Deep learning performance on MRI prostate gland segmentation: evaluation of two commercially available algorithms compared with an expert radiologist.

Purpose: Accurate whole-gland prostate segmentation is crucial for successful ultrasound-MRI fusion biopsy, focal cancer treatment, and radiation therapy techniques. Commercially available artificial intelligence (AI) models, using deep learning algorithms (DLAs) for prostate gland segmentation, are rapidly increasing in numbers. Typically, their performance in a true clinical context is scarcely examined or published. We used a heterogenous clinical MRI dataset in this study aiming to contribute to validation of AI-models. Approach: We included 123 patients in this retrospective multicenter (7 hospitals), multiscanner (8 scanners, 2 vendors, 1.5T and 3T) study comparing prostate contour assessment by 2 commercially available Food and Drug Association (FDA)-cleared and CE-marked algorithms (DLA1 and DLA2) using an expert radiologist's manual contours as a reference standard (RSexp) in this clinical heterogeneous MRI dataset. No in-house training of the DLAs was performed before testing. Several methods for comparing segmentation overlap were used, the Dice similarity coefficient (DSC) being the most important. Results: The DSC mean and standard deviation for DLA1 versus the radiologist reference standard (RSexp) was 0.90±0.05 and for DLA2 versus RSexp it was 0.89±0.04. A paired t-test to compare the DSC for DLA1 and DLA2 showed no statistically significant difference (p=0.8). Conclusions: Two commercially available DL algorithms (FDA-cleared and CE-marked) can perform accurate whole-gland prostate segmentation on a par with expert radiologist manual planimetry on a real-world clinical dataset. Implementing AI models in the clinical routine may free up time that can be better invested in complex work tasks, adding more patient value.

A pilot study of an organised population-based testing programme for prostate cancer.

OBJECTIVE: To determine the feasibility of a digitally automated population-based programme for organised prostate cancer testing (OPT) in Southern Sweden. PATIENTS AND METHODS: A pilot project for a regional OPT was conducted between September 2020 and February 2021, inviting 999 randomly selected men aged 50, 56, or 62 years. Risk stratification was based on prostate-specific antigen (PSA) level, PSA density (PSAD), and bi-parametric prostate magnetic resonance imaging (MRI). Men with a PSA level of 3-99 ng/mL had an MRI, and men with elevated PSA level (≥3 ng/mL) had a urological check-up, including a digital rectal examination and transrectal ultrasonography (TRUS). Indications for targeted and/or systematic transrectal prostate biopsies were suspicious lesions on MRI (Prostate Imaging-Reporting and Data System [PI-RADS] 4-5) and/or PSAD > 0.15 ng/mL/mL. Additional indications for prostate biopsies were palpable tumours, PSA ratio < 0.1, or cancer suspicion on TRUS. Patient selection, mail correspondence, data collection, and algorithm processing were performed by an automated digital management system. Feasibility is reported descriptively. RESULTS: A total of 418 men had a PSA test (42%), with increasing participation rates by age (50 years, 38%; 56 years, 44%; and 62 years, 45%). Among these, 35 men (8%) had elevated PSA levels (≥3 ng/mL: one of 139, aged 50 years; 10/143, aged 56 years; and 24/146, aged 62 years). On MRI, 16 men (48%) had a negative scan (PI-RADS < 3), seven men (21%) had PI-RADS 3, nine men (27%) had PI-RADS 4, and one man (3%) had PI-RADS 5. All men with PI-RADS 4 or 5 underwent prostate biopsies, as well as two men with PI-RADS 3 due to PSAD > 0.15 ng/mL/mL or a suspicious finding on TRUS. Prostate cancer was diagnosed in 10 men. Six men underwent active treatment, whereas four men were assigned to active surveillance. CONCLUSION: Our OPT model is feasible from an operational point of view, but due to the limited scale of this study no conclusions can be made regarding the efficacy of the diagnostic model or outcome.

Population-based Organised Prostate Cancer Testing: Results from the First Invitation of 50-year-old Men.

