SBIR-STTR Award

Determining meat quality attributes in food animals is essential for genetic selection, sorting, and marketing. Ultrasound techniques have the potential to benefit the swine industry for improvements in both quality and yield measures by non-destructive means. There is a need to continually improve percentage intramuscular fat estimation accuracy and reduced bias, especially before integration of percentage IMF levels into payment systems. A unique opportunity exists due to recent availability of a programmable ultrasound scanner with access to both unprocessed, radio frequency (RF) ultrasound backscattered signals and corresponding B-mode images, and recent reports that combining parameters from both signals and images improves accuracy, sensitivity, and specificity for tissue characterization. We hypothesize that the accuracy of percentage IMF estimation in the longissimus dorsi muscle can be improved by combining the RF signal analysis with currently used B-mode image texture analysis. Technical objectives are to: 1) demonstrate the feasibility of simultaneously acquiring B-mode images and RF signals from live animals and carcasses using the new generation of programmable digital ultrasound scanner; and 2) develop signal processing parameters for characterizing percentage IMF and prove that by combining parameters derived from RF signals with texture parameters from images, percentage IMF estimation can be improved. Phase I will include developing a protocol to acquire both ultrasound images and RF signals using a state-of-the-art ultrasound research system; in-house testing of the system, and developing procedures for scanning and processing data. The Phase I output will be the proof-of-concept that RF ultrasound signals can enhance existing image processing-based IMF prediction in swine. To our knowledge, this would be the first feasibility study to integrate ultrasound data collection and processing of both B-mode images and RF signals in ultrasound applications for food animals. Phase II will culminate in developing a prototype integrated acquisition and processing system for evaluation of feedlot animals and carcasses. The proposed hybrid system has the potential to significantly improve marketable ultrasound technology and have a long lasting impact on food animal industries. The Biotronics team is in unique position to lead this effort to develop the next generation ultrasound meat quality evaluation system. OBJECTIVES: The proposed work aims at improving the accuracy, repeatability and bias of ultrasound estimation of intramuscular fat (IMF) in live animals and carcasses. Although ultrasound B-mode and image texture analysis techniques have demonstrated tremendous benefits to the beef industry and have potential to benefit swine industry, three factors create a unique opportunity addressed in this proposal. First is the need for higher accuracy and reduced bias, especially for evaluation of percentage IMF in swine (due to narrower IMF range), feedlot cattle, and carcasses. Second, a programmable ultrasound scanner with significant research capabilities has recently become commercially available. Third, researchers in human medicine have demonstrated that when features extracted from ultrasound raw or backscatter signal analysis were used in combination with the texture features extracted from the corresponding B-mode images, the classification accuracy, sensitivity, and specificity for prostate cancer tissue characterization was significantly improved. B-mode ultrasound technology for livestock evaluation has lagged in keeping up the technological advances in ultrasound. It is challenging, time-consuming, and resource-consuming to incorporate new ultrasound technology for further improvements. In general, the B-mode ultrasound technology works best for ranking animals in a contemporary group. There is a varying degree of bias in the predicted IMF values due to differences in ultrasound equipment, technician, environmental effects, and scanning conditions. For genetic evaluation, undesirable biases get removed in statistical analysis for ranking the animals. However, for online evaluation, these biases will be inherent in the predicted results. Applications such as feedlot (chute-side) cattle evaluation and on-line carcass evaluation in the packing plant demand the highest accuracy possible with the ultrasound technology because decisions are based upon actual measures. In these applications, each individual animal or carcass is evaluated for its quality attribute such as percentage IMF. Carcass meat quality evaluation is progressing from subjective visual assessment to objective measures. The objective methods generally improve consistency and accuracy. The authors of this proposal hypothesize that by using state-of-the art ultrasound technology to acquire and analyze ultrasound raw signals, often referred to as backscatter signal or A-mode signal or radio-frequency (RF) signal, a significant improvement will be made in the IMF prediction accuracy, and a reduction in the bias will be realized. APPROACH: The proposed work builds on a foundation of research-based knowledge in B-mode ultrasound and intends to exploit a unique opportunity for being the first to use a hybrid ultrasound system for meat evaluation. The availability of state-of-the-art research ultrasound scanner on lease from ISU and the on-going SBIR Phase-II for the development of swine carcass percentage IMF evaluation (using ultrasound images, without RF signals) provide a unique and highly leveraged opportunity. If successful, the proposed