Automated heart segmentation using a convolutional neural network accelerates 3D model creation for cardiac surgery

Funding Acknowledgements

Type of funding sources: Private grant(s) and/or Sponsorship. Main funding source(s): Japan Organization of Occupational Health and Safety

onbehalf

J3D

Introduction

Improvements in cardiovascular imaging have contributed to the accurate diagnosis and treatment of cardiovascular diseases. In recent years, we can find a new demand from surgical departments, i.e., 3D heart models of patients for preoperative simulations. However, a 3D surgical model needs a detailed segmentation of heart structures from clinical images because these details are essential for achieving accurate polygon data of an exact 3D model. Thus, in most cases, manual segmentation is required for complicated heart shapes such as the chamber wall, valves, and papillary muscles, which causes a prolonged duration time.

Purpose

We aim to achieve an automated heart segmentation using a convolutional neural network (CNN) trained using deep learning techniques for the rapid creation of 3D surgical heart models of individual patients.

Methods

We constructed our original CNN program based on the latest artificial intelligence techniques and trained it to extract shapes of the heart from cardiac computed tomography (CT) images. The training data was 361 slices selected from CT scans of 10 patients. We used data augmentation to increase the amount and diversity of the data into 24,052 slices. The training result with the best Intersection over Union (IoU), one of the evaluation metrics used in deep learning, was saved. Finally, we used the best-trained CNN to construct 3D polygon data of two surgical cases of hypertrophic obstructive cardiomyopathy (HOCM) for preoperative assessments.

Results

The IoU attained 0.85 after deep learning. The time required to complete the 3D polygon data for the first HOCM case was 5 minutes for segmentation by the trained CNN and 3 hours for data correction by a human operator. Similarly, the time required for the second case was 5 minutes for segmentation without manual correction. We had to correct the segmentation for the first case because we needed an exact 3D model for the preoperative assessment (Fig. 1). According to our records of the other eight 3D heart models in the lab, the work for a 3D polygon shape from CT images needs a median 30 hours (quartiles 23-50 hours) when the procedure is fully manual and non-continuous with breaks in between.

Conclusion

The CNN-based segmentation aided the constructing heart shapes from cardiac CT images of preoperative patients. Although the performance, reaching IoU of 0.85, was insufficient for fully automatic segmentation, the methodology can shorten the process duration from several hours to several minutes for detailed segmentation of heart structures.

We previously applied the CNN-based segmentation to the aorta, aortic stenosis valves, and atheromatous plaque in clinical images, demonstrating adequate segmentation performance. The proposed methodology can be applied as a fundamental technology of cardiovascular imaging for obtaining the actual structures of a target object as 3D coordinate values or a 3D model within a reasonable duration time.

Abstract Figure 1

Autologous skeletal myoblast sheet prevents cardiomyocyte ischemia and right heart dysfunction in pressure-overloaded right heart porcine model

Introduction

Severe heart failure (HF) with congenital heart disease (CHD) have demonstrated life threatening disorder despite of remarkable progress in medical therapies. Autologous skeletal myoblast sheet transplantation therapy showed clinical efficacy for left ventricular dysfunction by cytokine paracrine effects, which are expected to be sufficiently effective against right ventricular (RV) dysfunction which is often seen in end-stage of CHD patients with severe HF.

Hypothesis

An autologous skeletal myoblast sheet transplantation alleviates RV dysfunction in a pressure-overloaded right heart in a porcine model.

Methods

Five-to-six-month-old Göttingen mini-pigs underwent pulmonary artery banding with vascular occluding system. To create the porcine model of chronic pressure-overloaded right heart, vascular occluding system was gradually inflated, over a month, to make pulmonary stenosis to banding velocity >3.0 m/s measured by echocardiography (UCG), and then fixed for another month. Two months after banding, autologous skeletal myoblast sheet was placed on the epicardium of the RV free wall and followed for 2 months. Groups were as follows: control (C, n=5), sheet implantation (S, n=5). Cardiac function was measured using UCG, cardiac computed tomography (CT), and cardiac catheterization (Cath). Two months after sheet implantation, hearts were dissected for histologic analysis.

