Detection of Differences in Longitudinal Cartilage Thickness Loss Using a Deep-Learning Automated Segmentation Algorithm: Data From the Foundation for the National Institutes of Health Biomarkers Study of the Osteoarthritis Initiative.

OBJECTIVE: To study the longitudinal performance of fully automated cartilage segmentation in knees with radiographic osteoarthritis (OA), we evaluated the sensitivity to change in progressor knees from the Foundation for the National Institutes of Health OA Biomarkers Consortium between the automated and previously reported manual expert segmentation, and we determined whether differences in progression rates between predefined cohorts can be detected by the fully automated approach. METHODS: The OA Initiative Biomarker Consortium was a nested case-control study. Progressor knees had both medial tibiofemoral radiographic joint space width loss (≥0.7 mm) and a persistent increase in Western Ontario and McMaster Universities Osteoarthritis Index pain scores (≥9 on a 0-100 scale) after 2 years from baseline (n = 194), whereas non-progressor knees did not have either of both (n = 200). Deep-learning automated algorithms trained on radiographic OA knees or knees of a healthy reference cohort (HRC) were used to automatically segment medial femorotibial compartment (MFTC) and lateral femorotibial cartilage on baseline and 2-year follow-up magnetic resonance imaging. Findings were compared with previously published manual expert segmentation. RESULTS: The mean ± SD MFTC cartilage loss in the progressor cohort was -181 ± 245 µm by manual segmentation (standardized response mean [SRM] -0.74), -144 ± 200 µm by the radiographic OA-based model (SRM -0.72), and -69 ± 231 µm by HRC-based model segmentation (SRM -0.30). Cohen's d for rates of progression between progressor versus the non-progressor cohort was -0.84 (P < 0.001) for manual, -0.68 (P < 0.001) for the automated radiographic OA model, and -0.14 (P = 0.18) for automated HRC model segmentation. CONCLUSION: A fully automated deep-learning segmentation approach not only displays similar sensitivity to change of longitudinal cartilage thickness loss in knee OA as did manual expert segmentation but also effectively differentiates longitudinal rates of loss of cartilage thickness between cohorts with different progression profiles.

Association of Superficial Cartilage Transverse Relaxation Time With Osteoarthritis Disease Progression: Data From the Foundation for the National Institutes of Health Biomarker Study of the Osteoarthritis Initiative.

OBJECTIVE: To study whether layer-specific cartilage transverse relaxation time (T2) and/or longitudinal change is associated with clinically relevant knee osteoarthritis (OA) disease progression. METHODS: The Foundation for the National Institutes of Health Biomarker Consortium was a nested case-control study on 600 knees from 600 Osteoarthritis Initiative participants. Progressor knees had both medial tibiofemoral radiographic joint space width (JSW) loss (≥0.7 mm) and a persistent increase in Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain score (≥9 on a 0-100 scale) at 24-48 months from baseline (n = 194). Multiecho spin-echo (MESE) magnetic resonance images (MRIs) for cartilage T2 analysis had been acquired in the right knees only (97 progressor knees). These were compared to 104 control knees without JSW or pain progression. Fifty-three knees had JSW progression, and 57 pain progression only. Cartilage thickness segmentations obtained from double-echo steady-state MRI were matched to MESE MRI to extract superficial and deep femorotibial cartilage T2. Superficial medial femorotibial compartment (MFTC) T2 at baseline was the primary, and change in deep MFTC T2 between baseline and 12 months was the secondary analytic outcome of this post hoc exploratory study. RESULTS: Baseline superficial MFTC T2 was significantly elevated in progressor knees (adjusted mean 47.2 msec [95% confidence interval (95% CI) 46.5, 48.0]) and JSW progression only knees (adjusted mean 47.3 msec [95% CI 46.3, 48.3]), respectively, versus non-progressor knees (45.8 msec [95% CI 45.0, 46.5]) after adjustment for age, sex, body mass index, WOMAC pain score, and medial joint space narrowing grade (analysis of covariance). Change in T2 was not significantly associated with case status. CONCLUSION: Baseline superficial, but not deep, medial cartilage T2 is associated with clinically relevant disease progression in knee OA.

Local MRI-based measures of thigh adipose tissue derived from fully automated deep convolutional neural network-based segmentation show a comparable responsiveness to bidirectional change in body weight as from quality controlled manual segmentation.