technology could be packaged with the customized ultrasound scanner for the animal and packing plant applications. The existing programs of genetic improvement using expected progeny differences (EPD), coupled with the ultrasound live animal measurement programs, provide an established vehicle to easily and quickly adopt percentage IMF measurement and prediction tools for producers and packers. Iowa State University researchers demonstrated that the basic technology works in swine, showing rank correlations between predicted ultrasound percentage IMF and actual ether extract to be greater than 0.50. Recently, Biotronics demonstrated that the rank correlations can be improved up to 0.64 with improved image processing and statistical model development procedures. The authors believe that this rank correlation may be further improved upon for the swine industry with the new RF ultrasound equipment that will allow combining the best image parameters and RF signal parameters. Integration of percentage IMF in payment systems requires an objective, reliable and highly accurate B-mode technology to non-destructively determine percentage IMF in live animals and carcasses. It is also the belief of the authors that the addition of RF ultrasound signal based parameters to existing image processing parameters will improve the percentage IMF prediction accuracy and reduce the bias. The approach is to develop a hybrid system and protocol for capturing both RF and image ultrasound data and develop databases of both types of ultrasound data using unique state-of-the-art ultrasound equipment along with loineye tissue samples to characterize the actual percentage IMF through chemical analysis and then developing new preliminary percentage IMF prediction models. Both live pigs and pork carcasses will be scanned along with the collection of chemical intramuscular fat data to develop the research database. The proposed Phase I will provide a significant technology base from which to begin Phase II of this project. The scanning protocol will be developed for combined RF and image ultrasound and boundaries will be established for equipment gain and other settings. New RF signal processing parameters will be developed that show significant predictive capability for IMF and complement the image texture processing parameters. This knowledge base will allow the hybrid ultrasound (RF signals plus B-mode image) capture and processing procedures of Phase II to be highly focused and directed towards additional improvement in overall system accuracy. Phase II will culminate in a prototype system and related software implementation that can be commercialized

Determining meat quality attributes in food animals is essential for genetic selection, sorting, and marketing. Ultrasound techniques have the potential to benefit the swine industry for improvements in both quality and yield measures by non-destructive and non-invasive means. Biotronics, Inc. has developed and markets products for evaluation of percent intramuscular fat (% IMF) in longissimus dorsi muscle and body composition traits in live pigs using conventional real-time ultrasound (B-mode) imaging technology. It has recently introduced this technology in a commercial prototype for hot carcass instrument grading. The premise of the project is that swine breeders can be rewarded financially for production of pigs resulting in higher levels of % IMF if the pork packing plant realizes premiums for their product. There is a need to continually improve accuracy and reduce bias for estimation of % IMF, especially for integration of % IMF levels into payment systems. The overall objective of this SBIR project is to significantly improve the % IMF and % lean evaluation by utilizing the new hybrid digital ultrasound technology. A unique opportunity exists due to recent availability of a programmable ultrasound scanner with access to both unprocessed, radio frequency (RF) ultrasound backscattered signals and corresponding B-mode images, and recent reports that combining parameters from both signals and images improves accuracy, sensitivity, and specificity for tissue characterization. The Phase-I of this project successfully demonstrated the feasibility of simultaneously acquiring ultrasound images and RF signals from live pigs using a new generation of programmable digital ultrasound scanner and demonstrated greater than 10% improvements in % IMF estimation accuracy by combining parameters from RF signals with texture parameters from images. During the Phase-II, we will assemble a commercial prototype of the hybrid system for hot carcass instrument grading and will test the system in the packing plant by scanning carcasses at a line speed of 1200 per hour. We will refine the methods for real-time acquisition and processing of hybrid ultrasound signal, spectrum and image data. We will implement algorithms for fully automatic and accurate determination of % IMF and % lean, in near real-time at the line speed, by automatically identifying acceptable quality data frames and anatomical landmarks. Benefits of the proposed next generation system will include significant improvement in % IMF in an overall customized system design that integrates digital ultrasound and computer into one device and reduces the number of components. This will also lead to better quality control and longer lifecycle of the proposed commercial carcass instrument grading system. The Biotronics team is in unique position to lead this effort to develop the next generation ultrasound meat quality evaluation system. The proposed hybrid system has the