Anecdotal example of how this model could be used for a new patient in observational mode. A 67-year-old male was evaluated for complaints of lower urinary tract symptoms for the past 3 months in June 2025. On evaluation, his Prostate Specific Antigen (PSA) was evaluated, and biopsy was reported as acinar adenocarcinoma (GS 4+3). A PSMA PET scan done in August 2025 showed locally advanced carcinoma of the prostate and a well-defined soft tissue density lesion measuring 19 mm × 28.5 mm in the anterior mediastinum abutting the ascending thoracic aorta with no fat/calcified components (incidental finding). Based on these incidental findings, the thoracic lesion, which could be a thymic epithelial tumor (TET), was evaluated using the Radiomics8 risk model.
Output from our prediction model with UQ
Model Predicted Risk Class	Prob. of HR Class (%)	Dist. to Nearest LR	Dist. to Nearest HR	LR UQ Score	HR UQ Score	Final Risk Class
LR	10	0.21	0.99	0.04	4.76	LR
The model predicted this patient to be low risk, with a probability of being high risk of 10%. Based on the distance metric, the patient’s features were ‘closer (0.21)’ to LR examples than the HR (0.99) examples in the training cohort. The final lower UQ score for LR (0.04) indicates that this is most likely to be LR and hence was labelled so in the final risk class. Implementing the TLP (Figure 3) by using a simple threshold (>75% = uncertainty high; 50–75% = uncertainty medium; <50% = uncertainty low), this patient’s distance (LR = 0.21) when compared with Table 1 was categorized in the amber category = uncertainty medium—exercise caution. The clinicians received the predicted risk and the degree of uncertainty in the classification in a two-step report. This aligned the multi-disciplinary team’s decision to go ahead with prostate cancer treatment and to take a ‘wait and watch’ strategy regarding the TET, given the patient’s age.