BACKGROUND: The European Union recently recommended evaluation of the feasibility of organised prostate cancer screening. In Sweden, regional population-based organised prostate cancer testing (OPT) programmes were introduced in 2020. OBJECTIVE: To describe initial participation rates and diagnostic outcomes. DESIGN, SETTING, AND PARTICIPANTS: The three most populated Swedish regions invited all men aged 50 yr to OPT by a letter in 2020-2022. Men with prostate-specific antigen (PSA) ≥3 ng/ml were referred for prostate magnetic resonance imaging (MRI). PSA assays differed across regions. Men with Prostate Imaging Reporting and Data System (PI-RADS) 1-3 and PSA density ≥0.15 ng/ml/cm3 or PI-RADS 4-5 were referred for a biopsy. Data were obtained from the Swedish Register for Organised Prostate Cancer Testing. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Overall and regional participation rates, PSA distributions, PI-RADS score distributions, cancer detection, and treatment were evaluated. RESULTS AND LIMITATIONS: A total of 23 855 (35%) of 68 060 invited men participated; 696 (2.9%) had PSA ≥3 ng/ml, and of them, 306 (44%) had a biopsy indication and 221 (32%) had a biopsy. On biopsy, 93 (42%) had Gleason grade group ≥2 (0.39% of PSA-tested men) and 44 (20%) Gleason grade group 1 cancer. Most men with cancer had treatment with curative intent (70%) or were under active surveillance (28%). Across regions, proportions of men with PSA ≥3 ng/ml ranged from 2.3% to 4.0%, and those with PI-RADS score 4-5 ranged from 12% to 21%. A limitation is that results are applicable only to first testing of men in their early 50s. CONCLUSIONS: The OPT programmes are feasible with good compliance to the diagnostic pathway. The use of MRI and PSA density avoided a biopsy for over half of the men with PSA ≥3 ng/ml. Inter-regional differences in diagnostic outcomes show a need for standardisation of the diagnostic pathway's components. PATIENT SUMMARY: We report the diagnostic outcomes of inviting 68 000 50-yr-old men to organised prostate cancer testing.

Correlation between ADC, ADC ratio, and Gleason Grade group in prostate cancer patients undergoing radical prostatectomy: Retrospective multicenter study with different MRI scanners.

Background: MRI is an important tool in the prostate cancer work-up, with special emphasis on the ADC sequence. This study aimed to investigate the correlation between ADC and ADC ratio compared to tumor aggressiveness determined by a histopathological examination after radical prostatectomy. Methods: Ninety-eight patients with prostate cancer underwent MRI at five different hospitals prior to radical prostatectomy. Images were retrospectively analyzed individually by two radiologists. The ADC of the index lesion and reference tissues (contralateral normal prostatic, normal peripheral zone, and urine) was recorded. Absolute ADC and different ADC ratios were compared to tumor aggressivity according to the ISUP Gleason Grade Groups extracted from the pathology report using Spearman's rank correlation coefficient (ρ). ROC curves were used to evaluate the ability to discriminate between ISUP 1-2 and ISUP 3-5 and intra class correlation and Bland-Altman plots for interrater reliability. Results: All patients had prostate cancer classified as ISUP grade ≥ 2. No correlation was found between ADC and ISUP grade. We found no benefit of using the ADC ratio over absolute ADC. The AUC for all metrics was close to 0.5, and no threshold could be extracted for prediction of tumor aggressivity. The interrater reliability was substantial to almost perfect for all variables analyzed. Conclusions: ADC and ADC ratio did not correlate with tumor aggressiveness defined by ISUP grade in this multicenter MRI study. The result of this study is opposite to previous research in the field.

Deep learning algorithm performs similarly to radiologists in the assessment of prostate volume on MRI.

