Breast Cancer Diagnosis Tool

By Rachit Singh

This project uses Logistic Regression on the Wisconsin Breast Cancer Dataset (569 samples, 30 features) to predict tumor malignancy, integrated into a Streamlit app for real-time diagnosis.

Breast Cancer Diagnosis Tool

This project applies Logistic Regression to predict whether a tumor is malignant or benign using the Wisconsin Breast Cancer Dataset. With 30 features from 569 observations, the model is integrated into a Streamlit web app for real-time tumor diagnosis.

1. Function load_and_prepare_data

Algorithm:

1. Load dataset $D$ from CSV.
2. Drop unnecessary columns ('Unnamed: 32', 'id').
3. Map 'diagnosis' column values ('M', 'B') to numerical values (1, 0).
4. Return D′ (processed dataframe).

2. Function add_sidebar

Algorithm:

1. For each column $k$ in D′:
   • Create a slider in the sidebar.
   • Set the range as [min, max] of $k$, and the default as mean of k.
2. Store slider values in input_dict.
3. Return input_dict.

3. Function get_scaled_values

Algorithm:

1. For each key-value pair (k,x) in $\text{input\_dict}$:
   • Calculate min(X_k) and max(X_k).
   • Scale $x$ using the min-max scaling formula.
2. Return $\text{scaled\_dict}$ containing scaled values.

4. Function get_radar_chart

Algorithm:

1. Group input values into R_mean, R_se, R_worst.
2. Create a radar chart with:
   • R_mean, R_se, R_worst as radii.
   • C as angular axes.
3. Return the radar chart.

5. Function add_predictions

Algorithm:

1. Load model and scaler from saved files.
2. Transform input_data using the scaler.
3. Use the model to predict y and probabilities P.
4. Display the result (Benign or Malignant) and probabilities

Important Note: This application is designed to support medical professionals in the diagnostic process; however, it is not intended to replace a professional medical diagnosis.

(Visit the Project)