1 Background Scott Automation Robotics(SCOTT)has developed an Automated Beef Rib Cutting machine using dual energy x-ray(DEXA)technology to determine cutting lines.The purpose of this project is to determine whether the same x-ray technology can deliver Objective Carcase Measurement(OCM)of bone,fat and muscle. SCOTT will utilise the x-ray technology present in this system to perform preliminary offline trials,analysing beef primal cuts and portions to determine whether the technology can deliver OCM of bone,fat and muscle composition.Murdoch University will be engaged to develop the trialling methodology,assist in conducting the trials and to perform the analysis required to assess whether the Automated Rib Cutting x-ray system is suitable for beef OCM calculations Below are examples of the images SCOTT has successfully obtained in their x-ray Rib Cutting project: Identifying sensing technologies able to improve current processing and/or provide a platform for OCM and automation is a key focus for automation RD&E suppliers such as SCOTT and the industry as a whole.Currently,there is no single technology proven to be able to measure carcase OCM characteristics while also being able to enhance or provide a platform for automation,particularly for beef processing.Being able to utilise a single x-ray system as a platform for Automation and OCM will provide a major step in sensing automation for red meat processing. Potential benefits of successfully advancing the use of pre-developed technology for OCM includes Utilisation of common technology for OCM and automated cutting (cost,footprint, enhanced return on investment) Page 6 of 44
Page 6 of 44 1 Background Scott Automation & Robotics (SCOTT) has developed an Automated Beef Rib Cutting machine using dual energy x-ray (DEXA) technology to determine cutting lines. The purpose of this project is to determine whether the same x-ray technology can deliver Objective Carcase Measurement (OCM) of bone, fat and muscle. SCOTT will utilise the x-ray technology present in this system to perform preliminary offline trials, analysing beef primal cuts and portions to determine whether the technology can deliver OCM of bone, fat and muscle composition. Murdoch University will be engaged to develop the trialling methodology, assist in conducting the trials and to perform the analysis required to assess whether the Automated Rib Cutting x-ray system is suitable for beef OCM calculations. Below are examples of the images SCOTT has successfully obtained in their x-ray Rib Cutting project: Identifying sensing technologies able to improve current processing and/or provide a platform for OCM and automation is a key focus for automation RD&E suppliers such as SCOTT and the industry as a whole. Currently, there is no single technology proven to be able to measure carcase OCM characteristics while also being able to enhance or provide a platform for automation, particularly for beef processing. Being able to utilise a single x-ray system as a platform for Automation and OCM will provide a major step in sensing automation for red meat processing. Potential benefits of successfully advancing the use of pre-developed technology for OCM includes: Utilisation of common technology for OCM and automated cutting (cost, footprint, enhanced return on investment)
Ensuring maximum meat and economic yield from each and every carcase The ability to better meet customer/market requirements Automated grading and carcase assessment The ability to influence livestock quality and price Beef Rib Cutting using existing x-ray system installed for automated cutting production An additional expected outcome of successful OCM trials is to enable and assist future development strategies for process automation with a view of establishing a considered strategy for future R&D project investment. Existing and new automated cutting systems developments would benefit immediately if successful by implementing OCM together with automated cut line detection in a single x-ray system. A path to industry adoption could be tested on existing SCOTT developments. 2 Project Objectives An Automated Beef Rib Cutting system has been installed and is in production at an Australian beef-processing facility.The system contains two x-ray tubes and two separate x- ray detectors located adjacent to each other on a conveyor system.The principal role of this system is to meet the imaging requirements of the automated rib cutting system utilised by this plant.However,its potential for determining carcase composition requires investigation. The project will provide the following outcome: Confirm whether the dual energy x-ray technology used in SCOTT's Beef Rib Cutting Project can be used to provide Objective Carcase Measurement of Bone,Fat and Muscle in beef primal cuts and portions. A Final Report including videos,images,results and outlining challenges and success in achieving project goals and outlining any future development steps to be submitted to MLA for review and approval. Page 7 of 44