BACKGROUND: Thigh intermuscular (IMF) and subcutaneous (SCF) fat are associated with joint function, inflammation and knee osteoarthritis. Fully automated segmentation from MRI is important to study the above relationship in larger cohorts. However, such algorithms are not clinically evaluated for longitudinal studies. Our aim was to evaluate a fully automated U-Net segmentation approach and its ability to detect longitudinal changes in thigh IMF and SCF during weight changes compared to manual segmentation. METHODS: 103 Osteoarthritis Initiative subjects, were studied, 52 with> 10% weight loss, and 51 with> 10% weight gain over 2-years. Longitudinal change in IMF and SCF were determined from baseline and year-2 axial thigh MRIs using U-Net segmentation. The standardised response mean (SRM) was used as measure of sensitivity to change. RESULTS: The U-Net took substantially less time (single-slice MRI:< 1 s) and IMF and SCF showed very similar sensitivity to change as manual segmentation: With an average weight gain of + 14%, we observed an + 12% /+ 26% increase in IMF / SCF (SRM=0.99 /1.03) using the U-Net, compared with + 21% /+ 27% (SRM=0.60 /1.07) for manual segmentation. During an average weight loss of - 18%, we observed an - 14% /- 22% reduction in IMF /SCF (SRM = - 1.04 /-1.20) using the U-Net, compared with - 16% /- 22% (SRM = - 0.70 /-1.23) for manual segmentation. CONCLUSION: U-Net segmentation replicates longitudinal changes of IMF and SCF associated with weight changes with a similar sensitivity to change as manual segmentation. This method is applicable to large databases for studying relationships between IMF and SCF and various disease conditions.

Longitudinal Change in Knee Cartilage Thickness and Function in Subjects with and without MRI-Diagnosed Cartilage Damage.

OBJECTIVE: Cartilage damage diagnosed by magnetic resonance imaging (MRI) is highly prevalent in the population. In this article, we explore whether such cartilage damage is associated with greater longitudinal change in 3D cartilage thickness and knee function in subjects without (risk factors of) knee osteoarthritis. DESIGN: Eighty-two knees of Osteoarthritis Initiative healthy reference cohort participants had baseline and 4-year follow-up MRI and knee function data. Baseline presence of semiquantitatively assessed MRI-based cartilage damage (MOAKS [MRI Osteoarthritis Knee Score] ≥ grade 1.0) was recorded by an experienced radiologist. Longitudinal femorotibial cartilage thickness change was determined after segmentation, using location-independent methodology. Knee function was evaluated by patient-reported outcomes and functional performance measures. Statistical comparisons included analysis of covariance adjusting for age, sex, and body mass index. RESULTS: Forty-five percent of the participants had cartilage damage in at least one femorotibial subregion; the cartilage thickness change score was 15% greater in participants with than in those without damage (1216 ± 434 vs. 1058 ± 277 µm). This difference reached borderline statistical significance with and without adjustment for age, sex, and body mass index (P = 0.05). No significant differences in the change of patient-reported outcomes of knee function (PASE [physical activity score of the elderly] and WOMAC [Western Ontario McMaster Osteoarthritis Index]) or chair stand test results were detected. Of those without femorotibial damage, 58% had cartilage damage in at least one femoropatellar subregion; these had a 9% greater femorotibial cartilage change score than those without femoropatellar or femorotibial damage (difference not statistically significant). CONCLUSIONS: In the absence of osteoarthritis risk factors, semiquantitatively assessed MRI-based cartilage damage appears to be associated with greater longitudinal location-independent femorotibial cartilage thickness changes, but not with greater functional deteriorations.

Accuracy and longitudinal reproducibility of quantitative femorotibial cartilage measures derived from automated U-Net-based segmentation of two different MRI contrasts: data from the osteoarthritis initiative healthy reference cohort.

OBJECTIVE: To evaluate the agreement, accuracy, and longitudinal reproducibility of quantitative cartilage morphometry from 2D U-Net-based automated segmentations for 3T coronal fast low angle shot (corFLASH) and sagittal double echo at steady-state (sagDESS) MRI. METHODS: 2D U-Nets were trained using manual, quality-controlled femorotibial cartilage segmentations available for 92 Osteoarthritis Initiative healthy reference cohort participants from both corFLASH and sagDESS (n = 50/21/21 training/validation/test-set). Cartilage morphometry was computed from automated and manual segmentations for knees from the test-set. Agreement and accuracy were evaluated from baseline visits (dice similarity coefficient: DSC, correlation analysis, systematic offset). The longitudinal reproducibility was assessed from year-1 and -2 follow-up visits (root-mean-squared coefficient of variation, RMSCV%). RESULTS: Automated segmentations showed high agreement (DSC 0.89-0.92) and high correlations (r ≥ 0.92) with manual ground truth for both corFLASH and sagDESS and only small systematic offsets (≤ 10.1%). The automated measurements showed a similar test-retest reproducibility over 1 year (RMSCV% 1.0-4.5%) as manual measurements (RMSCV% 0.5-2.5%). DISCUSSION: The 2D U-Net-based automated segmentation method yielded high agreement compared with manual segmentation and also demonstrated high accuracy and longitudinal test-retest reproducibility for morphometric analysis of articular cartilage derived from it, using both corFLASH and sagDESS.

Development of artificial tissue-like structures for a hybrid epidural anesthesia simulator.

Puncturing the epidural space and lumbar puncture are common procedures in anesthesia. They are carried out blind, where a needle is advanced from posterior between two adjacent vertebrae. Two different approaches are common practice for this technique, the midline and the paramedian one. The learning curve characteristics of both approaches significantly depends on the number of punctures carried out by a medical novice. For the training of these blind procedures a hybrid simulator requires artificial structures imitating the tissues which are penetrated by the needle. Within this work a patient phantom for spinal needle insertion procedures was developed and validated successfully against literature as well as by a study carried out with medical experts.