potential to significantly improve marketable ultrasound technology and have a long lasting economic impact on all segments of the pork industry, with the ultimate benefit to be realized by the consumer. OBJECTIVES: The long-term goal of the project is to develop marketable technology that aimed at improving the quality of food animal products. Specifically, the proposed work will develop a new hybrid system to improve ultrasound-based evaluation of quality attributes in live swine and beef animals and carcasses. It is hypothesized that by combining ultrasound raw signal analysis with currently used image analysis will result in improvement in accuracy, repeatability and bias in evaluation of % intramuscular fat (IMF) in the longissimus dorsi (LD) muscle by non-invasive and non-destructive means. This improvement will be beneficial for seedstock evaluation (quality improvement), feedlot evaluation (quality based marketing decisions and production efficiency), and packing plant (evaluation of quality of end products, quality based marketing and production efficiency). The proposed work has potential to develop significantly improved marketable ultrasound technology that will have a long lasting impact in multiple segments of food animal industries. Technical objectives for Phase II are to: 1) Refine hybrid ultrasound processing algorithms to computationally efficiently characterize % IMF levels in the pork LD muscle tissue in live animals and hot carcasses; 2) Develop, implement and validate technology for completely automating the IMF evaluation procedure including the steps of automatic identification of acceptable quality data, anatomical landmarks and suitable region-of-interest for IMF evaluation; 3) Assemble the first hybrid prototype system for real-time hybrid data capturing and processing and demonstrate an operational system for hot carcass scanning in the packing plant; and 4) Scan at least 500 carcasses at a packing plant to evaluate the system performance in the packing plant environment at the line speed of 1200 carcasses per hour. The full R&D program will lead to commercialization of the hybrid ultrasound system using the Biotronics developed hot carcass grading platform that currently utilizes image-based technology for evaluation of swine carcass composition and % IMF in a packing plant at a line speed of 1200 carcasses per hour. The Phase-II expected output will be a fully functional commercial prototype that incorporates the new generation of hybrid ultrasound acquisition and processing technology for near real-time evaluation of pig carcasses for improved quality and yield evaluation. The hybrid system will also provide additional benefits of improved overall system design by integration of ultrasound scanner and computer into one single digital programmable system that requires no signal conversion, no frame grabber and no multiple external interface cables and components. This will also lead to better quality control and longer lifecycle of the product lines. APPROACH: The Phase-I of this project successfully demonstrated the feasibility of simultaneously acquiring ultrasound images and RF signals from livestock using a new generation of programmable digital ultrasound scanner; developed signal processing parameters for characterizing IMF; and demonstrated that IMF estimation can be improved by combining parameters derived from RF signals with texture parameters from images. This Phase I work provides a significant technology base from which to begin Phase II of this project. During the Phase-II, the work plan for the four objectives will include a total of 12 work tasks for the duration of 24 months. We will further develop and refine the methods for more efficient acquisition of the hybrid data and processing of signal, spectrum and image parameters. Additionally, we will develop, implement and validate fully automatic data acquisition scheme using pressure sensors and analysis scheme to accurately determine % IMF and % lean by identifying acceptable quality data frames, identifying anatomical landmarks (e.g., interfaces for the ribs, fat and muscles) for ROI and presenting the results in near real-time at 1200 carcasses per hour. We will assemble a commercial prototype of the hybrid system for hot carcass evaluation and will test the system in the packing plant environment by scanning at least 500 carcasses. We will test the prototype hybrid system in a packing plant environment for scanning and evaluating carcasses in near real-time at a line speed of about 1200 per hour. This will validate the functionality of the hybrid technology in a commercial prototype as compared to the current image-only technology. The predicted % IMF for each carcass will be compared to its respective marbling score and chemical % IMF. The primary test statistic will be the root mean square error of the predicted results as compared to the actual values. Ability of the models to differentiate loins into defined % IMF specification groups will also be analyzed. Benefits of the proposed next generation system will include significant improvement in % IMF and highly customized overall system design by integration of ultrasound scanner and computer into one single digital programmable system that requires no signal conversion, no frame grabber and no multiple external interface cables and components. This will also lead to better quality control and longer lifecycle of the product lines. We anticipate incorporating this new technology into the new generation of our existing commercial product platform, BioQscan(R) for carcass evaluation