Page 7 of 44 Ensuring maximum meat and economic yield from each and every carcase The ability to better meet customer/market requirements Automated grading and carcase assessment The ability to influence livestock quality and price Beef Rib Cutting using existing x-ray system installed for automated cutting production An additional expected outcome of successful OCM trials is to enable and assist future development strategies for process automation with a view of establishing a considered strategy for future R&D project investment. Existing and new automated cutting systems developments would benefit immediately if successful by implementing OCM together with automated cut line detection in a single x-ray system. A path to industry adoption could be tested on existing SCOTT developments. 2 Project Objectives An Automated Beef Rib Cutting system has been installed and is in production at an Australian beef-processing facility. The system contains two x-ray tubes and two separate xray detectors located adjacent to each other on a conveyor system. The principal role of this system is to meet the imaging requirements of the automated rib cutting system utilised by this plant. However, its potential for determining carcase composition requires investigation. The project will provide the following outcome: Confirm whether the dual energy x-ray technology used in SCOTT’s Beef Rib Cutting Project can be used to provide Objective Carcase Measurement of Bone, Fat and Muscle in beef primal cuts and portions. A Final Report including videos, images, results and outlining challenges and success in achieving project goals and outlining any future development steps to be submitted to MLA for review and approval
3 Methodology 3.1 DEXA Scans of Tissue Phantoms Samples of lean and fat tissue were sourced from lamb carcases and used to create mixtures of the following fat:muscle ratio's:0:100,25:75,50:50,75:25,or 100:0.These samples were then ground and homogenised,after which subsamples were taken for the determination of chemical fat and lean percentage,and percent dry matter,as reported in Error!Reference source not found.below Table 1.Dry matter,chemical fat and chemical lean percentage of mixtures of fat and lean. Fat:Lean ratio Percent Dry Matter Chemical Fat Chemical Lean 100:0 26.6 88.0 12.0 75:25 36.3 60.6 39.4 50:50 53.4 40.0 60.0 25:75 70.2 18.3 81.7 0:100 91.4 6.2 93.8 These mixtures were then used to create calibration blocks of 3 different uniform sizes using custom built moulds which were 10mm,80mm,or 160mm thick.Thus 3 calibration blocks were created for each of the 5 fat:lean mixtures,with thicknesses of 10mm,80mm,or 160mm.X-Ray images were then generated of the phantoms.This entire process was repeated 3 times using 3 sets of 3 calibration blocks. Prior to carrying out image analysis,sections within each image were selected which corresponded to the calibration tissue.The corresponding pixels within the low and high energy images were then used to calculate an R-value for these pixels according to the following formula: (R=In(ILow/AirAtten)/In(IHigh/AirAtten)); Where: lLow represents the pixel value in the low energy image(ZnSe Photodiode) h represents the pixel value in the high energy image(Csl Photodiode) AirAen represents the pixel value corresponding to the un-attenuated photons(lo)in the white part of each image. Equation 1-R-value calculation The R-values for the pixels of each calibration block were then averaged to give a single R Value representing that block.This data was represented graphically relative to the corresponding chemical fat for that block. Page 8 of 44
Page 8 of 44 3 Methodology 3.1 DEXA Scans of Tissue Phantoms Samples of lean and fat tissue were sourced from lamb carcases and used to create mixtures of the following fat:muscle ratio’s: 0:100, 25:75, 50:50, 75:25, or 100:0. These samples were then ground and homogenised, after which subsamples were taken for the determination of chemical fat and lean percentage, and percent dry matter, as reported in Error! Reference source not found. below. Table 1. Dry matter, chemical fat and chemical lean percentage of mixtures of fat and lean. Fat:Lean ratio Percent Dry Matter Chemical Fat % Chemical Lean % 100:0 26.6 88.0 12.0 75:25 36.3 60.6 39.4 50:50 53.4 40.0 60.0 25:75 70.2 18.3 81.7 0:100 91.4 6.2 93.8 These mixtures were then used to create calibration blocks of 3 different uniform sizes using custom built moulds which were 10mm, 80mm, or 160mm thick. Thus 3 calibration blocks were created for each of the 5 fat:lean mixtures, with thicknesses of 10mm, 80mm, or 160mm. X-Ray images were then generated of the phantoms. This entire process was repeated 3 times using 3 sets of 3 calibration blocks. Prior to carrying out image analysis, sections within each image were selected which corresponded to the calibration tissue. The corresponding pixels within the low and high energy images were then used to calculate an R-value for these pixels according to the following formula: (R = ln(ILow/AirAtten) / ln(IHigh/AirAtten)); Where: ILow represents the pixel value in the low energy image (ZnSe Photodiode) IHigh represents the pixel value in the high energy image (CsI Photodiode) AirAtten represents the pixel value corresponding to the un-attenuated photons (I0) in the white part of each image. Equation 1 - R-value calculation The R-values for the pixels of each calibration block were then averaged to give a single R Value representing that block. This data was represented graphically relative to the corresponding chemical fat % for that block