Inexpensive bone cement substitute for vertebral cement augmentation training.

Vertebral compression fractures are treated surgically for approximately 25 years. In percutaneous cement augmentation techniques bone cement is applied to a fractured vertebra under fluoroscopic evidence to stabilize the bone fragments. Complications due to leakage of the low viscosity bone cement are reported in 5 to 15% of all routine cases. During the intraoperative application of bone cement surgeons rely on visiohaptic feedback and hence need to be familiar with the cement's rheology properties. Therefore, training is necessary. A hybrid simulator for cement augmentation training was developed but the usage of expensive real cement limits its purpose as a training modality. Twentythree inexpensive bone substitutes were developed and tested with the objective to mimic real bone cement. Cement application measurements were conducted and a mathematical model of the measurement setup was created. Compared with real bone cement, a cement substitute based on Technovit 3040 in combination with radical catchers and additional additives was identified as an appropriate substitute for cement augmentation training.

Development of trabecular bone surrogates for kyphoplasty-balloon dilatation training.

Vertebral compression fractures can limit quality of life. Cement augmentation techniques show good results in attaining pain relief. Kyphoplasty enables a better restoration of vertebra height due to a dilatable balloon tamp, which is inflated in the fractured vertebra. Surgical training of vertebral cement augmentation techniques is currently performed on patients or specimens. To enable another training possibility for surgical residents, a new hybrid patient simulator was developed. Artificial vertebrae allocate a realistic haptic feedback during needle insertion. Based on these results, new polyurethane foam recipes were developed to either enable a realistic needle insertion as well as a balloon tamp dilatation. Needle insertion forces of the newly developed foams were compared against commercially available artificial trabecular bone material and balloon tamp dilatations were performed in manufactured materials. Based on the matching needle insertion forces, two suitable material compositions for needle insertion and balloon dilatation training were found. This investigation is considered as a prior study before evaluation on human specimen.

Assessment parameters for a novel simulator in minimally invasive spine surgery.

Surgical simulators provide a safe environment where novice surgeons can acquire their surgical skills. Although the number of patients with diseases of the musculoskeletal system is growing, the development of orthopedic simulators is still in it's infancy. The aim of this work was to identify simulation-based assessment parameters for a novel simulator in minimally invasive spine surgery. Apart from parameters targeting the duration and the surgeon's economy of motion during percutaneous bone access, parameters characterizing the movement smoothness were also examined with respect to their suitability. The results indicated, that the overall duration, the number of instrument movements, the number of velocity peaks and the Movement Arrest Period Ratio are the most promising predictors of expertise. Targeting performance improvement, the overall duration (p = 0.001), the number of instrument movements (p = 0.003) and the traveled instrument path length (p = 0.009) detected significant differences between subsequent trials. Using these parameters, a study can be designed targeting the validity and reliability of the simulation-based assessment.

Artificial muscles for a novel simulator in minimally invasive spine surgery.

Vertebroplasty and kyphoplasty are commonly used minimally invasive methods to treat vertebral compression fractures. Novice surgeons gather surgical skills in different ways, mainly by "learning by doing" or training on models, specimens or simulators. Currently, a new training modality, an augmented reality simulator for minimally invasive spine surgeries, is going to be developed. An important step in investigating this simulator is the accurate establishment of artificial tissues. Especially vertebrae and muscles, reproducing a comparable haptical feedback during tool insertion, are necessary. Two artificial tissues were developed to imitate natural muscle tissue. The axial insertion force was used as validation parameter. It appropriates the mechanical properties of artificial and natural muscles. Validation was performed on insertion measurement data from fifteen artificial muscle tissues compared to human muscles measurement data. Based on the resulting forces during needle insertion into human muscles, a suitable material composition for manufacturing artificial muscles was found.

Foam phantom development for artificial vertebrae used for surgical training.

Currently the surgical training of kyphoplasty and vertebroplasty is performed on patients or specimens. To improve patient safety, a project was initiated to develop an Augmented Reality simulator for the surgical training of these interventions. Artificial vertebral segments should be integrated to provide realistic haptic feedback. To reach this, resulting forces during needle insertions (trans- and extrapedicular) into formalin-fixed vertebral specimens were measured. The same insertion procedure was also performed on six customized polyurethane blocks with varying mechanical parameters. Based on the results of these measurements, a specific foam phantom was generated and the insertion force measured. Additionally a parametric model for the needle insertion into bone was designed calculating three characteristic parameters for all insertion measurements. The resulting insertion force for the foam phantom was comparable to the specimen measurements and the parametric model provided comprehensible characteristic parameters. Based on the resulting force during needle insertion into human vertebrae, a possible foam recipe for manufacturing artificial segments was found. Furthermore, the parametric model provides characteristic parameters for the assessment of phantoms as well as the development of its production process.