Tracking Prostate Cancer Metastases in Scans express

Original: "Towards Routine AI-Based PET/CT and SPECT/CT Lesion Segmentation and Tracking in PSMA Theranostics"
arXiv:2603.22666 · 2026-03-24 · CC BY 4.0 · 1 min
Automated scan comparison reveals how each tumor responds to treatment.

📄 Showing the "Simple" version — "Popular" is not ready yet. Add it to favorites to help prioritize it.

When treating prostate cancer, doctors often see not one tumor but many lesions in bones and lymph nodes. Each lesion can respond to therapy differently: one shrinks, another stays the same, a third grows. Previously, the disease was assessed as a whole, like judging the weather by the average temperature over a year. Now it's possible to track each lesion individually.

This is done using special scans: PET/CT with a substance that attaches to prostate cancer cells, and SPECT/CT after administering a therapeutic radioactive drug. A computer program identifies all suspicious areas on the scans and then matches them across different examinations, like an air traffic controller tracking each plane from takeoff to landing.

Previously, manual scan comparison took hours and was imprecise. Now, the algorithm does it in minutes and doesn't miss subtle changes.

This approach helps doctors notice in time when a tumor stops responding to treatment and adjust their strategy. It also allows for more accurate radiation dose calculation for each lesion, sparing healthy tissue. Ultimately, treatment becomes more personal: not "against cancer in general," but against specific lesions in a specific person.

PET/CT and SPECT/CT are like two sides of the same coin: the first scan shows where the cancer is hiding, and the second shows how much drug has reached it.

In Star Trek, the medical tricorder scans the body and instantly finds all problems. Similar algorithms are gradually entering real-world oncology.

Concepts
tumor segmentation medical imaging oncology radiotherapy ai planning tumor imaging characterization medical image segmentation PI-RADS classification computed tomography spectral analysis ai personalized medicine positron emission tomography prostate MRI protein language model brain cancer detection therapy biomedical spectroscopy mass spectrometry benchmark precision measurement spectroscopy ct scan miniaturized spectrometry