实验室版 API

实验室版提供ECGFounder、文本生成单导联、单导联生成十二导联、三导联生成十二导联、PPG信号生成单导联以及单导联生成心脏超声的核心能力。

注意事项

🔬 欢迎使用实验室 API！
目前所有接口均处于公测阶段，功能可能会不定期调整。
为确保正常访问和数据安全，请进入 科研合作页面 联系工作人员获取专属访问密钥。
💡 温馨提示：请妥善保管您的密钥，避免泄露或滥用。

适用场景

• 医学研究
• 心电模型验证

---

MCMA接口

POST /api/v1/experimental/mcma

通过输入单导联信号可以生成十二导联信号。

请求参数

参数	类型	必填	说明
ecgData	array	✅	ECG信号数据数组（ADC值）
ecgSamplerate	number	✅	采样率，单位Hz（推荐250-500）
original	boolean	❌	是否输出原始波形，默认true

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/mcma" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ecgData": [512, 515, 520, 518, 525, ...],
    "ecgSampleRate": 500
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": [
    [
      -0.0005600000149570405,
      0.05680999904870987,
      -0.002580000087618828,
      0.02824000082910061,
      ...
      ]
    ]
}

可视化展示

import ecg_plot
ecg_plot.plot(gen_ecg12, sample_rate=fs, title='ECG 12')
ecg_plot.show()

文献

@article{chen2024multi,
  title={Multi-channel masked autoencoder and comprehensive evaluations for reconstructing 12-lead ECG from arbitrary single-lead ECG},
  author={Chen, Jiarong and Wu, Wanqing and Liu, Tong and Hong, Shenda},
  journal={npj Cardiovascular Health},
  volume={1},
  number={1},
  pages={34},
  year={2024},
  publisher={Nature Publishing Group UK London}
}

Chen, J., Wu, W., Liu, T., & Hong, S. (2024). Multi-channel masked autoencoder and comprehensive evaluations for reconstructing 12-lead ECG from arbitrary single-lead ECG. npj Cardiovascular Health, 1(1), 34.

---

DiffuSETS接口

POST /api/v1/experimental/diffuSets

通过输入心电图诊断结果、年龄、性别以及心率即可生成对应的心电图。

请求参数

参数	类型	必填	说明
text	str	✅	临床诊断文本，多个诊断需要用'|'分隔开
age	int	✅	年龄
sex	str	✅	性别
hr	int	✅	心率
batch	int	❌	选择生成ECG的数量

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/diffuSets" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": 'Sinus rhythm|Normal ECG.',  # Clinical text report, multi-reports should be split by '|'
    "age": 50,  # Age of patient
    "sex": 'M',
    "hr": 80,
    'batch': 1
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": {
    "textImg": "[\"iVBORw0KGgoAAAANSUhI=...\"]"
  }
}

可视化展示

import time
import requests
import json
import base64
import io
from PIL import Image

def base64_to_image(base64_str):
    # 解码Base64字符串
    image_data = base64.b64decode(base64_str)
    # 使用io.BytesIO将解码后的数据转换为文件对象
    image_file = io.BytesIO(image_data)
    # 使用PIL打开图片
    image = Image.open(image_file)
    return image

ecg_img = base64_to_image(ecg_data)
ecg_img.show()  

文献

@article{lai2025diffusets,
  title={DiffuSETS: 12-Lead ECG generation conditioned on clinical text reports and patient-specific information},
  author={Lai, Yongfan and Chen, Jiabo and Zhao, Qinghao and Zhang, Deyun and Wang, Yue and Geng, Shijia and Li, Hongyan and Hong, Shenda},
  journal={Patterns},
  year={2025},
  publisher={Elsevier}
}

Lai, Y., Chen, J., Zhao, Q., Zhang, D., Wang, Y., Geng, S., ... & Hong, S. (2025). DiffuSETS: 12-Lead ECG generation conditioned on clinical text reports and patient-specific information. Patterns.