OBJECTIVES: Prostate volume (PV) in combination with prostate specific antigen (PSA) yields PSA density which is an increasingly important biomarker. Calculating PV from MRI is a time-consuming, radiologist-dependent task. The aim of this study was to assess whether a deep learning algorithm can replace PI-RADS 2.1 based ellipsoid formula (EF) for calculating PV. METHODS: Eight different measures of PV were retrospectively collected for each of 124 patients who underwent radical prostatectomy and preoperative MRI of the prostate (multicenter and multi-scanner MRI's 1.5 and 3 T). Agreement between volumes obtained from the deep learning algorithm (PVDL) and ellipsoid formula by two radiologists (PVEF1 and PVEF2) was evaluated against the reference standard PV obtained by manual planimetry by an expert radiologist (PVMPE). A sensitivity analysis was performed using a prostatectomy specimen as the reference standard. Inter-reader agreement was evaluated between the radiologists using the ellipsoid formula and between the expert and inexperienced radiologists performing manual planimetry. RESULTS: PVDL showed better agreement and precision than PVEF1 and PVEF2 using the reference standard PVMPE (mean difference [95% limits of agreement] PVDL: -0.33 [-10.80; 10.14], PVEF1: -3.83 [-19.55; 11.89], PVEF2: -3.05 [-18.55; 12.45]) or the PV determined based on specimen weight (PVDL: -4.22 [-22.52; 14.07], PVEF1: -7.89 [-30.50; 14.73], PVEF2: -6.97 [-30.13; 16.18]). Inter-reader agreement was excellent between the two experienced radiologists using the ellipsoid formula and was good between expert and inexperienced radiologists performing manual planimetry. CONCLUSION: Deep learning algorithm performs similarly to radiologists in the assessment of prostate volume on MRI. KEY POINTS: • A commercially available deep learning algorithm performs similarly to radiologists in the assessment of prostate volume on MRI. • The deep-learning algorithm was previously untrained on this heterogenous multicenter day-to-day practice MRI data set.

Designing and Implementing a Population-based Organised Prostate Cancer Testing Programme.

BACKGROUND: European guidelines recommend that well-informed men at elevated risk of having prostate cancer (PCa) should be offered prostate-specific antigen (PSA) testing with risk-stratified follow-up. The Swedish National Board of Health and Welfare recommends against screening for PCa but supports regional implementation of organised prostate cancer testing (OPT). OBJECTIVE: To report the process for designing and implementing OPT programmes. DESIGN, SETTING, AND PARTICIPANTS: Population-based OPT programmes in two Swedish regions, designed to include men aged between 50 and 74 yr, launched in September 2020 for 50-yr-old men. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: The number of men invited, the participation rate, and the numbers of magnetic resonance imaging (MRI) scans, urological visits, and biopsies from September 2020 to June 2021 were recorded. RESULTS AND LIMITATIONS: Two Swedish regions co-designed an OPT programme with a risk-stratified diagnostic algorithm based on prostate-specific antigen (PSA), PSA density, MRI findings, and age. An automated administrative system was developed on a nationwide web-based platform. Invitation letters and test results are automatically generated and sent out by post. Men with PSA ≥3ng/ml, a suspicious MRI lesion, and/or PSA density ≥0.15 ng/ml/cm3 are referred for a prostate biopsy. Test results are registered for quality control and research. By June 2021, a total of 16 515 men were invited, of whom 6309 (38%) participated; 147 had an MRI scan and 39 underwent prostate biopsy. The OPT framework, algorithm, and diagnostic pathways have been working well. CONCLUSIONS: We designed and implemented a framework for OPT with a high grade of automation. The framework and organisational experiences may be of value for others who plan a programme for early detection of PCa. PATIENT SUMMARY: We describe the implementation of an organised testing programme for early detection of prostate cancer in two Swedish regions. This model is the first of its kind and may serve as a template for similar programmes.

Developmental learning in a pain-related system: evidence for a cross-modality mechanism.

The nociceptive spinal reflex system performs highly precise sensorimotor transformations that require functionally specified synaptic strengths. The specification is gradually attained during early development and appears to be learning dependent. Here we determine the time course of this specification for heat-nociceptive tail withdrawal reflexes and analyze which types of primary afferents are important for the learning by applying various forms of noninvasive sensory deprivations. The percentage of erroneous heat-nociceptive tail withdrawal reflexes (i.e., movements directed toward the stimulation) decreased gradually from 64.1 +/- 2.5% (mean +/- SEM) to <10% during postnatal days 10-21. This improvement was completely blocked by anesthetizing the tail during the adaptation period, confirming that an experience-dependent mechanism is involved in the specification of synaptic strengths. However, the results show that the adaptation occurs to a significant extent despite local analgesia and protection of the tail from noxious input, provided that tactile sensitivity is preserved. Therefore, it appears that a nociceptive input is not necessary for the adaptation, and that input from tactile receptors can be used to guide the nociceptive synaptic organization during development. Sensory deprivation in the adult rat failed to affect the heat-nociceptive withdrawal reflex system, indicating that the adaptation has a "critical period" during early development. These findings provide a key to the puzzle of how pain-related systems can be functionally adapted through experience despite the rare occurrence of noxious input during early life.