a b/Readme.md 

  1
  
# 1. Goal: 使用 YoloV8 对医学影像CT/磁共振图像进行肺炎区域分割:





  
  

  2
  





  
  

  3
  
### part1 :





  
  

  4
  
A. init data [nii(Nifit) -> png]: 将CT原始数据从 三维转二维, 保存png格式





  
  

  5
  
B. mask data [nii -> json]: 将CT的分割标注数据从 三维转二维, 再转Labelme可以识别的json文件





  
  

  6
  





  
  

  7
  
### part2:





  
  

  8
  
**Algorithm: Yolo-V8**





  
  

  9
  
A. make dataset [train/val/test + label]: 对二维数据和json文件进行数据划分, 并json转换Yolo标签格式





  
  

  10
  
B. Segmentation [Yolo-V8]: 使用YoloV8进行分割





  
  

  11
  





  
  

  12
  





  
  

  13
  





  
  

  14
  
# 2. JSON format (eg:) :





  
  

  15
  
{





  
  

  16
  
  "version": "4.5.6",  // Labelme software version





  
  

  17
  
  "flags": {},         // Additional flag information





  
  

  18
  
  "shapes": [





  
  

  19
  
    {





  
  

  20
  
      "label": "dog",  // Label name





  
  

  21
  
      "points": [      // Polygon vertex points of the annotation 





  
  

  22
  
        [298, 151],





  
  

  23
  
        [324, 151],





  
  

  24
  
        [324, 160],





  
  

  25
  
        [317, 160],





  
  

  26
  
        [317, 168],





  
  

  27
  
        [307, 168],





  
  

  28
  
        [307, 151],





  
  

  29
  
        [298, 151]





  
  

  30
  
      ],





  
  

  31
  
      "group_id": null,





  
  

  32
  
      "shape_type": "polygon",  // Shape type





  
  

  33
  
      "flags": {}





  
  

  34
  
    },





  
  

  35
  
    {





  
  

  36
  
      "label": "cat",





  
  

  37
  
      "points": [





  
  

  38
  
        [400, 200],





  
  

  39
  
        [425, 200],





  
  

  40
  
        [425, 210],





  
  

  41
  
        [415, 210],





  
  

  42
  
        [415, 220],





  
  

  43
  
        [405, 220],





  
  

  44
  
        [405, 200],





  
  

  45
  
        [400, 200]





  
  

  46
  
      ],





  
  

  47
  
      "group_id": null,





  
  

  48
  
      "shape_type": "polygon",





  
  

  49
  
      "flags": {}





  
  

  50
  
    }





  
  

  51
  
    ...





  
  

  52
  
  ],





  
  

  53
  
  "imagePath": "path/to/image.jpg",  // Image file path





  
  

  54
  
  "imageData": null,                 // Image data (Base64 encoded)





  
  

  55
  
  "imageHeight": 480,                // Image height





  
  

  56
  
  "imageWidth": 640                  // Image width





  
  

  57
  
}





  
  

  58
  





  
  

  59
  





  
  

  60
  





  
  

  61
  





  
  

  62
  





  
  

  63
  
YOLO .txt format:





  
  

  64
  
one object -> one row





  
  

  65
  
(classID, x_center, y_center, width_normalized, height_normalized)





  
  

  66
  





  
  

  67
  
(eg):





  
  

  68
  
1 0.375000 0.625000 0.312500 0.416667





  
  

  69
  
0 0.656250 0.291667 0.156250 0.208333





  
  

  70
  





  
  

  71
  





  
  

  72
  





  
  

  73
  





  
  

  74
  





  
  

  75
  
dataset.yaml format(eg):





  
  

  76
  
train: /path/to/your/dataset/images/train





  
  

  77
  
val: /path/to/your/dataset/images/val





  
  

  78
  
nc: 5





  
  

  79
  
names: ['cat', 'dog', 'person', 'car', 'truck']