---

WearECG接口

POST /api/v1/experimental/lead3Lead12

通过输入三导联可以生成十二导联信号。

请求参数

参数	类型	必填	说明
ecgData	array	✅	ECG信号数据数组（ADC值）
ecgSampleRate	number	✅	采样率，单位Hz（推荐250-500）

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/lead3Lead12" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ecgData": [[512, 515, 520, 518, 525, ...],[512, 515, 520, 518, 525, ...],[512, 515, 520, 518, 525, ...]],
    "ecgSampleRate": 500
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": [
    [
      -0.0025084596127271652,
      0.0118071788037084,
      0.014166647854523678,
      ...
      ]
    ]
}

可视化展示

import ecg_plot
ecg_plot.plot(gen_ecg12, sample_rate=fs, title='ECG 12')
ecg_plot.show()

---

PPGFlowECG接口

POST /api/v1/experimental/ppgFlowEcg

通过输入PPG信号可以生成高质量ECG信号。

请求参数

参数	类型	必填	说明
ppgData	array	✅	PPG信号数据数组。上传红光、绿光或者红外光数据
ppgSampleRate	number	✅	采样率，单位Hz（推荐250-500）

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/ppgFlowEcg" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ppgData": [512, 515, 520, 518, 525, ...],
    "ecgSampleRate": 500
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": [
    [
      -0.0005600000149570405,
      0.05680999904870987,
      -0.002580000087618828,
      0.02824000082910061,
      ...
      ]
    ]
}

可视化展示

文献

@article{fang2025ppgflowecg,
  title={PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection},
  author={Fang, Xiaocheng and Jin, Jiarui and Wang, Haoyu and Liu, Che and Cai, Jieyi and Nie, Guangkun and Li, Jun and Li, Hongyan and Hong, Shenda},
  journal={arXiv preprint arXiv:2509.19774},
  year={2025}
}

Fang, X., Jin, J., Wang, H., Liu, C., Cai, J., Nie, G., ... & Hong, S. (2025). PPGFlowECG: Latent Rectified Flow with Cross-Modal Encoding for PPG-Guided ECG Generation and Cardiovascular Disease Detection. arXiv preprint arXiv:2509.19774.

---

ECHOPulse接口

POST /api/v1/experimental/ecgEcho

通过输入ECG信号可以生成对应的心脏超声视频。

请求参数

参数	类型	必填	说明
ecgData	array	✅	ECG信号数据数组（ADC值）
ecgSampleRate	number	✅	采样率，单位Hz（推荐250-500）

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/ecgEcho" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ecgData": [512, 515, 520, 518, 525, ...],
    "ecgSampleRate": 500
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": {
    "ecgEcho": [
      [
        [
          [
            [
              0.0013988255523145199,
              0.004246973432600498,
              0.0010940954089164734,
              0.004930938594043255,
              -0.0024434151127934456,
              0.00477356743067503,
              ...
            ]
          ]
        ]
      ]
    ]
  }
}

可视化展示

保存ECHO的视频帧

def save_video_and_plot_frames(video_tensor, save_dir='output'):
    os.makedirs(save_dir, exist_ok=True)
    T = video_tensor.shape[2]  # 时间维度
    fig, axes = plt.subplots(1, T, figsize=(T * 3, 3))

    for t in range(T):
        # 取出第 t 帧
        frame = video_tensor[0, :, t, :, :]  # [C, H, W]

        # 转成 [H, W, C] numpy
        frame_np = frame.permute(1, 2, 0).cpu().numpy()

        # 归一化到 [0,1] 避免 imshow clipping
        frame_min, frame_max = frame_np.min(), frame_np.max()
        if frame_max > frame_min:
            frame_np = (frame_np - frame_min) / (frame_max - frame_min)
        else:
            frame_np = np.zeros_like(frame_np)  # 防止除零

        ax = axes[t]
        ax.imshow(frame_np)
        ax.set_title(f'Frame {t}')
        ax.axis('off')

    for i in range(T, len(axes)):
        axes[i].axis('off')

    plt.tight_layout()
    plt.savefig(os.path.join(save_dir, 'echo_frames.png'))
    plt.show()
    plt.close()

    print(f"All frames plotted in {save_dir}")

