Texture analysis of equivocal Likert scored 3/5 peripheral zone prostate lesions on mpMRI
Aishah Azam1, Dario Picone2, Mrishta Brizmohun Appayya2, Balaji Ganeshan3, Nikolaos Dikaios2, Raymond Endozo3, Ashley Groves3, Hashim Ahmed4, and Shonit Punwani1

1University College London Hospital, London, United Kingdom, 2Centre for Medical Imaging, University College London, London, United Kingdom, 3Institute of Nuclear Medicine, University College London, London, United Kingdom, 4Division of Surgery, University College London Hospital, London, United Kingdom

Synopsis

Approximately 30% of multiparametricMRIs for peripheral zone prostate cancer are scored as “equivocal”. Texture analysis was performed on 66 patients with “equivocal” mpMRIs using TexRad software to see if there were any particular textural differences between patients with significant cancer and non-significant disease on biopsy. We found average entropy on ADC sequences of the PZ is reduced in patients with significant cancer (p=0.003). Using an entropy threshold of <5.99 demonstrates a sensitivity of 0.88 and specificity is 0.60 for detecting significant PZ prostate cancer. Therefore, ADC entropy can help assess “equivocal” studies and enable selection of patients for biopsy and treatment.

Target audience:

Radiologists, oncologists, urologists, physicists with an interest in prostate cancer diagnostics.

Purpose:

Approximately 70% of prostate cancers arise in the peripheral zone (PZ) (1) and these cancers tend to be more aggressive than those found in the transition zone (2). The use of multiparametric Magnetic Resonance Imaging (mpMRI) has transformed the traditional prostate cancer pathway allowing the accurate localisation of prostate cancer (3). However, in approximately 30% of patients, the mpMRI study is equivocal for tumour (PiRADS/Likert score 3/5). It is unclear what to do when a tumour is reported as “equivocal”, whether to perform a biopsy or to repeat the mpMRI study after an interval; this has a direct implication for individual patient and the health economics of prostate mpMRI as a diagnostic technique.

Image texture analysis encompasses a mathematical approach to evaluate pixel intensity and position. This is then used to derive textural features including entropy, skewness and kurtosis which each have biological relevance to heterogeneity (4). Textural analysis has been applied in many different imaging modalities and has been shown to be able to identify tumours and even assess prognosis in brain, rectal, breast and liver ca (5-9). Our aim was to determine whether textural analysis of peripheral zone mp-MRI signal could help improve classification of ‘equivocal’ reported studies.

Materials and methods:

330 men enrolled in the PICTURE study (10) underwent mpMRI of the prostate which were all prospectively scored on a Likert scale by a radiologist with 8 years of experience blinded to histopathological results. Patients underwent mpMRI using a 3T static magnet with a specific combination of sequences including: T2 weighted, apparent diffusion co-efficient (ADC) and dynamic contrast enhancement (DCE) imaging. From the cohort, 66 patients had prostate mpMRIs scored as “equivocal” (Likert 3/5) and were eligible for inclusion for textural analysis. All patients subsequently had a targeted template biopsy of the primary lesion scored 3/5; and irrespective of mpMRI score completed a full 20-core template biopsy of the remainder of their prostate (10). In each patient, 3 representative slices (at apex, mid-gland and base for each of DCE [early post-contrast T1], ADC and T2 weighted images) were selected for textural analysis.

The peripheral zone was manually segmented and then textural analysis performed using TexRAD software (TexRAD Ltd www.texrad.com, Feedback Plc) (5). TexRad quantified pixel distribution within the segmented PZ of the prostate gland and this was followed by histogram analysis to generate values for entropy, skewness and kurtosis. T2, ADC and DCE textural parameters were separately recorded and averaged (across the 3 slices) for each patient. Textural features were separately assessed for predicting patients with significant prostate cancer using receiver operating characteristic (ROC) areas under curve (AUC) analysis. Significant cancer was defined as the presence of a Gleason 4 disease component at biopsy.

