Natural history of prostatic lesions on serial multiparametric magnetic resonance imaging

Biparametric MRI in prostate cancer during active surveillance: is it safe?

Active surveillance (AS) is the preferred option for patients presenting with low-intermediate-risk prostate cancer. MRI now plays a crucial role for baseline assessment and ongoing monitoring of AS. The Prostate Cancer Radiological Estimation of Change in Sequential Evaluation (PRECISE) recommendations aid radiological assessment of progression; however, current guidelines do not advise on MRI protocols nor on frequency. Biparametric (bp) imaging without contrast administration offers advantages such as reduced costs and increased throughput, with similar outcomes to multiparametric (mp) MRI shown in the biopsy naïve setting. In AS follow-up, the paradigm shifts from MRI lesion detection to assessment of progression, and patients have the further safety net of continuing clinical surveillance. As such, bpMRI may be appropriate in clinically stable patients on routine AS follow-up pathways; however, there is currently limited published evidence for this approach. It should be noted that mpMRI may be mandated in certain patients and potentially offers additional advantages, including improving image quality, new lesion detection, and staging accuracy. Recently developed AI solutions have enabled higher quality and faster scanning protocols, which may help mitigate against disadvantages of bpMRI. In this article, we explore the current role of MRI in AS and address the need for contrast-enhanced sequences. CLINICAL RELEVANCE STATEMENT: Active surveillance is the preferred plan for patients with lower-risk prostate cancer, and MRI plays a crucial role in patient selection and monitoring; however, current guidelines do not currently recommend how or when to perform MRI in follow-up. KEY POINTS: Noncontrast biparametric MRI has reduced costs and increased throughput and may be appropriate for monitoring stable patients. Multiparametric MRI may be mandated in certain patients, and contrast potentially offers additional advantages. AI solutions enable higher quality, faster scanning protocols, and could mitigate the disadvantages of biparametric imaging.

Stability of Multi-Parametric Prostate MRI Radiomic Features to Variations in Segmentation

Stability analysis remains a fundamental step in developing a successful imaging biomarker to personalize oncological strategies. This study proposes an in silico contour generation method for simulating segmentation variations to identify stable radiomic features. Ground-truth annotation provided for the whole prostate gland on the multi-parametric MRI sequences (T2w, ADC, and SUB-DCE) were perturbed to mimic segmentation differences observed among human annotators. In total, we generated 15 synthetic contours for a given image-segmentation pair. One thousand two hundred twenty-four unfiltered/filtered radiomic features were extracted applying Pyradiomics, followed by stability assessment using ICC(1,1). Stable features identified in the internal population were then compared with an external population to discover and report robust features. Finally, we also investigated the impact of a wide range of filtering strategies on the stability of features. The percentage of unfiltered (filtered) features that remained robust subjected to segmentation variations were T2w-36% (81%), ADC-36% (94%), and SUB-43% (93%). Our findings suggest that segmentation variations can significantly impact radiomic feature stability but can be mitigated by including pre-filtering strategies as part of the feature extraction pipeline.

Bridging the gap between prostate radiology and pathology through machine learning

Background

Prostate cancer remains the second deadliest cancer for American men despite clinical advancements. Currently, magnetic resonance imaging (MRI) is considered the most sensitive non‐invasive imaging modality that enables visualization, detection, and localization of prostate cancer, and is increasingly used to guide targeted biopsies for prostate cancer diagnosis. However, its utility remains limited due to high rates of false positives and false negatives as well as low inter‐reader agreements.

Purpose

Machine learning methods to detect and localize cancer on prostate MRI can help standardize radiologist interpretations. However, existing machine learning methods vary not only in model architecture, but also in the ground truth labeling strategies used for model training. We compare different labeling strategies and the effects they have on the performance of different machine learning models for prostate cancer detection on MRI.

Methods

Four different deep learning models (SPCNet, U‐Net, branched U‐Net, and DeepLabv3+) were trained to detect prostate cancer on MRI using 75 patients with radical prostatectomy, and evaluated using 40 patients with radical prostatectomy and 275 patients with targeted biopsy. Each deep learning model was trained with four different label types: pathology‐confirmed radiologist labels, pathologist labels on whole‐mount histopathology images, and lesion‐level and pixel‐level digital pathologist labels (previously validated deep learning algorithm on histopathology images to predict pixel‐level Gleason patterns) on whole‐mount histopathology images. The pathologist and digital pathologist labels (collectively referred to as pathology labels) were mapped onto pre‐operative MRI using an automated MRI‐histopathology registration platform.

