This dataset contains clinical, genomic, imaging, biomarker, symptom, treatment, and outcome data for 50,000 breast cancer patients, designed to support research into metastatic breast cancer prediction, risk stratification, and treatment response modeling.

File: oncology_metastatic_breast_cancer_50k-1.csv Records: 50,000 patients Features: 92 columns Target variable: Metastatic_Label (binary: 0 = non-metastatic, 1 = metastatic)

Class Distribution

Label	Count	Percentage
Non-metastatic (0)	41,250	82.5%
Metastatic (1)	8,750	17.5%

The dataset is imbalanced. Consider SMOTE, class weighting, or stratified sampling when training classifiers.

Data Sources

Source	Records
Oncology Center	14,271
Breast Imaging Clinic	11,884
Diagnostic Lab	9,567
Genomic Testing Facility	7,070
Longitudinal Monitoring Program	4,826

Column Reference

Patient Demographics

Column	Type	Description
`Patient_ID`	string	Unique patient identifier (e.g., PT000001)
`Age`	int	Patient age in years (range: 22–90, mean: ~54)
`Gender`	string	Patient gender
`Ethnicity`	string	Patient ethnicity
`Menopausal_Status`	int (0/1)	0 = pre-menopausal, 1 = post-menopausal
`BMI`	float	Body mass index

Risk Factors & Family History

Column	Type	Description
`BRCA1_Mutation`	float (0/1)	BRCA1 gene mutation status
`BRCA2_Mutation`	float (0/1)	BRCA2 gene mutation status
`TP53_Mutation`	float (0/1)	TP53 gene mutation status
`Family_History_Breast_Cancer`	float (0/1)	Family history of breast cancer
`Genetic_Predisposition_Score`	float	Composite genetic risk score
`Smoking_History`	string	Smoking status (Never / Former / Current)
`Alcohol_Use`	string	Alcohol consumption level
`Physical_Activity_MET_hrs_wk`	float	Physical activity in MET-hours per week
`Hormonal_Contraceptive_Use`	float (0/1)	History of hormonal contraceptive use
`HRT_Use`	float (0/1)	Hormone replacement therapy use
`Diet_Quality_Score`	float	Diet quality score
`Prior_Radiation_Exposure`	float (0/1)	Prior radiation exposure
`Prior_Benign_Breast_Disease`	float (0/1)	History of benign breast disease
`Previous_Cancer`	float (0/1)	History of any prior cancer
`Chronic_Inflammation`	float (0/1)	Presence of chronic inflammation
`Diabetes`	float (0/1)	Diabetes diagnosis

Tumor Pathology

Column	Type	Description
`Molecular_Subtype`	string	Luminal_A / Luminal_B / HER2_Enriched / Triple_Negative / Other
`Histological_Subtype`	string	IDC (invasive ductal) / ILC (invasive lobular)
`Tumor_Size_mm`	float	Tumor size in millimeters
`Tumor_Grade`	float	Histological grade (1–3)
`ER_Status`	float (0/1)	Estrogen receptor status
`PR_Status`	float (0/1)	Progesterone receptor status
`HER2_Status`	float (0/1)	HER2 receptor status
`Ki67_Index_Pct`	float	Ki-67 proliferation index (%)
`Lymphovascular_Invasion`	float (0/1)	Presence of lymphovascular invasion
`Lymph_Node_Count_Positive`	float	Number of positive lymph nodes
`Lymph_Node_Involvement`	float (0/1)	Any lymph node involvement
`TNM_Stage`	float	TNM staging (1–4)

Imaging

Column	Type	Description
`Mammography_BIRADS_Score`	float	BIRADS score from mammography (1–6)
`MRI_Lesion_Characteristic`	string	MRI lesion morphology description
`Ultrasound_Morphology`	string	Ultrasound lesion shape
`Calcification_Pattern`	string	Calcification classification
`Tumor_Density_Score`	float	Tumor density on imaging
`Suspicious_Lymph_Node_Imaging_Flag`	float (0/1)	Suspicious lymph nodes on imaging

Serum Biomarkers

Column	Type	Description
`CA15_3_U_mL`	float	CA 15-3 tumor marker (U/mL)
`CEA_ng_mL`	float	Carcinoembryonic antigen (ng/mL)
`CTC_Count_per_7_5mL`	float	Circulating tumor cells per 7.5 mL blood
`ctDNA_Level_ng_mL`	float	Circulating tumor DNA level (ng/mL)
`ctDNA_Trend`	string	ctDNA trend direction (Rising / Stable / Falling)
`CRP_mg_L`	float	C-reactive protein (mg/L)
`ESR_mm_hr`	float	Erythrocyte sedimentation rate (mm/hr)
`Hemoglobin_g_dL`	float	Hemoglobin (g/dL)
`WBC_x10_9_L`	float	White blood cell count (×10⁹/L)
`ALT_U_L`	float	Alanine aminotransferase (U/L)
`AST_U_L`	float	Aspartate aminotransferase (U/L)
`ALP_U_L`	float	Alkaline phosphatase (U/L)

Genomics

Column	Type	Description
`Gene_Expression_Risk_Score`	float	Gene expression-based risk score
`Tumor_Mutation_Burden`	float	Tumor mutational burden
`Copy_Number_Variation_Flag`	float (0/1)	Presence of copy number variations
`RNA_Biomarker_Signature_Score`	float	RNA-based biomarker signature score

