Mohamed Mahmoud, Greif Philipp A, Diamond James, Sharafeldin Osama, Maxwell Perry, Montironi Rodolfo, O'Brien Aidan, Young Michael, Hamilton Peter W
Biomedical Imaging and Informatics Research Group, The Queen's University of Belfast, Belfast, Ireland.
BJU Int. 2009 Feb;103(3):391-8. doi: 10.1111/j.1464-410X.2008.08063.x. Epub 2008 Nov 14.
To assess the value of studying chromatin organization using high-resolution digital image analysis to predict the response to hormonal-deprivation therapy (HDT) in patients with prostate cancer, using pretreatment prostate tissues.
A tissue microarray (TMA) was constructed using pretreatment paraffin-embedded tissues from transurethral resection of the prostate (TURP) samples (48 patients, 96 cores). None of the patients had received any treatment for prostate cancer before TURP. The patients' medical records for 5 years after treatment were assessed; patients were divided, based on their prostatic specific antigen (PSA) levels after treatment, into those optimally responsive to HDT (14) and those resistant to HDT (34). The latter were further subclassified based on their nadir PSA level. Imaging comprised a calibrated digital image-analysis system with software for densitometric and texture analysis, the latter being assessed on manually segmented nuclei (> or =30 nuclei/core).
Most of the measured digital texture features assessing chromatin density and distribution were significantly different between the prognostic groups (P = 0.001). In the training set, 12 of 14 HDT-responsive and 23 (68%) of HDT-resistant patients were accurately predicted. However, all HDT-resistant patients with a nadir PSA level of >5 ng/mL were accurately predicted. The overall classification sensitivity was 47%, specificity 94% with a positive predictive value of 85%. However, the sensitivity was 100% between patients optimally responsive to HDT and those poorly responsive with a nadir PSA level of >5 ng/mL.
Quantitative image analysis of chromatin phenotype showed promising value in predicting before treatment the response to HDT in patients diagnosed with prostatic adenocarcinoma. However, further work using larger data sets is required before adapting the technique in routine clinical practice.
利用高分辨率数字图像分析研究染色质组织,以预测前列腺癌患者对激素剥夺疗法(HDT)的反应,采用治疗前的前列腺组织进行评估。
使用经尿道前列腺切除术(TURP)样本(48例患者,96个芯块)的治疗前石蜡包埋组织构建组织微阵列(TMA)。所有患者在TURP前均未接受过前列腺癌的任何治疗。评估患者治疗后5年的病历;根据治疗后的前列腺特异性抗原(PSA)水平,将患者分为对HDT反应最佳的患者(14例)和对HDT耐药的患者(34例)。后者根据其最低PSA水平进一步细分。成像包括一个校准的数字图像分析系统,带有用于密度测定和纹理分析的软件,后者在手动分割的细胞核(每个芯块≥30个细胞核)上进行评估。
评估染色质密度和分布的大多数测量数字纹理特征在预后组之间存在显著差异(P = 0.001)。在训练集中,14例HDT反应者中的12例和34例HDT耐药患者中的23例(68%)被准确预测。然而,所有最低PSA水平>5 ng/mL的HDT耐药患者均被准确预测。总体分类敏感性为47%,特异性为94%,阳性预测值为85%。然而,在对HDT反应最佳的患者和最低PSA水平>5 ng/mL的反应较差的患者之间,敏感性为100%。
染色质表型的定量图像分析在预测前列腺腺癌患者治疗前对HDT的反应方面显示出有前景的价值。然而,在将该技术应用于常规临床实践之前,需要使用更大的数据集进行进一步的研究。