Results

Radiologist labels missed cancers (ROC‐AUC: 0.75‐0.84), had lower lesion volumes (~68% of pathology lesions), and lower Dice overlaps (0.24‐0.28) when compared with pathology labels. Consequently, machine learning models trained with radiologist labels also showed inferior performance compared to models trained with pathology labels. Digital pathologist labels showed high concordance with pathologist labels of cancer (lesion ROC‐AUC: 0.97‐1, lesion Dice: 0.75‐0.93). Machine learning models trained with digital pathologist labels had the highest lesion detection rates in the radical prostatectomy cohort (aggressive lesion ROC‐AUC: 0.91‐0.94), and had generalizable and comparable performance to pathologist label‐trained‐models in the targeted biopsy cohort (aggressive lesion ROC‐AUC: 0.87‐0.88), irrespective of the deep learning architecture. Moreover, machine learning models trained with pixel‐level digital pathologist labels were able to selectively identify aggressive and indolent cancer components in mixed lesions on MRI, which is not possible with any human‐annotated label type.

Conclusions

Machine learning models for prostate MRI interpretation that are trained with digital pathologist labels showed higher or comparable performance with pathologist label‐trained models in both radical prostatectomy and targeted biopsy cohort. Digital pathologist labels can reduce challenges associated with human annotations, including labor, time, inter‐ and intra‐reader variability, and can help bridge the gap between prostate radiology and pathology by enabling the training of reliable machine learning models to detect and localize prostate cancer on MRI.

Alternatives for MRI in Prostate Cancer Diagnostics-Review of Current Ultrasound-Based Techniques

Simple Summary

Prostate cancer (PCa) is the most common solid malignant tumor in men worldwide with various clinical manifestations. Due to overdiagnosis and overtreatment of a clinically insignificant disease, multiparametric magnetic resonance imaging is recommended for every patient before performing prostate biopsy. However, the diagnostic pathway currently has many limitations and is still far from ideal. Therefore, further alternatives need to be investigated. As the novel ultrasound-based techniques, such as shear wave elastography, contrast-enhanced ultrasound or high frequency micro-ultrasound are able to, overcome the limitations of magnetic resonance imaging presenting good performance in recent studies, we have summarized and compared the results of each technique in the detection of PCa. Furthermore, we analyzed the future perspectives for ultrasound modalities that may soon significantly improve their diagnostic value.

Abstract

The purpose of this review is to present the current role of ultrasound-based techniques in the diagnostic pathway of prostate cancer (PCa). With overdiagnosis and overtreatment of a clinically insignificant PCa over the past years, multiparametric magnetic resonance imaging (mpMRI) started to be recommended for every patient suspected of PCa before performing a biopsy. It enabled targeted sampling of the suspicious prostate regions, improving the accuracy of the traditional systematic biopsy. However, mpMRI is associated with high costs, relatively low availability, long and separate procedure, or exposure to the contrast agent. The novel ultrasound modalities, such as shear wave elastography (SWE), contrast-enhanced ultrasound (CEUS), or high frequency micro-ultrasound (MicroUS), may be capable of maintaining the performance of mpMRI without its limitations. Moreover, the real-time lesion visualization during biopsy would significantly simplify the diagnostic process. Another value of these new techniques is the ability to enhance the performance of mpMRI by creating the image fusion of multiple modalities. Such models might be further analyzed by artificial intelligence to mark the regions of interest for investigators and help to decide about the biopsy indications. The dynamic development and promising results of new ultrasound-based techniques should encourage researchers to thoroughly study their utilization in prostate imaging.

Repeat MRI during active surveillance: natural history of prostatic lesions and upgrading rates

OBJECTIVES

To assess upgrading rates in patients on active surveillance (AS) for prostate cancer (PCa) after serial multiparametric magnetic resonance imaging (mpMRI).