Figure 1:Dynamic DEXA scanning-Carcass phantom mid scan Figure 1 above is a snapshot of the carcass phantom as it is being scanned.The five blocks that can be seen are each various combinations of meat/fat/bone compositions and thicknesses.Each block had to be rearranged from various"slides"of meat/fat/bone for every scan. Figure 2 below depicts partially processed images of two different"carcass phantom" configurations.It can be seen that each tile is a slightly different shade and this corresponds to the fat,meat and bone composition of each tile. Figure 2:Dynamic DEXA scans of two carcass phantoms 3.2 DEXA and CT scanning of six beef sides The DEXA hardware at the beef processing plant was used to capture dual energy images of 6 beef half-carcases,scanned in 2 batches of 4(batch 1),and 2 (batch 2)carcases,with each batch collected on separate days.The DEXA hardware consisted of two X-ray tubes (one operating at high voltage and one operating at low voltage)and two GADOX photodiodes located at separate points along a conveyor used to maintain carcass orientation.These two x-ray tube/detector combinations produce the high and low energy images which are then used to calculate an R-value for these pixels according to Equation 1. The average R-value for all of the pixels in the carcase image was calculated,and the image was then reconstructed after removing any pixels with R-values lying above this mean R- value.Pixel R-values were then converted to proportion lean tissue and weighted based on thickness using the equations derived in the section above,and then averaged to reflect an average R-value for the whole carcase.These carcase R-values were then used to predict CT lean%,fat%,and bone%measured on these same carcases. Page 9 of 44
Page 9 of 44 Figure 1: Dynamic DEXA scanning - Carcass phantom mid scan Figure 1 above is a snapshot of the carcass phantom as it is being scanned. The five blocks that can be seen are each various combinations of meat/fat/bone compositions and thicknesses. Each block had to be rearranged from various “slides” of meat/fat/bone for every scan. Figure 2 below depicts partially processed images of two different “carcass phantom” configurations. It can be seen that each tile is a slightly different shade and this corresponds to the fat, meat and bone composition of each tile. Figure 2: Dynamic DEXA scans of two carcass phantoms 3.2 DEXA and CT scanning of six beef sides The DEXA hardware at the beef processing plant was used to capture dual energy images of 6 beef half- carcases, scanned in 2 batches of 4 (batch 1), and 2 (batch 2) carcases, with each batch collected on separate days. The DEXA hardware consisted of two X-ray tubes (one operating at high voltage and one operating at low voltage) and two GADOX photodiodes located at separate points along a conveyor used to maintain carcass orientation. These two x-ray tube/detector combinations produce the high and low energy images which are then used to calculate an R-value for these pixels according to Equation 1. The average R-value for all of the pixels in the carcase image was calculated, and the image was then reconstructed after removing any pixels with R-values lying above this mean Rvalue. Pixel R-values were then converted to proportion lean tissue and weighted based on thickness using the equations derived in the section above, and then averaged to reflect an average R-value for the whole carcase. These carcase R-values were then used to predict CT lean%, fat%, and bone% measured on these same carcases
Figure 3:DEXA scans of two carcass sides The beef carcasses that were DEXA scanned were broken down into primals,vacuum packed and sent to a CT scanner.The CT scanning of 53 cartons of product to determine lean meat,fat and bone composition and distribution was successfully completed.Full bone out was conducted so that manual objective measurements could be taken. Figure 4 shows a series of images of the vacuum packed"primals"being scanned. easurement ning System Page 10 of 44
Page 10 of 44 Figure 3: DEXA scans of two carcass sides The beef carcasses that were DEXA scanned were broken down into primals, vacuum packed and sent to a CT scanner. The CT scanning of 53 cartons of product to determine lean meat, fat and bone composition and distribution was successfully completed. Full bone out was conducted so that manual objective measurements could be taken. Figure 4 shows a series of images of the vacuum packed “primals” being scanned