liveBook · Manning

Cancer

As explained in Deep Learning with PyTorch, Second Edition

The property of a tumor being cancerous and capable of harmful growth, as distinguished from benign tissue.

Cancer describes a malignant tumor: harmful abnormal tissue that can be difficult to distinguish from healthy anatomy in medical images. section 11.4.1

In a CT scan, cancer may occupy an extremely small fraction of the data. Even when a tumor exists, up to 99.9999% of voxels may not be cancer, making detection from images especially subtle. section 11.4.1

Finding one nodule

One CT scan contains roughly 1,000 structures that may look like tumors to an untrained eye, while a specialist identifies exactly one as a nodule. A close-up multislice crop and a malignant nodule illustrate why location and visual differences matter. section 11.4.1

Figure 11.5 A CT scan with approximately 1,000 structures that look like tumors to the untrained eye. Exactly one has been identified as a nodule when reviewed by a human specialist. The rest are normal anatomical structures, such as blood vessels, lesions, and other nonproblematic lumps.
figure

Use the figure to compare many tumor-like structures with the single nodule identified by a specialist. section 11.4.1

How Deep Learning with PyTorch, Second Edition covers it

How deep it goes
Explains extreme voxel imbalance and visual subtlety, but stops short of defining cancer biology or diagnostic criteria.
Examples
Uses CT scans, a specialist-identified nodule, multislice crops, and a visually discrepant malignant nodule.
What you can do after
Evaluate why cancer detection from complete CT images requires specialized modeling rather than indiscriminate data training.
How it connects
Connects cancer with malignant tumors, nodules, CT scans, object detection, and image classification.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with PyTorch, Second Edition. The section links point to the parts of the book it draws on.