METHODS

We conducted a retrospective analysis of 558 patients. Five different criteria for mpMRI progression were used: 1) a Prostate Imaging Reporting and Data System (PI-RADS) score increase; 2) a lesion size increase; 3) an extraprostatic extension score increase; 4) overall mpMRI progression; and 5) the number of criteria met for mpMRI progression (0 vs 1 vs 2-3). In addition, two definitions of PCa upgrading were evaluated: 1) International Society of Urological Pathology Grade Group (ISUP GG) ≥2 with >10% of pattern 4 and 2) ISUP GG ≥ 3. Estimated annual percent changes methodology was used to show the temporal trends of mpMRI progression criteria. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of mpMRI progression criteria were also analysed. Multivariable logistic regression models tested PCa upgrading rates.

RESULTS

Lower rates over time for all mpMRI progression criteria were observed. The NPV of serial mpMRI scans ranged from 90.5% to 93.5% (ISUP GG≥2 with >10% of pattern 4 PCa upgrading) and from 98% to 99% (ISUP GG≥3 PCa upgrading), depending on the criteria used for mpMRI progression. A prostate-specific antigen density (PSAD) threshold of 0.15 ng/mL/mL was used to substratify those patients who would be able to skip a prostate biopsy. In multivariable logistic regression models assessing PCa upgrading rates, all five mpMRI progression criteria achieved independent predictor status.

CONCLUSION

During AS, approximately 27% of patients experience mpMRI progression at first repeat MRI. However, the rates of mpMRI progression decrease over time at subsequent mpMRI scans. Patients with stable mpMRI findings and with PSAD < 0.15 ng/mL/mL could safely skip surveillance biopsies. Conversely, patients who experience mpMRI progression should undergo a prostate biopsy.

Prostate Cancer Screening with Magnetic Resonance Imaging: Results from the Second Round of the Göteborg Prostate Cancer Screening 2 Trial

BACKGROUND

The Göteborg 2 prostate cancer (PC) screening (G2) trial evaluates screening with prostate-specific antigen (PSA) followed by magnetic resonance imaging (MRI) in case of elevated PSA levels.

OBJECTIVE

To assess the safety of using a 2-yr interval in men who were previously screened positive with PSA but had negative MRI or positive MRI with a negative biopsy.

DESIGN, SETTING, AND PARTICIPANTS

A total of 61 201 men aged 50-60 yr were randomized and 38 366 were invited for screening (years 2015-2020). Men with positive MRI (Prostate Imaging Reporting and Data System [PI-RADS] score ≥3) were scheduled for targeted biopsies. Men with negative MRI or negative biopsies were reinvited after 2 yr. Round 1 and 2 MRI scans (PI-RADS ≥3) of men not diagnosed with PC in round 1 were re-read and classified according to Prostate Cancer Radiological Estimation of Change in Sequential Evaluation (PRECISE) by two radiologists. Interval PCs (detected outside the program before invitation to round 2) were identified by linking to the Regional PC Registry.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

Tabulation of overall detection of PC was done.

RESULTS AND LIMITATIONS

Between October 2017 and June 2020, 474 men with round 1 elevated PSA and MRI underwent a second screening. Of those, 19% had nonelevated PSA in round 2 and were not examined further. Of the remaining 376 men, 89% had negative MRI. Targeted biopsies yielded 14 PCs: nine grade group (GG) 1 and five GG 2-3. In men with PI-RADS ≥3 and PC diagnosed in round 2, only two (GG 1) progressed according to the PRECISE criteria and the remainder were stable. Ten interval PCs were diagnosed: seven GG 1, one GG 2, and two GG 5. The two GG 5 PCs were PI-RADS 4 and 5 with negative round 1 biopsy.

CONCLUSIONS

A 2-yr interval seems to be safe in men with negative MRI, while men with PI-RADS 4 and 5 lesions with negative biopsies should have a closer follow-up.

PATIENT SUMMARY

In prostate cancer screening, a 2-yr follow-up seems to be safe if magnetic resonance imaging did not show highly suspicious findings.