Results

Before sheet implantation, RV dysfunction was equal in groups; however, 2 months after sheet implantation, RV dysfunction and myocardial ischemia was significantly ameliorated in group S than group C. On CT, RV ejection fraction exacerbation were well controlled in Group S compared to Group C (S 44.9±2.2 vs C 31.9±2.1% [p=0.0042]). UCG and Cath revealed well maintained systolic and diastolic function in Group S compared to Group C (Tei index: S 0.42±0.06 vs C 0.70±0.07 [p=0.0240], Fraction Area Change: S 45.8±7.8 vs C 19.5±1.3% [p=0.0240], Isovolumic Relaxation Time; S 44.3±9.2 vs C 97.3±9.5 ms [p=0.0304]). On C11-Acetate Positron Emission Tomography, myocardial ischemia was more prominent in Group C compared to Group S (K mono-Rest/Stress: S 3.17±0.69 vs C 2.03±0.65 min-1 [p=0.0421], Myocardial Blood Flow-Rest/Stress: S 3.22±0.39 vs C 2.13±0.92 min-1 [p=0.0421]). In histologic analysis, Group S presented less progressed hypertrophic change in periodic acid-Schiff stain (S 13.5±0.9 vs C 18.0±3.0 μg [p=0.0240]), anti-fibrotic changes in picrosirius red stain (S 3.0±0.3 vs C 4.2±0.2% [p=0.0421]), more angiogenesis in CD31 expression (S 18.3±1.5 vs C 10.7±2.8 / 104 μm2 [p=0.0240]), and less production of reactive oxygen species in fluorescent immunostaining (S 5.9±1.7 vs C 18.4±1.7% [p=0.0304]).

Conclusion

Autologous skeletal myoblast sheet transplantation alleviates cardiomyocyte Ischemia and RV dysfunction in a porcine model of pressure-overloaded right heart.

Funding Acknowledgement

Type of funding source: None

Serial evaluation of cardiac allograft vasculopathy after heart transplantation by dual-modality intravascular ultrasound and optical coherence tomography

Background

Cardiac allograft vasculopathy is a combination of the development of de novo plaque and the progression of donor-transmitted plaque.

Purpose

This study aimed to evaluate the development of de novo plaque and the progression of donor-transmitted plaque within 12-month after heart transplantation (HTx) using serial intravascular ultrasound (IVUS) and optical coherence tomography (OCT). The association between inflammatory cytokines and plaque progression was also examined.

Methods

We prospectively enrolled 40 recipients to conduct serial three-vessel IVUS and OCT analysis at 8-week and 12-month after HTx. De novo plaque was defined as having maximum intimal thickness (MIT) ≥0.5 mm at 12-month in the absence of donor-transmitted plaques (MIT ≥0.5 mm at 8-week). Serum cytokines were screened with a bead-based multiplex assay.

Results

A total of 13 de novo plaques (fibrous, n=10; fibroatheroma, n=3) were detected in eight recipients. Serum interleukin (IL)-31 at 8-week was associated with the development of de novo plaques (p=0.009). A total of 31 donor-transmitted plaques (fibrous, n=12; fibroatheroma, n=11; fibrocalcific, n=8) were detected in 17 recipients. Multiple regression analysis revealed that fibrous (p=0.026) and fibroatheroma (p=0.012) observed at 8-week were significantly associated with subsequent plaque progression within 12-month after HTx. ΔPlaque burden was significant higher in de novo plaque than donor-transmitted plaque (38.8% [29.6–41.2] versus 8.7% [1.3–13.6], p<0.001). The prevalence of macrophage accumulation was lower in de novo plaque than in donor-transmitted plaque (8% versus 52%, p=0.006). Serum IL-31 at 8-week was correlated with the progression of donor-transmitted plaque as well as de novo plaque (r=0.663, p=0.029) although other cytokines like IL-1β, IL-6, IL-17, and tissue necrotic factor alpha were not.

Conclusions

In de novo plaques, fibrous plaque was the most common and macrophage accumulation was rarely observed. In donor-transmitted plaque, fibrous and fibroatheroma were independent predictor for the subsequent plaque progression. Serum interleukin-31 surge at subacute phase may play pathogenic role in cardiac allograft vasculopathy.

