AI-based Measurements of Tumour Burden in PSMA PET-CT

In Sweden, prostate cancer is diagnosed in 10,000 men annually and the mortality rate of 2,400 is among the highest worldwide. Some prostate cancers are at high risk of metastatic progression to lethal disease and require correct staging or detection of recurrence and multidisciplinary treatments.

The investigators have developed an AI-based method to detect and quantify tumours and metastases in 18F-PSMA-1007 PET-CT scans in patients with prostate cancer. The method can find tumours in the prostate and metastases in pelvic lymph nodes, distant lymph nodes and in bone, both in patients referred to the PET-CT scan for primary staging of high-risk prostate cancer for secondary staging due to recurrence.

Patients referred to clinically indicated PSMA PET-CT due to either initial staging of primary high-risk prostate cancer or due to biochemical recurrence will be eligible for inclusion. The AI-based method will automatically calculate TV, TU and number of suspected lesions and this information will be stored in a database. The values will after a 5 year follow-up period be analysed with regard to overall survival (OS) and progression-free survival (PFS).

The primary aim of the present study is to evaluate how tumour burden, measured as TV and as tumour uptake (TU: TV x SUVmean) in the prostate/prostate bed, pelvic lymph nodes, distant lymph nodes, bone and as the total tumour burden predicts overall survival (OS) in patients with prostate cancer (newly diagnosed and patients with biochemical recurrence). A secondary aim is to evaluate how the AI-derived measurements predict time to biochemical recurrence in a sub-cohort of patients with newly diagnosed high-risk prostate cancer. Tertiary aims are to evaluate the difference in TV and TU measured with two different segmentation methods (a threshold of 50% of SUVmax in each lesion and a threshold of SUV 4) in relation to OS and biochemical PFS. The impact of the number of automatically calculated suspected lesions will also be investigated regarding OS and biochemical PFS as well as to the difference in tumour burden measured with AI and manually.

• Tumour burden (cm3) in relation to biochemical recurrence [Time frame: 5 years]
• Number of tumours/metastases in relation to OS [Time frame: 5 years]
• Comparing two different segmentation methods in relation to OS [Time frame: 5 years]
• Comparing total tumour burden (cm3) measured manually and by the AI-based mehtod [Time frame: 5 years]