Kobe University · Graduate School of Medicine / Faculty of Medicine

Research keywords深層学習・バイオインフォマティクス・人工知能・循環器内科・Artificial・intelligence・Internal・medicine

Research fieldsMolecular biology・Circulatory organs internal medicine・生命、健康、医療情報学

This is a public-data preview. Personalized fit, contact angles, and saved workflows require sign-in.

• Assessment of transthyretin instability in patients with wild-type transthyretin amyloid cardiomyopathy2024 · 2024年09月, SCIENTIFIC REPORTS, 14(1) (1), 20508 - 20508, 英語, 国際誌
• Identifying heart failure dynamics using multi-point electrocardiograms and deep learning.2020 · AIMS: Heart failure (HF) hospitalizations are associated with poor survival outcomes, emphasizing the need for early intervention. Deep learning algorithms have shown promise in HF detection through electrocardiogram (ECG). However, their utility in ongoing HF monitoring remains uncertain. This study developed a deep learning model using 12-lead ECGs collected at 2 different time points to evaluate HF status changes, aiming to enhance early intervention and continuous monitoring in various healthcare settings. METHODS AND RESULTS: We analysed 30 171 ECGs from 6531 adult patients at Kobe University Hospital. The participants were randomly assigned to training, validation, and test datasets. A Transformer-based model was developed to classify HF status into deteriorated, improved, and no-change classes based on ECG waveform signals at two different time points. Performance metrics, such as the area under the receiver operating characteristic curve (AUROC) and accuracy, were calculated, and attention mapping via gradient-weighted class activation mapping was utilized to interpret the model's decision-making ability. The patients had an average age of 64.6 years (±15.4 years) and brain natriuretic peptide of 66.3 pg/mL (24.6-175.1 pg/mL). For HF status classification, the model achieved an AUROC of 0.889 [95% confidence interval (CI): 0.879-0.898] and an accuracy of 0.871 (95% CI: 0.864-0.878). CONCLUSION: Transformer-based deep learning model demonstrated high accuracy in detecting HF status changes, highlighting its potential as a non-invasive, efficient tool for HF monitoring. The reliance of the model on ECGs reduces the need for invasive, costly diagnostics, aligning with clinical needs for accessible HF management. IRB INFORMATION: Kobe University Hospital Clinical & Translational Research Center (reference number: B220208).
• Prediction of difficulty in cryoballoon ablation with a three-dimensional deep learning model using polygonal mesh representation.2015 · BACKGROUND: Cryoballoon ablation (CBA) is useful for pulmonary vein (PV) isolation. However, some cases are challenging, requiring multiple applications and/or touch-up ablations. Although several predictors of CBA difficulty have been reported, none have assessed the spatial location and morphology of the left atrium and PVs. This study aimed to develop a three-dimensional (3D) deep learning (DL) model to predict CBA difficulty and compare its accuracy with conventional manual measurement. METHODS: A 28-mm cryoballoon (Arctic Front Advance, Medtronic) was used in all cases. CBA difficulty was defined as requiring touch-up ablation and/or more than three applications per PV. We developed a DL model that can learn polygonal meshes and predict CBA difficulty. In the conventional method, predictors included a thinner left lateral ridge, higher left superior PV (LSPV) ovality index, longer LSPV ostium-bifurcation distance, and shorter right inferior PV ostium-bifurcation distance. RESULTS: A total of 189 patients who underwent CBA for drug-resistant atrial fibrillation between January 2015 and January 2022 were included. The DL model was superior to the conventional method in accuracy (0.793 vs. 0.630, p = .042) and specificity (0.796 vs. 0.609, p = .022), with the AUC-ROC of 0.821. CONCLUSIONS: We developed a 3D DL model that can detect CBA difficulty using a polygonal mesh representation. By predicting difficult cases in advance, strategies can be developed to increase success rates.
• Prediction of difficulty in cryoballoon ablation with a three-dimensional deep learning model using polygonal mesh representation2015 · Background: Cryoballoon ablation (CBA) is useful for pulmonary vein (PV) isolation. However, some cases are challenging, requiring multiple applications and/or touch-up ablations. Although several predictors of CBA difficulty have been reported, none have assessed the spatial location and morphology of the left atrium and PVs. This study aimed to develop a three-dimensional (3D) deep learning (DL) model to predict CBA difficulty and compare its accuracy with conventional manual measurement. Methods: A 28-mm cryoballoon (Arctic Front Advance, Medtronic) was used in all cases. CBA difficulty was defined as requiring touch-up ablation and/or more than three applications per PV. We developed a DL model that can learn polygonal meshes and predict CBA difficulty. In the conventional method, predictors included a thinner left lateral ridge, higher left superior PV (LSPV) ovality index, longer LSPV ostium-bifurcation distance, and shorter right inferior PV ostium-bifurcation distance. Results: A total of 189 patients who underwent CBA for drug-resistant atrial fibrillation between January 2015 and January 2022 were included. The DL model was superior to the conventional method in accuracy (0.793 vs. 0.630, p = .042) and specificity (0.796 vs. 0.609, p = .022), with the AUC-ROC of 0.821. Conclusions: We developed a 3D DL model that can detect CBA difficulty using a polygonal mesh representation. By predicting difficult cases in advance, strategies can be developed to increase success rates.

• 小児心臓検診用マルチモーダル人工知能(AI)モデルの開発研究 西森 誠 日本学術振興会, 科学研究費助成事業, 若手研究, 神戸大学, 2022年04月01日 - 2024年03月31日2022 · KUID parsed section
• 小児心臓検診用マルチモーダル人工知能(AI)モデルの開発研究

• Official profile