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): Grants-in-Aid for Scientific Research (KAKENHI)

Functional engineered heart tissue cultured in a rotating wall vessel bioreactor improve cardiac function in the distressed rat heart

Introduction

How to construct massive cardiac tissue and culture it with functional improvement may be crucial as cardiomyogenesis in failed heart. We previously presented that dynamic culture in a rotating wall vessel (RWV) bioreactor could provide a better culture environment for maintenance of the engineered 3D cardiac tissue. However, it is unknown about the effect of the tissue cultured in a RWV bioreactor on engraftment and improvement of function in the distressed rat heart.

Hypothesis

We hypothesized that the engineered 3D cardiac tissue cultured in a RWV bioreactor could improve its engraftment and lead recovery of cardiac function in rat infarction model.

Methods

We made engineered cardiac tissue by seeding 2.0 × 106 human induced pluripotent stem cell derived cardiomyocytes on the PLGA fiber sheet. It was cultured in the RWV bioreactor for seven days (RWV group). For the control, static culture has been done. After in vitro assessment, these tissues were transplanted to myocardial infarction model nude rats (sham, control, and RWV group, n=10, respectively) and cardiac performance was evaluated by ultrasonography. Four weeks after transplantation, we evaluated their hearts by histological analysis.

Results

The RWV group demonstrated maturation of cardiomyocytes evidenced by significantly higher expression of Troponin T (TnT), sarcomeric α actinin (SAA), connexin 43 (Cx43) and myosin heavy chain 7 (MYH7) than the control by Western blots (TnT; 2.7±1.0 vs. 1.0±0.4, p<0.01, SAA; 2.1±0.7 vs. 1.0±0.2, p<0.01, Cx 43; 2.0±0.6 vs. 1.0±0.1, p<0.05, MYH7; 10.9±2.7 vs. 1.0±0.1, p<0.01). In the culture supernatant, the concentration of cytokines related to angiogenesis was significantly higher in the RWV group than in the control (VEGF; 29.6±7.4 vs. 12.2±4.3pg/ml, p<0.01, HGF; 72.7±9.9 vs. 42.6±5.9pg/ml, p<0.01). Four weeks after transplantation, the left ventricular ejection fraction was significantly improved in the RWV group than in the control (RWV vs. control; 47±4.9 vs. 38±6.9%, p<0.01). On histological analysis, more engineered cardiac tissue survived in the RWV group than in the control (RWV vs. control; 7/10 vs. 3/10, p=0.18). A vascular-like structure double-stained with isolectin B4 and smooth muscle actin was partially observed in the transplanted tissue. LV remodeling exhibiting extracellular collagen deposition (fibrotic area, RWV vs. control; 17±4.3 vs. 24±5.2%, p<0.05) and cardiomyocyte hypertrophy (RWV vs. control; 16±1.7 vs. 18±2.1μm, p<0.05) was significantly attenuated in RWV group than in the control. Neovascularization was significantly noted in the RWV group compared with the control (capillary density, RWV vs. control; 545±113 vs. 356±92, p<0.01).

Conclusion

Functional engineered 3D cardiac tissue cultured in a RWV bioreactor could induce angiogenesis and improved its engraftment, leading significant improvement of cardiac function in rat infarction model.

Dynamic culture in a RWV bioreactor

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): Japan Society for the Promotion of Science

Occlusal support is associated with nutritional improvement and recovery of physical function in patients recovering from hip fracture

OBJECTIVE

This study aimed to elucidate the association between occlusal support and nutritional improvement and recovery of activities of daily living (ADLs) among elderly patients recovering from hip fracture.

BACKGROUND

Many patients with hip fracture are malnourished, and malnutrition is associated with poor functional outcomes. Poor oral status is one reason for malnutrition because loss of occlusal support leads to masticatory disorders and can cause nutrient deficiencies.

METHODS

We evaluated 202 elderly patients aged 65 years and older (mean age, 84.9 ± 7.9 years). We assessed nutritional status using the Mini Nutritional Assessment-Short Form and ADLs using functional independence measure (FIM) scores. Occlusal support was recorded in accordance with the Eichner Index. We categorised participants into two groups according to the presence or absence of occlusal support, and statistical analyses were performed to investigate the differences between the groups.