Symptoms

Column	Type	Description
`Symptom_Breast_Lump`	float (0/1)	Palpable breast lump
`Symptom_Nipple_Discharge`	float (0/1)	Nipple discharge
`Symptom_Skin_Dimpling`	float (0/1)	Skin dimpling
`Symptom_Breast_Pain`	float (0/1)	Breast pain
`Symptom_Fatigue`	float (0/1)	Fatigue
`Symptom_Bone_Pain`	float (0/1)	Bone pain
`Symptom_Unexplained_Weight_Loss`	float (0/1)	Unexplained weight loss
`Symptom_Dyspnea`	float (0/1)	Shortness of breath

Metastasis

Column	Type	Description
`Metastatic_Label`	int (0/1)	Primary target: metastatic (1) vs non-metastatic (0)
`Metastasis_Site`	string	Site of metastasis (Multi_Site, or organ-specific); mostly NaN for label=0
`Micro_Metastatic_Spread_Flag`	float (0/1)	Micro-metastatic spread detected
`Bone_Metastasis_Likelihood`	float	Predicted probability of bone metastasis
`Liver_Metastasis_Likelihood`	float	Predicted probability of liver metastasis
`Lung_Metastasis_Likelihood`	float	Predicted probability of lung metastasis
`Brain_Metastasis_Likelihood`	float	Predicted probability of brain metastasis

Treatment

Column	Type	Description
`Surgery_Type`	string	Type of surgery (e.g., Mastectomy, Lumpectomy)
`Chemotherapy_Regimen`	string	Chemotherapy regimen used (e.g., AC-T, Docetaxel)
`Hormone_Therapy`	float (0/1)	Hormone therapy administered
`Targeted_Therapy_HER2`	float (0/1)	HER2-targeted therapy administered
`Radiation_Therapy`	float (0/1)	Radiation therapy administered

Longitudinal Monitoring & Outcomes

Column	Type	Description
`Follow_Up_Interval_Months`	float	Follow-up interval in months
`Tumor_Size_Change_mm`	float	Change in tumor size since last visit (mm)
`CA15_3_Trend`	string	CA 15-3 trend (Rising / Stable / Falling)
`ctDNA_Trajectory`	string	ctDNA trajectory over follow-up
`Imaging_Progression_Flag`	float (0/1)	Imaging-confirmed disease progression
`Treatment_Response_Trajectory`	string	Trajectory of treatment response
`Time_To_Metastasis_Months`	float	Time from diagnosis to metastasis (months)
`Recurrence_Event_Count`	float	Number of recurrence events
`Last_Recurrence_Months_Ago`	float	Months since last recurrence
`Recurrence_Risk_Score`	float	Estimated recurrence risk score
`Survival_Probability_12mo`	float	Estimated 12-month survival probability
`Survival_Probability_24mo`	float	Estimated 24-month survival probability
`Survival_Probability_60mo`	float	Estimated 60-month survival probability
`Treatment_Response`	string	Overall treatment response category
`Diagnostic_Uncertainty_Flag`	float (0/1)	Flagged for diagnostic uncertainty
`Data_Source`	string	Originating clinical data source

Missing Data

Several columns have notable missingness:

Column	Missing Count	Notes
`Metastasis_Site`	~41,532	Expected — only populated for metastatic cases
`Last_Recurrence_Months_Ago`	~33,001	Only relevant for patients with recurrence
`Alcohol_Use`	~18,488	Self-reported, frequently missing
`Chemotherapy_Regimen`	~17,369	Not all patients receive chemotherapy
`Calcification_Pattern`	~14,985	Imaging finding, not always present
`Time_To_Metastasis_Months`	~10,461	Only for metastatic patients
`Surgery_Type`	~9,205	Not all patients undergo surgery

Suggested Use Cases

Metastasis prediction — binary classification using Metastatic_Label
Survival analysis — using Survival_Probability_12mo/24mo/60mo and Time_To_Metastasis_Months
Treatment response modeling — using Treatment_Response and Treatment_Response_Trajectory
Site-specific metastasis risk — using bone/liver/lung/brain likelihood scores
Biomarker research — exploring ctDNA, CA 15-3, and CTC as early metastasis signals
Multi-modal fusion — combining imaging, genomics, and clinical features

Quick Start

import pandas as pd

df = pd.read_csv("oncology_metastatic_breast_cancer_50k-1.csv")

print(df.shape)          # (50000, 92)
print(df["Metastatic_Label"].value_counts())
print(df.isnull().sum().sort_values(ascending=False).head(10))

Notes

All patient IDs are anonymized (format: PT######).
Binary flags are encoded as 0.0 / 1.0 floats; convert with .astype(int) after handling nulls.
The Metastasis_Site column is NaN for the vast majority of records (non-metastatic patients) — treat as a conditional field, not a missing-data problem.
Survival probability columns are model-estimated values, not observed outcomes.

Downloads last month: 6

Total file size:

23.6 MB

Collection including Auric-Grid/E1L2-Early-Metastatic-Breast-Cancer

Edition 1

Collection

Users can get access to these datasets for free. • 11 items • Updated Jul 2 • 1