保存结果为gif

def video_tensor_to_gif_with_ecg(video_tensor, ecg_data, output_path, title, duration=125, fps=8):
    # 将视频内容转换成numpy阵列
    video_np = np.array(video_tensor)
    video_np = np.squeeze(video_np, axis=0)
    # 改变维度顺序 (C, T, H, W) -> (T, H, W, C)
    video_np = np.transpose(video_np, (1, 2, 3, 0))

    # EKG 数据准备(缩小维度)
    ecg = np.array(ecg_data)
    ecg = ecg[0, 1, :]  # Assuming this is the correct dimension for EKG data

    # 确认帧数
    num_frames = 11
    # EKG 数据根据帧数重新抽样
    x_original = np.linspace(0, 1, len(ecg))
    x_resampled = np.linspace(0, 1, num_frames)
    ecg_resampled = np.interp(x_resampled, x_original, ecg)

    # 设定画布
    fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(8, 10), gridspec_kw={'height_ratios': [3, 1]})

    # 设置初始化图像
    im = ax1.imshow(np.zeros((128, 128, 3)), animated=True)
    ax1.set_title(title)
    line, = ax2.plot([], [], 'r-')

    ax1.axis('off')
    ax2.set_xlim(0, num_frames)
    ax2.set_ylim(ecg_resampled.min(), ecg_resampled.max())
    ax2.set_xlabel('Frame')
    ax2.set_ylabel('EKG')
    ax2.set_title('Lead I EKG')

    def update(frame):
        im.set_array(video_np[frame])
        line.set_data(range(frame + 1), ecg_resampled[:frame + 1])
        return im, line

    num_frames = video_np.shape[0]
    anim = FuncAnimation(fig, update, frames=num_frames, interval=duration, blit=True)

    anim.save(output_path, writer='pillow', fps=fps)
    plt.close(fig)

    print(f"GIF saved to {output_path}")

可视化展示

• 生成的ECHO帧（调用save_video_and_plot_frames）

• 转成gif（调用video_tensor_to_gif_with_ecg）

文献

@article{li2024echopulse,
  title={Echopulse: Ecg controlled echocardio-grams video generation},
  author={Li, Yiwei and Kim, Sekeun and Wu, Zihao and Jiang, Hanqi and Pan, Yi and Jin, Pengfei and Song, Sifan and Shi, Yucheng and Liu, Tianming and Li, Quanzheng and others},
  journal={arXiv preprint arXiv:2410.03143},
  year={2024}
}

Li, Y., Kim, S., Wu, Z., Jiang, H., Pan, Y., Jin, P., ... & Li, X. (2024). Echopulse: Ecg controlled echocardio-grams video generation. arXiv preprint arXiv:2410.03143.

---

ECGTwins接口

POST /api/v1/experimental/ecgTwins

通过输入年龄、心率、节律、性别可以生成个性化ECG信号。

请求参数

参数	类型	必填	说明
reference_file	file	✅	参考文件，目前是固定的 示例文件
text	str	✅	临床诊断文本，多个诊断需要用'|'分隔开
age	int	✅	年龄
sex	str	✅	性别
hr	int	✅	心率
batch	int	❌	选择生成ECG的数量

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/diffuSets" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -F "file=@normal_1-66b62bfc8acabc11cf5cc34cafe860f4.pt"
  -d '{
    "text": 'Sinus rhythm|Normal ECG.',  # Clinical text report, multi-reports should be split by '|'
    "age": 50,  # Age of patient
    "sex": 'M',
    "hr": 80,
    'batch': 1
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": {
    "ECGTwin_imgs": [
        [
          0.0013988255523145199,
          0.004246973432600498,
          0.0010940954089164734,
          0.004930938594043255,
          -0.0024434151127934456,
          0.00477356743067503,
          ...
        ]
    ]
  }
}

可视化展示

import requests
import json
import base64
import io
from PIL import Image

def base64_to_image(base64_str):
    # 解码Base64字符串
    image_data = base64.b64decode(base64_str)
    # 使用io.BytesIO将解码后的数据转换为文件对象
    image_file = io.BytesIO(image_data)
    # 使用PIL打开图片
    image = Image.open(image_file)
    return image

ecg_img = base64_to_image(ecg_data)
ecg_img.show()
    

文献

@article{lai2025ecgtwin,
  title={ECGTwin: Personalized ECG Generation Using Controllable Diffusion Model},
  author={Lai, Yongfan and Liu, Bo and Guan, Xinyan and Zhao, Qinghao and Li, Hongyan and Hong, Shenda},
  journal={arXiv preprint arXiv:2508.02720},
  year={2025}
}

Lai, Y., Liu, B., Guan, X., Zhao, Q., Li, H., & Hong, S. (2025). ECGTwin: Personalized ECG Generation Using Controllable Diffusion Model. arXiv preprint arXiv:2508.02720.