RESULTS

One group contained 152 participants (mean age, 85.5 ± 7.4 years) with occlusal support, and the other group contained 50 participants (mean age, 83.0 ± 8.9 years) without occlusal support. The group with occlusal support had greater gain of FIM and higher FIM efficiency than did the other group. Multivariate analyses showed that occlusal support was independently associated with nutritional improvement [odds ratio (OR) = 4.00, 95% confidence intervals (CI) = 1.90-8.43] and motor FIM efficiency (R²  = .338, P < .001).

CONCLUSIONS

Our findings suggest that occlusal support is associated with nutritional improvement and the recovery of ADLs in patients recovering from hip fracture.

P315In-vivo and vitro mitochondrial transfer from adipose-derived mesenchymal stem cell to ischemic cardiomyocyte

Background

Although transplantation of human Adipose-derived Mesenchymal stem cell (hADSC) shows efficacy in the treatment of ischemic cardiomyopathy, its therapeutic mechanisms have not been fully elucidated. It has been already reported that mitochondria transfer to recipient cells have impact on resistance to injury and tissue regeneration, however this phenomenon has not been elucidated in the damaged heart. Therefore, we hypothesized that ADSC transfer own mitochondria to cardiomyocytes in-vivo and in-vitro under ischemic condition, resulting in the functional recovery of cardiomyocyte.

Method and result

Transplantation of hADSC (group A) to the heart surface or sham operation (group C) was performed in rats that were subjected to LAD ligation 2 weeks prior to the treatment (n=10 each). The number of transplant cell was 1x106/body. Three days after transplantation, transferred hADSCs' mitochondria were observed in recipient cardiomyocytes histologically (Figure). Quantitative PCR analysis revealed that mitochondrial genome of recipient myocytes increased over time. The cardiac function assessed with echocardiography was significantly better in group A. Furthermore, live-imaging of hADSC transplantation revealed the suspected transfer of mitochondria to beating heart.

In-vitro, the co-culture of rat cardiomyocytes (rCM) and hADSC was observed with time-lapse photography and demonstrated mitochondrial transfer under the hypoxic condition. The measuring the oxygen consumption rate (OCR) of these cells showed that OCR of rCM was reinforced by co-culture with hADSC conspicuously.

Figure 1

Conclusion

Mitochondrial transfer from hADSC to rCM was suggested in-vivo and in-vitro ischemic condition and suspected to be related to functional recovery of ischemic cardiomyocyte.

P5391Systemic administration of high-mobility group box 1 can suppress adverse post-infarction ventricular remodeling in a rat infarction model by enhancing self-regeneration

Background

High-mobility group box 1 protein (HMGB1) reportedly enhances CXCR4-positive bone marrow-derived mesenchymal stem cell (BM-MSC) recruitment to damaged tissue to promote tissue regeneration.

Purpose

Our aim of this study is to evaluate whether systemic administration of HMGB1 might promote tissue repair in a rat myocardial infarction (MI) model.

Methods

We prepared 26 MI model rats with high ligation of the left coronary artery. Two weeks later, HMGB1 (3 mg/kg/day) or phosphate-buffered saline (control: 3 mL/kg/day) was administered for 4 days via femoral vein. Cardiac performance was evaluated by ultrasonography, left ventricular (LV) remodeling via immunostaining. We then used immunostaining to examine MSC recruitment to damaged tissue in green fluorescent protein bone marrow transplantation (GFP-BMT) model rats, and also performed intravital imaging using two-photon microscopy to visualize BM-cells recruitment in real time.

