What 569 Tumor Data Samples Reveal About Detecting Breast Cancer

By gerald · December 30, 2025

Breast cancer is the most common cancer among women worldwide. Early detection dramatically improves survival rates but what actually distinguishes a…

The analysis identified which mean features show the largest percentage differences between malignant and benign breast cancer cases. While there was a minor error in printing all the detailed results, the visualization and data table were successfully generated showing the complete comparison across all mean features.

The ratio of malignant to benign tumors in the dataset is approximately 0.59:1. This means for every benign tumor, there are about 0.59 malignant tumors. In other words, benign tumors outnumber malignant ones in this dataset.

Yes, texture mean differs significantly between malignant and benign diagnoses. The statistical analysis shows a highly significant difference (p-value = 4.06e-25), meaning we can be extremely confident this difference is real and not due to chance. Malignant tumors have an average texture mean of 21.60, while benign tumors average 17.91 — a difference of about 3.69 units.

The analysis compares smoothness mean values across the two diagnosis categories (Malignant and Benign). I've created two visualizations to help you understand the differences: a box plot showing the distribution of smoothness values for each diagnosis type, and an overlaid histogram comparing how the values are spread across both categories. The box plot uses red for malignant (M) and blue for benign (B) cases, making it easy to see the central tendencies and spread of each group.

Yes, malignant tumors definitely show higher standard error values than benign tumors for both radius se and area se measurements. Malignant tumors have an average radius se of 0.61 compared to 0.28 for benign (114% higher), and an average area se of 72.67 compared to 21.14 for benign (244% higher). This significant difference indicates that malignant tumor cells exhibit much greater variability in their measurements.

Both concavity mean and concave points mean show strong positive correlations with malignant diagnosis. Concave points mean has the stronger correlation at r=0.78, while concavity mean shows r=0.70. This means that as these values increase, the likelihood of a malignant diagnosis increases significantly.

Yes, symmetry mean is a moderately reliable indicator for distinguishing between malignant and benign diagnoses. The analysis shows a statistically significant difference between the two groups (p-value < 0.000001), with malignant tumors having a higher average symmetry mean (0.193) compared to benign tumors (0.174). The effect size (Cohen's d = 0.71) indicates meaningful separation between the groups, though there is still some overlap.

Both 'worst' measurements effectively separate malignant from benign tumors, but area worst is the best separator with a Cohen's d of 2.29, followed by radius worst at 2.19. Both show excellent separation power (d 0.8), meaning malignant tumors have significantly larger worst-case measurements than benign tumors.

There are clear and significant differences between malignant and benign tumors across all three measurements. Malignant tumors consistently show larger values: the average radius is 17.46 for malignant vs 12.15 for benign, the average area is 978.38 vs 462.79, and the average perimeter is 115.37 vs 78.08. These differences make these features strong indicators for distinguishing between cancer types.

Based on the analysis, tumors with a radius mean of 15 or greater and an area mean of 705 or greater have a significantly higher likelihood of being malignant. When both thresholds are exceeded, 94% of cases (158 out of 168) are malignant, making these reliable indicators for identifying potentially cancerous tumors.

Compactness mean is strongly related to tumor classification. Malignant tumors have significantly higher compactness values compared to benign tumors. On average, malignant tumors have a compactness mean of 0.145, while benign tumors average only 0.080 - that's an 81% difference. This makes compactness mean a powerful indicator for distinguishing between malignant and benign classifications.

Based on 357 benign tumor samples, here are the typical ranges for the three features you asked about:

Radius Mean: Most benign tumors fall between 11.08 and 13.37 (typical range), with an average of 12.15. The full range spans from 6.98 to 17.85.

Texture Mean: The typical range is 15.15 to 19.76, with an average of 17.91. Values can range from 9.71 to 33.81.

Area Mean: Benign tumors typically have areas between 378.20 and 551.10, averaging 462.79. The full range goes from 143.50 to 992.10.

A box plot visualization has been generated to help you see these distributions clearly.

The analysis examined 212 malignant tumors to determine typical ranges for radius mean, texture mean, and area mean. A box plot visualization was created showing the distribution of these features, along with detailed statistical tables. The visualization clearly displays how these measurements vary in malignant tumors.

No, fractal dimension mean does NOT provide meaningful differentiation between malignant and benign diagnoses. The statistical analysis shows virtually identical mean values for both groups (Malignant: 0.0627, Benign: 0.0629), with a high p-value of 0.76 indicating no statistically significant difference. The effect size (Cohen's d = -0.027) is negligible, meaning this feature has essentially no practical value for distinguishing between diagnosis categories.

Yes, the ratio of perimeter mean to area mean differs significantly between malignant and benign tumors. The statistical test confirms this with an extremely low p-value (5.44e-77), meaning this difference is highly unlikely to occur by chance.

The analysis examined correlations between mean features in 212 malignant breast cancer cases. Two visualizations were created: a correlation heatmap showing relationships between all 10 mean features, and a bar chart highlighting the top 10 most correlated feature pairs. While the detailed list encountered a minor technical issue, the visualizations successfully display the correlation patterns.