---

ECGFounder接口

POST /api/v1/experimental/ecgFounder

ECGFounder旨在构建一个 心电图（ECG, electrocardiogram）基础模型（foundation model），即一个通用的、预训练的模型，可以在很多下游 ECG 任务中拿来 fine-tune 或直接使用。

请求参数

参数	类型	必填	说明
ecgData	array	✅	ECG信号数据数组（ADC值）
ecgSampleRate	number	✅	采样率，单位Hz（推荐250-500）

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/ecgFounder" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "ecgData": [512, 515, 520, 518, 525, ...],
    "ecgSampleRate": 500
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": {
    "otherwise_normal_ecg": null,
    "rsr_OR_QR_PATTERN_IN_V1_SUGGESTS_RIGHT_VENTRICULAR_CONDUCTION_DELAY": null,
    "nonspecific_T_WAVE_ABNORMALITY_HAS_REPLACED_INVERTED_T_WAVES_IN": null,
    "inverted_T_WAVES_HAVE_REPLACED_NON_SPECIFIC_T_WAVE_ABNORMALITY_IN": null,
    "voltage_CRITERIA_FOR_LEFT_VENTRICULAR_HYPERTROPHY": null,
    "bifascicular_BLOCK": null,
    ...
  }
}

中英文对照

英文标签	中文翻译
ABNORMAL ECG	异常心电图
NORMAL SINUS RHYTHM	正常窦性心律
NORMAL ECG	正常心电图
SINUS RHYTHM	窦性心律
SINUS BRADYCARDIA	窦性心动过缓
ATRIAL FIBRILLATION	心房颤动
SINUS TACHYCARDIA	窦性心动过速
otherwise normal ecg	其他方面正常的心电图
LEFT AXIS DEVIATION	左轴偏移
PREMATURE VENTRICULAR COMPLEXES	室性早搏
BORDERLINE ECG	临界心电图
RIGHT BUNDLE BRANCH BLOCK	右束支传导阻滞
SEPTAL INFARCT	室间隔梗死
LEFT ATRIAL ENLARGEMENT	左心房扩大
NONSPECIFIC T WAVE ABNORMALITY	非特异性 T 波异常
LOW VOLTAGE QRS	低电压 QRS 波群
PREMATURE ATRIAL COMPLEXES	房性早搏
ANTERIOR INFARCT	前壁梗死
INCOMPLETE RIGHT BUNDLE BRANCH BLOCK	不完全性右束支传导阻滞
PREMATURE SUPRAVENTRICULAR COMPLEXES	上室性早搏
LEFT BUNDLE BRANCH BLOCK	左束支传导阻滞
NONSPECIFIC T WAVE ABNORMALITY NOW EVIDENT IN	当前出现非特异性 T 波异常
NONSPECIFIC T WAVE ABNORMALITY NO LONGER EVIDENT IN	非特异性 T 波异常已消失
T WAVE INVERSION NOW EVIDENT IN	当前出现 T 波倒置
LATERAL INFARCT	侧壁梗死
NONSPECIFIC ST ABNORMALITY	非特异性 ST 段异常
LEFT VENTRICULAR HYPERTROPHY	左心室肥大
T WAVE INVERSION NO LONGER EVIDENT IN	T 波倒置已消失
WITH RAPID VENTRICULAR RESPONSE	伴快速心室反应
QT HAS SHORTENED	QT 间期缩短
QT HAS LENGTHENED	QT 间期延长
FUSION COMPLEXES	融合波
ATRIAL FLUTTER	心房扑动
MARKED SINUS BRADYCARDIA	明显窦性心动过缓
WITH SINUS ARRHYTHMIA	伴窦性心律不齐
NONSPECIFIC ST AND T WAVE ABNORMALITY	非特异性 ST-T 波异常
LEFT ANTERIOR FASCICULAR BLOCK	左前分支传导阻滞