Update on Multiparametric Prostate MRI During Active Surveillance: Current and Future Trends and Role of the PRECISE Recommendations

Active surveillance for low-to-intermediate risk prostate cancer is a conservative management approach that aims to avoid or delay active treatment until there is evidence of disease progression. In recent years, multiparametric MRI (mpMRI) has been increasingly used in active surveillance and has shown great promise in patient selection and monitoring. This has been corroborated by publication of the Prostate Cancer Radiologic Estimation of Change in Sequential Evaluation (PRECISE) recommendations, which define the ideal reporting standards for mpMRI during active surveillance. The PRECISE recommendations include a system that assigns a score from 1 to 5 (the PRECISE score) for the assessment of radiologic change on serial mpMRI scans. PRECISE scores are defined as follows: a score of 3 indicates radiologic stability, a score of 1 or 2 denotes radiologic regression, and a score of 4 or 5 indicates radiologic progression. In the present study, we discuss current and future trends in the use of mpMRI during active surveillance and illustrate the natural history of prostate cancer on serial scans according to the PRECISE recommendations. We highlight how the ability to classify radiologic change on mpMRI with use of the PRECISE recommendations helps clinical decision making.

Multiparametric magnetic resonance imaging facilitates reclassification during active surveillance for prostate cancer

OBJECTIVE

To investigate the utility of multiparametric magnetic resonance imaging (mpMRI) in the reassessment and monitoring of patients on active surveillance (AS) for Grade Group (GG) 1 prostate cancer (PCa).

PATIENTS AND METHODS

We identified, from our prospectively maintained institutional review board-approved database, 181 consecutive men enrolled on AS for GG 1 PCa who underwent at least one surveillance mpMRI followed by MRI/prostate biopsy (PBx). A subset analysis was performed among 68 patients who underwent serial (at least two) mpMRI/PBx during AS. Pathological progression (PP) was defined as upgrade to GG ≥2 on follow up biopsy.

RESULTS

Baseline MRI was performed in 34 patients (19%). At a median follow-up of 2.2 years for the overall cohort, the PP was 12% (6/49) for Prostate Imaging Reporting and Data System (PI-RADS) 1-2 lesions and 37% (48/129) for the PI-RADS ≥3 lesions. The 2-year PP-free survival rate was 84%. Surveillance prostate-specific antigen density (P < 0.001) and surveillance PI-RADS ≥3 (P = 0.002) were independent predictors of PP on reassessment MRI/PBx. In the serial MRI cohort, the 2-year PP-free survival was 95% for the No-MRI-progression group vs 85% for the MRI-progression group (P = 0.02). MRI progression was significantly higher in the PP (62%) than in the No-PP (31%) group (P = 0.04). If serial MRI were used for PCa surveillance and biopsy were triggered based only on MRI progression, 63% of PBx might be postponed at the cost of missing 12% of GG ≥2 PCa in those with stable MRI. Conversely, this strategy would miss 38% of those with upgrading to GG ≥2 PCa on biopsy. Stable serial mpMRI correlates with no reclassification to GG ≥3 PCa during AS.

CONCLUSION

On surveillance mpMRI, PI-RADS ≥3 was associated with increased risk of PCa reclassification. Surveillance biopsy based only on MRI progression may avoid a large number of biopsies at the cost of missing many PCa reclassifications.

Five-year Outcomes of Magnetic Resonance Imaging-based Active Surveillance for Prostate Cancer: A Large Cohort Study

Background

Although the use of multiparametric magnetic resonance imaging (mpMRI) in active surveillance (AS) for prostate cancer is of increasing interest, existing data are derived from small cohorts.

Objective

We describe clinical, histological, and radiological outcomes from an established AS programme, where protocol-based biopsies were omitted in favour of MRI-led monitoring.

Design, setting, and participants

Data on 672 men enrolled in AS between August 2004 and November 2017 (inclusion criteria: Gleason 3 + 3 or 3 + 4 localised prostate cancer, presenting prostate-specific antigen <20 ng/ml, and baseline mpMRI) were collected from the University College London Hospital (UCLH) database.

Outcome measurements and statistical analysis

Primary outcomes were event-free survival (EFS; event defined as prostate cancer treatment, transition to watchful waiting, or death) and treatment-free survival (TFS). Secondary outcomes included rates of all-cause or prostate cancer–related mortality, metastasis, and upgrading to Gleason ≥4 + 3. Data on radiological and histological progression were also collected.

