医学影像AI处理

# 使用PyDICOM读取医学影像
# 技能：scientific-skills:pydicom

import pydicom
import numpy as np
import matplotlib.pyplot as plt

# 读取DICOM文件
dcm = pydicom.dcmread("CT_scan_001.dcm")

# 提取患者信息
patient_id = dcm.PatientID
patient_name = dcm.PatientName
study_date = dcm.StudyDate

print(f"患者ID: {patient_id}")
print(f"检查日期: {study_date}")

# 获取图像参数
slice_thickness = dcm.SliceThickness
pixel_spacing = dcm.PixelSpacing

print(f"层厚: {slice_thickness} mm")
print(f"像素间距: {pixel_spacing} mm")

# 提取图像数据
image_array = dcm.pixel_array.astype(np.float32)

# 显示图像
plt.figure(figsize=10, 10))
plt.imshow(image_array, cmap='gray')
plt.title('CT扫描图像')
plt.axis('off')
plt.show()

# 窗宽窗位调整
def apply_window_level(image, window_center, window_width):
    # 计算窗宽窗位范围
    window_min = window_center - window_width / 2
    window_max = window_center + window_width / 2

    # 限制范围并归一化到0-255
    windowed = np.clip(image, window_min, window_max)
    windowed = ((windowed - window_min) / (window_max - window_min) * 255).astype(np.uint8)

    return windowed

# 肺窗（显示肺部结构）
lung_window = apply_window_level(image_array, window_center=-600, window_width=1500)

# 纵隔窗（显示纵隔结构）
mediastinal_window = apply_window_level(image_array, window_center=40, window_width=400)

# 重采样到统一分辨率
from skimage.transform import resize

target_spacing = (1.0, 1.0)  # 1mm x 1mm
resized = resize(
    image_array,
    (512, 512),
    preserve_range=True,
    anti_aliasing=True
)

# 标准化到0-1
normalized = (image_array - image_array.min()) / (image_array.max() - image_array.min())

# 使用HistoLab分析病理切片
# 技能：scientific-skills:histolab

from histolab import Tile, TilesGrid
from histolab.filters import RandomRotation, RandomFlip
from histolab.data import polar_bear_slide

# 加载病理切片（SVS格式）
slide = polar_bear_slide()

# 创建瓦片网格（将大图分割成小块）
grid = TilesGrid(
    slide,
    tile_size=256,
    overlap=32,
    max_tiles=100
)

# 组织分割（区分组织和背景）
from histolab.segmentation import OtsuSegmentation

segmenter = OtsuSegmentation()
tissue_mask = segmenter(slide)

# 细胞检测
from histolab.detection import CellDetector

detector = CellDetector()
cells = detector.detect(
    slide,
    min_radius=5,
    max_radius=20
)

print(f"检测到 {len(cells)} 个细胞")

# 特征提取
from histolab.features import extract_features

features = extract_features(
    slide,
    cells,
    features=['area', 'perimeter', 'circularity', 'nucleus_intensity']
)

# 多平面重建（MPR）
import nibabel as nib
from scipy.ndimage import zoom

# 加载3D体积
nii = nib.load('brain_scan.nii')
volume = nii.get_fdata()

# 提取三个正交平面
axial = volume[:, :, volume.shape[2]//2]
sagittal = volume[:, volume.shape[1]//2, :]
coronal = volume[volume.shape[0]//2, :, :]

# 3D体绘制
from mpl_toolkits.mplot3d import Axes3D

def volume_rendering(volume, threshold):
    # 提取等值面
    verts, faces, _, _ = measure.marching_cubes(volume, threshold)

    # 3D显示
    fig = plt.figure(figsize=10, 10))
    ax = fig.add_subplot(111, projection='3d')
    ax.plot_trisurf(verts[:, 0], verts[:, 1], verts[:, 2], triangles=faces)
    plt.show()

# 病理切片标注
import matplotlib.patches as patches

fig, ax = plt.subplots(1, 1, figsize=12, 12))
ax.imshow(slide.get_thumbnail(2048))

# 绘制检测到的细胞
for cell in cells:
    circle = patches.Circle(
        (cell.x, cell.y),
        radius=cell.radius,
        fill=False,
        edgecolor='red',
        linewidth=2
    )
    ax.add_patch(circle)

plt.title('细胞检测结果')
plt.axis('off')
plt.show()