RIGHT AXIS DEVIATION	右轴偏移
ECTOPIC ATRIAL RHYTHM	异位房性心律
UNDETERMINED RHYTHM	未定型心律
ANTEROSEPTAL INFARCT	前间隔梗死
RIGHTWARD AXIS	右偏心电轴
ST NOW DEPRESSED IN	当前出现 ST 段压低
WITH SHORT PR	伴短 PR 间期
WITH MARKED SINUS ARRHYTHMIA	伴明显窦性心律不齐
ST NO LONGER DEPRESSED IN	ST 段压低已消失
INVERTED T WAVES HAVE REPLACED NONSPECIFIC T WAVE ABNORMALITY IN	T 波倒置取代了非特异性 T 波异常
NON-SPECIFIC CHANGE IN ST SEGMENT IN	ST 段非特异性改变
NONSPECIFIC T WAVE ABNORMALITY HAS REPLACED INVERTED T WAVES IN	非特异性 T 波异常取代了倒置 T 波
JUNCTIONAL RHYTHM	交界性心律
ELECTRONIC ATRIAL PACEMAKER	电子房起搏器
ABERRANT CONDUCTION	异常传导
ELECTRONIC VENTRICULAR PACEMAKER	电子心室起搏器
T WAVE INVERSION LESS EVIDENT IN	T 波倒置减轻
ANTEROLATERAL INFARCT	前外侧梗死
WITH REPOLARIZATION ABNORMALITY	伴复极异常
RSR' OR QR PATTERN IN V1 SUGGESTS RIGHT VENTRICULAR CONDUCTION DELAY	V1 导联 RSR' 或 QR 波形提示右心室传导延迟
T WAVE INVERSION MORE EVIDENT IN	T 波倒置更加明显
WIDE QRS RHYTHM	宽 QRS 波群心律
WITH PREMATURE VENTRICULAR OR ABERRANTLY CONDUCTED COMPLEXES	伴室性早搏或异常传导复合波
RIGHT ATRIAL ENLARGEMENT	右心房扩大
INFERIOR INFARCT	下壁梗死
INCOMPLETE LEFT BUNDLE BRANCH BLOCK	不完全性左束支传导阻滞
VOLTAGE CRITERIA FOR LEFT VENTRICULAR HYPERTROPHY	左心室肥大的电压标准
OR DIGITALIS EFFECT	或洋地黄效应
BIFASCICULAR BLOCK	双分支传导阻滞
ST NO LONGER ELEVATED IN	ST 段抬高已消失
WITH SLOW VENTRICULAR RESPONSE	伴缓慢心室反应
ST ELEVATION NOW PRESENT IN	当前出现 ST 段抬高
PREMATURE ECTOPIC COMPLEXES	早期异位复合波
LEFT POSTERIOR FASCICULAR BLOCK	左后分支传导阻滞
T WAVE AMPLITUDE HAS DECREASED IN	T 波振幅降低
WITH A COMPETING JUNCTIONAL PACEMAKER	伴竞争性交界性起搏
RIGHT SUPERIOR AXIS DEVIATION	右上轴偏移
BIATRIAL ENLARGEMENT	双心房扩大
VENTRICULAR-PACED RHYTHM	心室起搏心律
ATRIAL-PACED RHYTHM	房起搏心律
T WAVE AMPLITUDE HAS INCREASED IN	T 波振幅增高
WITH QRS WIDENING	伴 QRS 波群增宽
WITH 1ST DEGREE AV BLOCK	伴一度房室传导阻滞
PROLONGED QT	QT 间期延长
WITH PROLONGED AV CONDUCTION	伴房室传导延迟
RIGHT VENTRICULAR HYPERTROPHY	右心室肥大
WITH QRS WIDENING AND REPOLARIZATION ABNORMALITY	伴 QRS 波群增宽及复极异常
ATRIAL-SENSED VENTRICULAR-PACED RHYTHM	房感知-心室起搏心律
AV SEQUENTIAL OR DUAL CHAMBER ELECTRONIC PACEMAKER	房室顺序或双腔电子起搏器
PULMONARY DISEASE PATTERN	肺型心电图
ACUTE MI / STEMI	急性心肌梗死 / ST 段抬高型心梗
INFERIOR-POSTERIOR INFARCT	下后壁梗死
NONSPECIFIC INTRAVENTRICULAR CONDUCTION DELAY	非特异性心室内传导延迟