Results and limitations

More than 3800 person-years (py) of follow-up were accrued (median: 58 mo; interquartile range 37–82 mo). Approximately 84.7% (95% confidence interval [CI]: 82.0–87.6) and 71.8% (95% CI: 68.2–75.6) of patients remained on AS at 3 and 5 yr, respectively. EFS and TFS were lower in those with MRI-visible (Likert 4–5) disease or secondary Gleason pattern 4 at baseline (log-rank test; p <  0.001). In total, 216 men were treated. There were 24 deaths, none of which was prostate cancer related (6.3/1000 py; 95% CI: 4.1–9.5). Metastases developed in eight men (2.1 events/1000 py; 95% CI: 1.0–4.3), whereas 27 men upgraded to Gleason ≥4 + 3 on follow-up biopsy (7.7 events/1000 py; 95% CI: 5.2–11.3).

Conclusions

The rates of discontinuation, mortality, and metastasis in MRI-led surveillance are comparable with those of standard AS. MRI-visible disease and/or secondary Gleason grade 4 at baseline are associated with a greater likelihood of moving to active treatment at 5 yr. Further research will concentrate on optimising imaging intervals according to baseline risk.

Patient summary

In this report, we looked at the outcomes of magnetic resonance imaging (MRI)-based surveillance for prostate cancer in a UK cohort. We found that this strategy could allow routine biopsies to be avoided. Secondary Gleason pattern 4 and MRI visibility are associated with increased rates of treatment. We conclude that MRI-based surveillance should be considered for the monitoring of small prostate tumours.

Multiparametric MRI for prostate cancer diagnosis: current status and future directions

The current diagnostic pathway for prostate cancer has resulted in overdiagnosis and consequent overtreatment as well as underdiagnosis and missed diagnoses in many men. Multiparametric MRI (mpMRI) of the prostate has been identified as a test that could mitigate these diagnostic errors. The performance of mpMRI can vary depending on the population being studied, the execution of the MRI itself, the experience of the radiologist, whether additional biomarkers are considered and whether mpMRI-targeted biopsy is carried out alone or in addition to systematic biopsy. A number of challenges to implementation remain, such as ensuring high-quality execution and reporting of mpMRI and ensuring that this diagnostic pathway is cost-effective. Nevertheless, emerging clinical trial data support the adoption of this technology as part of the standard of care for the diagnosis of prostate cancer.

Using prognosis to guide inclusion criteria, define standardised endpoints and stratify follow-up in active surveillance for prostate cancer

OBJECTIVES

To test whether using disease prognosis can inform a rational approach to active surveillance (AS) for early prostate cancer.

PATIENTS AND METHODS

We previously developed the Cambridge Prognostics Groups (CPG) classification, a five-tiered model that uses prostate-specific antigen (PSA), Grade Group and Stage to predict cancer survival outcomes. We applied the CPG model to a UK and a Swedish prostate cancer cohort to test differences in prostate cancer mortality (PCM) in men managed conservatively or by upfront treatment in CPG2 and 3 (which subdivides the intermediate-risk classification) vs CPG1 (low-risk). We then applied the CPG model to a contemporary UK AS cohort, which was optimally characterised at baseline for disease burden, to identify predictors of true prognostic progression. Results were re-tested in an external AS cohort from Spain.

RESULTS

In a UK cohort (n = 3659) the 10-year PCM was 2.3% in CPG1, 1.5%/3.5% in treated/untreated CPG2, and 1.9%/8.6% in treated/untreated CPG3. In the Swedish cohort (n = 27 942) the10-year PCM was 1.0% in CPG1, 2.2%/2.7% in treated/untreated CPG2, and 6.1%/12.5% in treated/untreated CPG3. We then tested using progression to CPG3 as a hard endpoint in a modern AS cohort (n = 133). During follow-up (median 3.5 years) only 6% (eight of 133) progressed to CPG3. Predictors of progression were a PSA density ≥0.15 ng/mL/mL and CPG2 at diagnosis. Progression occurred in 1%, 8% and 21% of men with neither factor, only one, or both, respectively. In an independent Spanish AS cohort (n = 143) the corresponding rates were 3%, 10% and 14%, respectively.

CONCLUSION

Using disease prognosis allows a rational approach to inclusion criteria, discontinuation triggers and risk-stratified management in AS.

