The Experts below are selected from a list of 294 Experts worldwide ranked by ideXlab platform
Lubomir M Hadjiiski - One of the best experts on this subject based on the ideXlab platform.
-
automated pectoral muscle identification on mlo view mammograms comparison of deep neural network to Conventional Computer vision
Medical Physics, 2019Co-Authors: Chuan Zhou, Heangping Chan, Lubomir M Hadjiiski, Mark A Helvie, Yao LuAbstract:OBJECTIVES: The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using Conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. MATERIALS AND METHODS: Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. RESULTS: In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. CONCLUSION: In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05).
-
Automated pectoral muscle identification on MLO-view mammograms: Comparison of deep neural network to Conventional Computer vision.
Medical physics, 2019Co-Authors: Jun Wei, Heangping Chan, Mark A Helvie, Chuan Zhou, Lubomir M HadjiiskiAbstract:The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using Conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05). © 2019 American Association of Physicists in Medicine.
Johannes Beckmann - One of the best experts on this subject based on the ideXlab platform.
-
no difference in accuracy between pinless and Conventional Computer assisted surgery in total knee arthroplasty
Knee Surgery Sports Traumatology Arthroscopy, 2014Co-Authors: Clemens Baier, G. Maderbacher, H.r. Springorum, Florian Zeman, Wolfgang Fitz, Jens Schaumburger, Joachim Grifka, Johannes BeckmannAbstract:Purpose Many studies have demonstrated higher precision and better radiological results in Total knee arthroplasty (TKA) with Computer-assisted surgery (CAS). On the other hand, studies revealed a lengthening of operation time up to 20 min for this technique and demonstrated rare additional complications as fractures and neurovascular injuries caused by the array pins and any intraoperative array dislocation leads to abortion of CAS. To combine the advantages and eliminate the disadvantages of standard CAS, we evaluated the accuracy of a so-called pinless CT-free version of knee navigation (pinless CAS) abandoning the reference pins and reducing the necessary workflow to a minimum.
-
No difference in accuracy between pinless and Conventional Computer-assisted surgery in total knee arthroplasty
Knee surgery sports traumatology arthroscopy : official journal of the ESSKA, 2013Co-Authors: Clemens Baier, G. Maderbacher, H.r. Springorum, Florian Zeman, Wolfgang Fitz, Jens Schaumburger, Joachim Grifka, Johannes BeckmannAbstract:Many studies have demonstrated higher precision and better radiological results in Total knee arthroplasty (TKA) with Computer-assisted surgery (CAS). On the other hand, studies revealed a lengthening of operation time up to 20 min for this technique and demonstrated rare additional complications as fractures and neurovascular injuries caused by the array pins and any intraoperative array dislocation leads to abortion of CAS. To combine the advantages and eliminate the disadvantages of standard CAS, we evaluated the accuracy of a so-called pinless CT-free version of knee navigation (pinless CAS) abandoning the reference pins and reducing the necessary workflow to a minimum. The present study compares the accuracy of the reference methods of two different CT-free knee navigation software versions (Brainlab Knee 2.1 and Brainlab Knee Express 2.5). Thirty patients received TKA assisted by standard CAS. Intraoperatively, the proposed bony resections of standard CAS were matched with the new pinless CAS. Postoperatively, the results were checked by evaluating the radiographs concerning leg axis, femoral flexion and tibial slope. All results concerning precise cuts (femoral as well as tibial coronal/varus-valgus alignment, femoral flexion alignment and tibial slope, resection height) were comparable between both groups (n.s.). In femoral, we found a mean deviation of coronal alignment of 0.3° (SD 0.7) and flexion of 0.2° (SD 0.8). In tibial, we found a mean deviation of coronal alignment of 0.2° (SD 0.5) and slope of 0.2° (SD 0.6). The mean additional operation time for the pinless CAS was below 2 min. The postoperative mechanical leg axis was within the threshold of 3° in all patients, tibial slope and femoral flexion matched with CAS values. In clinical routine, pinless CAS can comprise the advantages of CAS leaving the disadvantages aside. It reduces surgical time and avoids complications associated with the tracking pins of Conventional CAS.
Chuan Zhou - One of the best experts on this subject based on the ideXlab platform.
-
automated pectoral muscle identification on mlo view mammograms comparison of deep neural network to Conventional Computer vision
Medical Physics, 2019Co-Authors: Chuan Zhou, Heangping Chan, Lubomir M Hadjiiski, Mark A Helvie, Yao LuAbstract:OBJECTIVES: The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using Conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. MATERIALS AND METHODS: Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. RESULTS: In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. CONCLUSION: In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05).
-
Automated pectoral muscle identification on MLO-view mammograms: Comparison of deep neural network to Conventional Computer vision.
Medical physics, 2019Co-Authors: Jun Wei, Heangping Chan, Mark A Helvie, Chuan Zhou, Lubomir M HadjiiskiAbstract:The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using Conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05). © 2019 American Association of Physicists in Medicine.
Matthew Brown - One of the best experts on this subject based on the ideXlab platform.
-
Initial Outcomes of Total Knee Arthroplasty Comparing Conventional, Computer Assisted Navigation, and Custom Cutting Block Techniques.
Journal of Bone and Joint Surgery-british Volume, 2013Co-Authors: Christopher Drinkwater, Wes Madsen, Matthew BrownAbstract:Hypothesis Custom cutting blocks can produce similar alignment compared to Computer navigated and Conventional total knee arthroplasty (TKA) techniques. Method We conducted a retrospective review of 37 patients who underwent TKA by a single surgeon in a teaching hospital setting. Groups were Conventional method (10), Computer assisted navigation (10), and custom blocks (18). The custom group was further subdivided to CT and MRI based blocks. Post-operative alignment was measured (blinded) using full length weight bearing radiographs at 18 weeks on average. Hospital records were reviewed to determine operative time, transfusion requirements, length of hospital stay, complications and cost. Results Post-operative mechanical axis was within 3 degrees of neutral in 100% of the navigation group, 70% of the Conventional group and 50% of the custom block group. Average alignment was within 1.8, 3.1 and 3.6 degrees of neutral for each group respectively. The operative time was greater for the Computer navigation group (86.7 min) compared to the Conventional (72.1) and MRI custom block groups which involved unfamiliar instrumentation (73.8). CT based block procedures involved otherwise familiar instruments and averaged 61.2 minutes. Length of hospital stay and complications were similar for all three groups. Total cost was the least for the Conventional group. Increased costs were associated with Computer equipment, pre-operative advanced imaging and custom blocks. Conclusions Custom cutting blocks in this small series obtain worse radiographic positioning of total knee arthroplasty components compared to Conventional and Computer navigation techniques. Further studies with greater number of patients, CT alignment analysis and long-term follow-up are required.
Yao Lu - One of the best experts on this subject based on the ideXlab platform.
-
automated pectoral muscle identification on mlo view mammograms comparison of deep neural network to Conventional Computer vision
Medical Physics, 2019Co-Authors: Chuan Zhou, Heangping Chan, Lubomir M Hadjiiski, Mark A Helvie, Yao LuAbstract:OBJECTIVES: The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using Conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation. MATERIALS AND METHODS: Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline. RESULTS: In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method. CONCLUSION: In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05).