PREMATURE VENTRICULAR AND FUSION COMPLEXES	室性早搏及融合波
IN A PATTERN OF BIGEMINY	二联律模式
AV DUAL-PACED RHYTHM	双腔起搏心律
SUPRAVENTRICULAR TACHYCARDIA	上室性心动过速
VENTRICULAR-PACED COMPLEXES	心室起搏复合波
WIDE QRS TACHYCARDIA	宽 QRS 心动过速
RSR' PATTERN IN V1	V1 导联 RSR' 波形
ST LESS DEPRESSED IN	ST 段压低减轻
VENTRICULAR TACHYCARDIA	心室性心动过速
EARLY REPOLARIZATION	早复极
ST MORE DEPRESSED IN	ST 段压低加重
ANTEROLATERAL LEADS	前外侧导联
ELECTRONIC DEMAND PACING	电子按需起搏
RBBB AND LEFT ANTERIOR FASCICULAR BLOCK	右束支阻滞伴左前分支阻滞
LATERAL INJURY PATTERN	侧壁损伤图形
BIVENTRICULAR PACEMAKER DETECTED	检测到双心室起搏器
SUSPECT UNSPECIFIED PACEMAKER FAILURE	疑似未明确的起搏器功能障碍
WOLFF-PARKINSON-WHITE	预激综合征（Wolff-Parkinson-White 综合征）
WITH VENTRICULAR ESCAPE COMPLEXES	伴心室逸搏复合波
INFERIOR INJURY PATTERN	下壁损伤图形
CONSIDER RIGHT VENTRICULAR INVOLVEMENT IN ACUTE INFERIOR INFARCT	急性下壁心梗伴右室受累可能
ST ELEVATION HAS REPLACED ST DEPRESSION IN	ST 段抬高取代 ST 段压低
NONSPECIFIC INTRAVENTRICULAR BLOCK	非特异性室内传导阻滞
MASKED BY FASCICULAR BLOCK	被分支阻滞掩盖
PEDIATRIC ECG ANALYSIS	儿科心电图分析
BLOCKED	阻滞
WITH UNDETERMINED RHYTHM IRREGULARITY	伴不确定心律不齐
LEFTWARD AXIS	左偏电轴
WITH 2ND DEGREE SA BLOCK MOBITZ I	伴二度窦房阻滞 Mobitz I 型
ACUTE	急性
ABNORMAL LEFT AXIS DEVIATION	异常左轴偏移
WITH COMPLETE HEART BLOCK	伴完全性房室传导阻滞
NO P-WAVES FOUND	未检测到 P 波
ST LESS ELEVATED IN	ST 段抬高减轻
WITH RETROGRADE CONDUCTION	伴逆向传导
ST MORE ELEVATED IN	ST 段抬高加重
JUNCTIONAL BRADYCARDIA	交界性心动过缓
WITH VARIABLE AV BLOCK	伴变异性房室传导阻滞
ANTERIOR INJURY PATTERN	前壁损伤图形
WITH JUNCTIONAL ESCAPE COMPLEXES	伴交界性逸搏复合波
ACUTE MI	急性心肌梗死
ACUTE PERICARDITIS	急性心包炎
POSTERIOR INFARCT	后壁梗死
IDIOVENTRICULAR RHYTHM	特发性心室心律
WITH 2ND DEGREE SA BLOCK MOBITZ II	伴二度窦房阻滞 Mobitz II 型
R IN AVL	AVL 导联 R 波
SINUS/ATRIAL CAPTURE	窦性/房性捕获
AV DUAL-PACED COMPLEXES	双腔起搏复合波
INFEROLATERAL INJURY PATTERN	下外侧损伤图形
RBBB AND LEFT POSTERIOR FASCICULAR BLOCK	右束支阻滞伴左后分支阻滞
ANTEROLATERAL INJURY PATTERN	前外侧损伤图形
ATRIAL-PACED COMPLEXES	房起搏复合波
WITH SINUS PAUSE	伴窦性停搏
BIVENTRICULAR HYPERTROPHY	双心室肥大
ABNORMAL RIGHT AXIS DEVIATION	异常右轴偏移
SUPRAVENTRICULAR COMPLEXES	上室性复合波
WITH 2ND DEGREE AV BLOCK MOBITZ I	伴二度房室传导阻滞 Mobitz I 型
WITH 2:1 AV CONDUCTION	伴 2:1 房室传导
WITH AV DISSOCIATION	伴房室分离
MULTIFOCAL ATRIAL TACHYCARDIA	多源性房性心动过速

