File size: 4,892 Bytes
fced272 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 | import torch
import numpy as np
class ClinicalContextVectorizer:
"""
将来自 clinical_latest.json 的 context_static 对象转换为一个扁平的数值向量。
"""
def __init__(self, vector_dim: int, config: dict = None):
self.vector_dim = vector_dim
self.config = config or self._get_default_config()
# 为分类特征构建映射
self.feature_map = {}
self.vector_layout = {}
current_idx = 0
for feature, details in self.config.items():
if details['type'] == 'categorical':
for value in details['values']:
self.feature_map[f"{feature}_{value}"] = current_idx
current_idx += 1
self.vector_layout[feature] = {'start': self.feature_map[f"{feature}_{details['values'][0]}"] , 'end': current_idx}
elif details['type'] == 'numerical':
self.feature_map[feature] = current_idx
self.vector_layout[feature] = {'start': current_idx, 'end': current_idx + 1}
current_idx += 1
elif details['type'] == 'binary_group':
for sub_feature in details['values']:
self.feature_map[sub_feature] = current_idx
current_idx += 1
self.vector_layout[feature] = {'start': self.feature_map[details['values'][0]], 'end': current_idx}
# 确保最终维度匹配
if current_idx != self.vector_dim:
raise ValueError(f"Configured vector dimension ({current_idx}) does not match expected dimension ({self.vector_dim})")
def _get_default_config(self):
# 这个配置应该与你的模型定义中的 clinical_input_dim 匹配
# 这里的维度是 3 (sex) + 4 (race) + 1 (age) + 1 (grade) + 12 (genomics) = 21
return {
"sex_at_birth": {"type": "categorical", "values": ["male", "female", "unknown"]},
"race": {"type": "categorical", "values": ["white", "black or african american", "asian", "unknown"]},
"age_at_diagnosis_years": {"type": "numerical", "norm_mean": 60, "norm_std": 15},
"who_grade": {"type": "numerical", "norm_mean": 3.5, "norm_std": 0.5},
"genomics": {
"type": "binary_group",
"values": [
"idh1", "idh2", "atrx", "mgmt_methylation", "braf_v600e",
"tert_promoter", "chr7_gain_chr10_loss", "h3_3a", "egfr_amp",
"pten", "cdkn2ab_deletion", "tp53_alteration"
]
}
}
def vectorize(self, context_static: dict) -> torch.Tensor:
"""将单个 context_static 字典转换为 Tensor"""
vec = torch.zeros(self.vector_dim)
# 处理顶层特征
for feature, details in self.config.items():
if feature == "genomics":
continue
value = context_static.get(feature)
if details['type'] == 'categorical':
value = value if value in details['values'] else 'unknown'
key = f"{feature}_{value}"
if key in self.feature_map:
vec[self.feature_map[key]] = 1.0
elif details['type'] == 'numerical':
if value is not None:
# 归一化
norm_val = (value - details['norm_mean']) / details['norm_std']
vec[self.feature_map[feature]] = norm_val
# 处理基因组学特征
genomics_data = context_static.get("genomics", {})
genomics_config = self.config.get("genomics", {})
if genomics_config:
for sub_feature in genomics_config['values']:
value = genomics_data.get(sub_feature)
if value is not None and sub_feature in self.feature_map:
# 0: no, 1: yes, 2: unknown. 我们将 unknown 映射为 0.5
if value == 0:
vec[self.feature_map[sub_feature]] = 0.0
elif value == 1:
vec[self.feature_map[sub_feature]] = 1.0
elif value == 2:
vec[self.feature_map[sub_feature]] = 0.5 # 代表未知
return vec
if __name__ == '__main__':
# 示例用法
vectorizer = ClinicalContextVectorizer(vector_dim=20)
sample_context = {"sex_at_birth": "female", "race": "white", "age_at_diagnosis_years": 57.0, "who_grade": 4, "genomics": {"idh1": 0, "idh2": 0, "atrx": 2, "mgmt_methylation": 2, "braf_v600e": 2, "tert_promoter": 2, "chr7_gain_chr10_loss": 2, "h3_3a": 2, "egfr_amp": 2, "pten": 0, "cdkn2ab_deletion": 0, "tp53_alteration": 0}}
tensor_vec = vectorizer.vectorize(sample_context)
print(tensor_vec)
print(tensor_vec.shape) |