Outcomes of Serial Multiparametric Magnetic Resonance Imaging and Subsequent Biopsy in Men with Low-risk Prostate Cancer Managed with Active Surveillance

BACKGROUND

Outcomes of serial multiparametric magnetic resonance imaging (mpMRI) and subsequent biopsy in monitoring prostate cancer (PCa) in men on active surveillance (AS) have not been defined clearly.

OBJECTIVE

To determine whether changes in serial mpMRI can predict pathological upgrade among men with grade group (GG) 1 PCa managed with AS.

DESIGN, SETTING, AND PARTICIPANTS

Retrospective analysis of men with GG1 on AS with at least two consecutive mpMRI examinations during 2012-2018 who underwent mpMRI/ultrasound fusion or systematic biopsies.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

Progression on serial mpMRI was evaluated as a predictor of pathological upgrading to GG≥2 on a follow-up biopsy using clinical, pathological, and imaging factors in binary logistic regression. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were determined.

RESULTS AND LIMITATIONS

Of 122 patients, 29 men (23.8%) experienced pathological upgrade on the follow-up biopsy. Progression on mpMRI was not associated with pathological upgrade. The sensitivity, specificity, PPV, and NPV of mpMRI progression for predicting pathological upgrade were 41.3%, 54.8%, 22.2%, and 75%, respectively. Age (odds ratio [OR] 1.17, p=0.006), Prostate Imaging Reporting and Data System (PI-RADS) score on initial mpMRI (4-5 vs ≤3, OR 7.48, p=0.01), number of positive systematic cores (OR 1.84, p=0.03), number of positive targeted cores (OR 0.44, p=0.04), and maximum percent of targeted core tumor involvement (OR 1.04, p=0.01) were significantly associated with pathological upgrade.

CONCLUSIONS

We did not observe an association between mpMRI progression and pathological upgrade; however, a PI-RADS score of 4-5 on initial mpMRI was predictive of subsequent pathological progression. The continued use of systematic and fusion biopsies appears necessary due to risks of reclassification over time.

PATIENT SUMMARY

Progression on serial multiparametric magnetic resonance imaging during active surveillance (AS) is not associated with progression on the follow-up biopsy. Both systematic and fusion biopsies are necessary to sufficiently capture progression during AS.

Four-year outcomes from a multiparametric magnetic resonance imaging (MRI)-based active surveillance programme: PSA dynamics and serial MRI scans allow omission of protocol biopsies

OBJECTIVES

To report outcomes from a multiparametric (mp) magnetic resonance imaging (MRI)-based active surveillance programme that did not include performing protocol biopsies after the first confirmatory biopsy.

PATIENTS AND METHODS

All patients diagnosed with Gleason 3 + 3 prostate cancer because of a raised PSA level who underwent mpMRI after diagnosis were included. Patients were recorded in a prospective clinical database and followed up with PSA monitoring and repeat MRI. In patients who remained on active surveillance after the first MRI (with or without confirmatory biopsy), we investigated PSA dynamics for association with subsequent progression. Comparison between first and second MRI scans was undertaken. Outcomes assessed were: progression to radical therapy at first MRI/confirmatory biopsy and progression to radical therapy in those who remained on active surveillance after first MRI.

RESULTS

A total of 211 patients were included, with a median of 4.2 years of follow-up. The rate of progression to radical therapy was significantly greater at all stages among patients with visible lesions than in those with initially negative MRI (47/125 (37.6%) vs 11/86 (12.8%); odds ratio 4.1 (95% CI 2.0-8.5), P < 0.001). Only 1/56 patients (1.8%) with negative initial MRI scans who underwent a confirmatory systematic biopsy had upgrading to Gleason 3 + 4 disease. PSA velocity was significantly associated with subsequent progression in patients with negative initial MRI (area under the curve 0.85 [95% CI 0.75-0.94]; P <0.001). Patients with high-risk visible lesions on first MRI who remained on active surveillance had a high risk of subsequent progression 19/76 (25.0%) vs 9/84 (10.7%) for patients with no visible lesions, despite reassuring targeted and systematic confirmatory biopsies and regardless of PSA dynamics.

CONCLUSION

Men with low-risk Gleason 3 + 3 prostate cancer on active surveillance can forgo protocol biopsies in favour of MRI and PSA monitoring with selective re-biopsy.