Results:

16/66 Likert 3/5 reported patients demonstrated significant cancer at biopsy. 15 patients had gleason 3+4, and 1 patient had Gleason 4+3. The remaining 50 patients with: Gleason score 3+3 cancers (28 patients) and benign histology (22 patients), were combined and classified together as non-significant disease. Mean PZ ADC entropy was significantly lower in patients with significant cancer at biopsy (5.698 vs. 5.883; p=0.003). There was no significant difference of mean peripheral zone kurtosis or skewness between patients with and without significant cancer (p=0.944, p=0.221). There was no significant difference of any T2 or DCE textural parameters between patients with and without significant cancer (p=0.051 to 0.742). ROC-AUC of mean peripheral zone ADC entropy was 0.749 (95% CI: 0.613-0.884). A mean peripheral zone ADC entropy threshold of <5.99 demonstrates a sensitivity of 0.88 and specificity is 0.60 for detecting significant PZ prostate cancer.

Discussion:

Entropy is a first-order statistic measure of texture irregularity and a lower entropy value in patients with significant cancer suggests a more a homogenous peripheral zone ADC map. Our results demonstrate that assessing the entropy of ADC of the peripheral zone in patients that have ‘equivocal’ mpMRI studies could help to further classify patients by selecting those patients most likely to have significant cancer for early biopsy and subsequent treatment.

Acknowledgements

This work has been supported by the KCL-UCL Comprehensive Cancer Imaging Centre funding [Cancer Research UK (CR-UK) & Engineering and Physical Sciences Research Council (EPSRC)].The majority of this work was undertaken at University College London Hospital and University College London, which receive a proportion of funding from the NIHR Biomedical Research Centre funding scheme [Department of Health UK].

References

1. McNeal JE, Redwine EA, Freiha FS, Stamey TA. Zonal distribution of prostatic adenocarcinoma. Correlation with histologic pattern and direction of spread. Am J Surg Pathol. 1988;(12):897-906.

2. Erdersdobler A, Fritz H, Schnoger S, Graefen M, Hammerer P, Huland H, et al. Tumour grade, proliferation, apoptosis, microvessel density, p53, and bcl-2 in prostate cancers: differences between tumours located in the transition zone and in the peripheral zone. European Urology. 2002;(1):40-6.

3. Fütterer JJ, Briganti A, De Visschere P, Emberton M, Giannarini G, Kirkham A, et al. Can Clinically Significant Prostate Cancer Be Detected with Multiparametric Magnetic Resonance Imaging? A Systematic Review of the Literature. European Urology. 2015:1-9.

4. Davnall F, Connie SPY, Ljungqvist G, Selmi M, Ng F, Sanghera B, et al. Assessment of tumor heterogeneity: an emerging tool for clinical practice? Insights imaging. 2012;(3):573-589.

5. Miles KA, Ganeshan B, Hayball MP. CT tecture analysis using the filtration-histogram method: what do the measurements mean? Cancer Imaging. 2013;13(3):400-406.

6. Eliat PA, Olivie D, Saikali S, Carsin B, Saint-Jalmes H, de Certaines JD. Can dynamic contrast-enhanced magnetic resonance imaging combined with texture analysis differentiate malignant glioneuronal tumors from other glioblastoma? Neurol Res Int. 2012.

7. Mayerhoefer ME, Schima W, Trattnig S, Pinker K, Berger-Kulemann V, Ba-Ssalamah A. Texture-based classification of focal liver lesions on MRI at 3.0 Tesla: a feasibility study in cysts and hemangiomas. J Magn Reson Imaging 2010; 32(2):352–359.

8. Lopes R, Ayache A, Makni N et al. Prostate cancer characterization on MR images using fractal features. Med Phys. 2011; 38(1):83–95.

9. Parikh J, Selmi M, Charles-Edwards G, Glendenning J, Ganeshan B, Verma H, Mansi J, Harries M, Tutt A, Goh V. Changes in primary breast cancer heterogeneity may augment midtreatment MR imaging assessment of response to neoadjuvant chemotherapy. Radiology 2014; 272(1):100-12.

10. Simmons LA, Ahmed HU, Moore CM, et al. The PICTURE study -- prostate imaging (multi-parametric MRI and Prostate HistoScanning™) compared to transperineal ultrasound guided biopsy for significant prostate cancer risk evaluation. Contemp Clin Trials. Jan 2014;37(1):69-83.

Figures

Average ADC entropy for patients with significant ca vs. non-significant disease ± stdev (5.698±0.266, 5.883±0.222, p = 0.003)

Receiver Operating Characteristic curve of average ADC Entropy for determination of significant peripheral zone prostate ca, Area Under Curve = 0.749.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2496