli O157:H7.

Pathogen survival in chorizos: ecological factors.

This study addressed health risks from ethnic sausages produced on a small scale, without inspection, in California and elsewhere. Mexican-style chorizo, a raw pork sausage that is not cured, fermented, or smoked, was contaminated experimentally in the batter with Escherichia coli O157:H7, Listeria monocytogenes, or Salmonella serotypes and stuffed into natural casings. Formulations were based on a market survey in California. Physical parameters that were controlled were pH, water activity (a(w)), and storage temperature. The pH was adjusted with vinegar, stabilizing at 5.0 within 24 h. Initial a(w) levels adjusted with salt were 0.97, 0.95, 0.93, 0.90, and 0.85; levels declined with time because of evaporation. Pathogen numbers declined with storage up to 7 days, with few brief exceptions. Main effects and interactions of constant temperature and pH with declining a(w) on survival of the pathogens were determined. Maximum death rates occurred at higher a(w) for E. coli O157:H7 and Salmonella than for L. monocytogenes. Salt used to adjust a(w) affected palatability. Spices (black pepper, chili pepper, chili powder, cumin, garlic, guajillo pepper, oregano, and paprika) comprised another, potentially significant aspect of the sausage formulation. Some (notably black pepper and cumin) carried an indigenous microflora that contributed significantly to the microbial load of the sausage batter. Only undiluted fresh and powdered garlic exhibited a significant antimicrobial effect on the pathogens. Although each of the tested formulations caused death of the inoculated pathogens, none of the death rates was sufficiently rapid to ensure safety within the probable shelf life of the product.

Survival of Salmonella and Escherichia coli O157:H7 in Chorizos

Mexican-style raw meat sausages (chorizos) are not regulated in California when they are produced in small ethnic food markets. These sausages are sold uncooked, but their formulation imparts a color that may lead the consumer to assume that they are already cooked, and thus the chorizos may sometimes be eaten without proper cooking. If pathogens are present in such cases, illness may result. Survival of Salmonella and Escherichia coli O157:H7 in chorizos was evaluated under different storage conditions selected based on an initial survey of uninspected chorizos in California. Chorizos were formulated with five different initial water activity (aw) values (0.85, 0.90, 0.93, 0.95, and 0.97), stored under four conditions (refrigeration at 6 to 8 degrees C, room temperature at 24 to 26 degrees C, under a hood at 24 to 26 degrees C with forced air circulation, and incubation at 30 to 31 degrees C with convective air circulation), and sampled after 1, 2, 4, and 7 days. The initial pH was 4.8 and remained near 5.0 from day 1 of the sampling period. Two separate studies of packs inoculated with five-strain cocktails of Salmonella and of E. coli O157:H7 were performed twice for each initial aw. The three lowest aw values (0.85, 0.90, and 0.93) and the incubation and hood storage conditions were more effective (P < or = 0.05) at reducing the target pathogen levels in chorizos than were the two highest aw values (0.95 and 0.97) and the refrigeration storage condition, regardless of storage time. These results provide a scientific basis for guidelines given to producers of uninspected chorizo and should reduce the probability of foodborne illness associated with these products.

Translocation of natural microflora from muscle surface to interior by blade tenderization

The effect of blade tenderization on translocation of natural microflora from the surface to the interior of longissimus dorsi steaks aged for 7, 14, and 21 days was evaluated. Samples from the exterior and interior of steaks from blade-tenderized (BT) and non-blade-tenderized (N-BT) strip loins were analyzed for aerobic plate, coliform, and Escherichia coli counts. Results showed that BT translocated microorganisms (aerobic plate counts) from the exterior to the interior of muscle. Microorganism numbers increased with extended storage (P<.05). Counts of coliforms and Escherichia coli recovered from BT steaks were comparable to those from N-BT steaks because of very low exterior counts, showing the importance of good hygiene.

Efficacy of buffered sodium citrate alone and in combination with sodium diacetate againstListeria monocytogenes on beef franks

We assessed the antimicrobial efficacy of buffered sodium citrate alone and in combination with sodium diacetate against L. monoyctogenes on beef frank samples stored at 39°F. Initial inoculum level of L. monocytogenes was 1.5 log colony forming units (CFU)/cm. After 6 weeks of incubation at 39°F, the pathogen reached 5.4 log CFU/cm in the control sample, but was 1.2 log CFU/cm and 0.85 log CFU/cm in samples treated with 1% buffered sodium citrate alone and in combination with 0.1% sodium diacetate, respectively. Use of buffered sodium citrate and the combination of buffered sodium citrate and sodium diacetate should improve safety of ready to eat foods by controlling L. monocytogenes during storage.