Results

Compared with control rats, there was a significant improvement in the left ventricular ejection fraction of the HMGB1 group (HMGB1 vs. control: 48.6% ± 5.5% vs. 33.6% ± 5.4%; p<0.01) at 4 weeks after each administration. LV remodeling, characterized by interstitial fibrosis, cardiomyocyte hypertrophy, and a decrease of capillary density, was significantly attenuated in the HMGB1 group compared with control rats. On QT-PCR analysis, VEGF mRNA expression was significantly higher in the HMGB1 group than in the control (border zone; 1.6±0.6 vs. 1.1±0.2; p=0.02, septal zone; 1.1±0.1 vs. 0.9±0.1; p<0.01). In GFP-BMT rats, GFP+/PDGFR+ cells were significantly mobilized to the border zone in the HMGB1 group compared with the control (1331±197 vs. 615±45 /mm2; p<0.01), leading to formation of newly developed vasculature (Figure 1). In intravital imaging, more GFP+ cells were mobilized to the infarction area in the HMGB1 group than in the control, which was further enhanced at 12h later. Additionally, SDF-1 expression in the peri-infarction area increased significantly in MI rats compared with normal rats (MI vs. normal; 2.1±0.4 vs. 0.9±0.1; p<0.01), in where some cell-adhesions of vascular endothelial cells were destroyed.

Conclusions

Systemic administration of HMGB1 mobilized BM-MSCs to the damaged myocardium via the SDF-1/CXCR4 signaling complex. Those BM-MSCs might migrate to extracellular matrix in the border zone via the gap of each endothelial cell, leading to induction of angiogenesis and reduced fibrosis.

Acknowledgement/Funding

None

Antigen Clearing from Porcine Heart Valves with Preservation of Structural Integrity

Bioprostheses currently used for replacement of diseased cardiovascular tissue are preserved and partially protected from immune rejection through chemical fixation. However, after implantation, chemically preserved (fixed) material has limited durability and lacks the ability to revitalize through cellular ingrowth and remodeling. As an alternative to fixation, we aimed at thoroughly removing antigens from tissue, leaving an intact scaffold, suitable for integration and revitalization in the host. Extensive washing of porcine heart valves with a mixture of two detergents (SDS and Triton X-100) yielded an intact matrix devoid of cells and depleted of soluble proteins that was minimally immunogenic in rabbits. A detailed characterization of the biomechanics and durability of the tissue is under way. If the lack of immunogenicity is confirmed in primates, our results would suggest that a detergent-washed, unfixed porcine heart valve can be an attractive non-inflammatory scaffold for heart valve regeneration in humans.

Immunohistochemical expression of fibrillin-1 and fibrillin-2 during tooth development

BACKGROUND AND OBJECTIVE

Oxytalan fibers are categorized as a microfibril assembly without elastin deposition, and are unique components in the periodontal ligament (PDL). However, little is known about their formation during PDL development. To clarify the mechanisms of oxytalan fiber formation in developing PDL, we performed immunohistochemical analysis to detect the direct expression of fibrillin-1 and fibrillin-2, which are major components of microfibrils.

MATERIAL AND METHODS

Frozen sections of lower molars from mice at several stages of growth were prepared without chemical fixation and decalcification using the film transfer method. Immunostaining was performed with anti-fibrillin-1 and -2, and anticytokeratin antibodies.

RESULTS

Fibrillin-1 was not expressed in the dental follicle during the crown forming stage. At postneonatal day 9, fibrillin-1 expression started with meshwork appearance between the epithelial cells from Hertwig's epithelial root sheath at the root dentin surface. Fibirillin-2 was detected much earlier than fibrillin-1 expression. Fibrillin-2 was expressed with a liner appearance, running parallel to the root axis in PDL, and was partially co-expressed with cytokeratin 14 expression in Hertwig's epithelial root sheath. Furthermore, we detected both fibrillin-1 and fibrillin-2 expression in human PDL. Fibrillin-1 was detected in fibers with a vertically oriented root axis in PDL. Fibrillin-2 was widely expressed in PDL, including around the epithelial cell rests of Malassez. Fibrillin-1 and fibrillin-2 were clearly co-expressed in thick fiber structures in human PDL.

CONCLUSION

Our results suggest that both fibrillin-1 and fibrillin-2 expression is required to form thick oxytalan fibers in PDL. Based on the expression patterns for fibrillin-1 and fibrillin-2, they have different functions during tooth root and PDL development. Early expression of fibrillin-2 may regulate dental epithelial cell behavior during root and PDL development.