文献

@article{li2025electrocardiogram,
  title={An Electrocardiogram Foundation Model Built on over 10 Million Recordings},
  author={Li, Jun and Aguirre, Aaron D and Junior, Valdery Moura and Jin, Jiarui and Liu, Che and Zhong, Lanhai and Sun, Chenxi and Clifford, Gari and Brandon Westover, M and Hong, Shenda},
  journal={NEJM AI},
  volume={2},
  number={7},
  pages={AIoa2401033},
  year={2025},
  publisher={Massachusetts Medical Society}
}

Li, J., Aguirre, A., Moura, J., Liu, C., Zhong, L., Sun, C., ... & Hong, S. (2024). An electrocardiogram foundation model built on over 10 million recordings with external evaluation across multiple domains. arXiv preprint arXiv:2410.04133.

---

心电图分析接口

POST /api/v1/experimental/ecgTwins

通过输入单导联信号可以生成十二导联信号。

请求参数

参数	类型	必填	说明
reference_file	file	✅	参考文件，目前是固定的 示例文件
text	str	✅	临床诊断文本，多个诊断需要用'|'分隔开
age	int	✅	年龄
sex	str	✅	性别
hr	int	✅	心率
batch	int	❌	选择生成ECG的数量

请求示例

curl -X POST "https://api.heartvoice.com.cn/api/v1/experimental/diffuSets" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -F "file=@normal_1-66b62bfc8acabc11cf5cc34cafe860f4.pt"
  -d '{
    "text": 'Sinus rhythm|Normal ECG.',  # Clinical text report, multi-reports should be split by '|'
    "age": 50,  # Age of patient
    "sex": 'M',
    "hr": 80,
    'batch': 1
  }'

响应示例

{
  "errorCode": "0",
  "msg": "成功",
  "data": {
    "ECGTwin_imgs": [
        [
          0.0013988255523145199,
          0.004246973432600498,
          0.0010940954089164734,
          0.004930938594043255,
          -0.0024434151127934456,
          0.00477356743067503,
          ...
        ]
    ]
  }
}

可视化展示

import requests
import json
import base64
import io
from PIL import Image

def base64_to_image(base64_str):
    # 解码Base64字符串
    image_data = base64.b64decode(base64_str)
    # 使用io.BytesIO将解码后的数据转换为文件对象
    image_file = io.BytesIO(image_data)
    # 使用PIL打开图片
    image = Image.open(image_file)
    return image

ecg_img = base64_to_image(ecg_data)
ecg_img.show()
    

文献

@article{lai2025ecgtwin,
  title={ECGTwin: Personalized ECG Generation Using Controllable Diffusion Model},
  author={Lai, Yongfan and Liu, Bo and Guan, Xinyan and Zhao, Qinghao and Li, Hongyan and Hong, Shenda},
  journal={arXiv preprint arXiv:2508.02720},
  year={2025}
}

Lai, Y., Liu, B., Guan, X., Zhao, Q., Li, H., & Hong, S. (2025). ECGTwin: Personalized ECG Generation Using Controllable Diffusion Model. arXiv preprint arXiv:2508.02720.

---