Comparing radiomic classifiers and classifier ensembles for detection of peripheral zone prostate tumors on T2-weighted MRI: a multi-site study.
*Prostate cancer; *Classifiers; *Comparison; *MRI; *Radiomics
BACKGROUND: For most computer-aided diagnosis (CAD) problems involving prostate cancer detection via medical imaging data, the choice of classifier has been largely ad hoc, or been motivated by classifier comparison studies that have involved large synthetic datasets. More significantly, it is currently unknown how classifier choices and trends generalize across multiple institutions, due to heterogeneous acquisition and intensity characteristics (especially when considering MR imaging data). In this work, we empirically evaluate and compare a number of different classifiers and classifier ensembles in a multi-site setting, for voxel-wise detection of prostate cancer (PCa) using radiomic texture features derived from high-resolution in vivo T2-weighted (T2w) MRI. METHODS: Twelve different supervised classifier schemes: Quadratic Discriminant Analysis (QDA), Support Vector Machines (SVMs), naive Bayes, Decision Trees (DTs), and their ensemble variants (bagging, boosting), were compared in terms of classification accuracy as well as execution time. Our study utilized 85 prostate cancer T2w MRI datasets acquired from across 3 different institutions (1 for discovery, 2 for independent validation), from patients who later underwent radical prostatectomy. Surrogate ground truth for disease extent on MRI was established by expert annotation of pre-operative MRI through spatial correlation with corresponding ex vivo whole-mount histology sections. Classifier accuracy in detecting PCa extent on MRI on a per-voxel basis was evaluated via area under the ROC curve. RESULTS: The boosted DT classifier yielded the highest cross-validated AUC (= 0.744) for detecting PCa in the discovery cohort. However, in independent validation, the boosted QDA classifier was identified as the most accurate and robust for voxel-wise detection of PCa extent (AUCs of 0.735, 0.683, 0.768 across the 3 sites). The next most accurate and robust classifier was the single QDA classifier, which also enjoyed the advantage of significantly lower computation times compared to any of the other methods. CONCLUSIONS: Our results therefore suggest that simpler classifiers (such as QDA and its ensemble variants) may be more robust, accurate, and efficient for prostate cancer CAD problems, especially in the context of multi-site validation.
Viswanath Satish E; Chirra Prathyush V; Yim Michael C; Rofsky Neil M; Purysko Andrei S; Rosen Mark A; Bloch B Nicolas; Madabhushi Anant
BMC medical imaging
2019
2019-02
<a href="http://doi.org/10.1186/s12880-019-0308-6" target="_blank" rel="noreferrer noopener">10.1186/s12880-019-0308-6</a>
Radiomic characterization of perirectal fat on MRI enables accurate assessment of tumor regression and lymph node metastasis in rectal cancers after chemoradiation
Evaluating tumor regression of rectal cancers via MRI after standard-of-care chemoradiation therapy (CRT) remains highly challenging for radiologists. While the tumor region-of-interest (ROI) on post-CRT rectal MRI is difficult to localize, an underexplored region is the perirectal fat (surrounding tumor and rectum) where residual cancer cells and positive lymph nodes are known to be present. Recent studies have shown that physiologic environments surrounding tumor regions may provide complementary information that is predictive of response to CRT and patient survival. We present initial results of characterizing perirectal fat regions on MRI via radiomics, towards capturing sub-visual details related to rectal tumor or nodal response to CRT. A total of 37 rectal cancer patients for whom MRIs as well as pathologic tumor staging were available post-CRT were included in this study. Region-wise radiomic features were extracted from expert annotated perirectal fat regions and a 2-stage feature selection was employed to identify the most relevant features. Radiomic entropy of perirectal fat was found to be over-expressed in patients with poor tumor or nodal response post-CRT, albeit with different spatial distributions. In a leave-one-patient-out cross validation setting, a quadratic discriminant analysis (QDA) classifier trained on top radiomic features from the perirectal fat achieved AUCs of 0.77 (for differentiating incomplete vs marked tumor regression) and 0.75 (for differentiating lymph node positive from negative patients). By comparison, perirectal fat intensities achieved significantly poorer AUCs in both tasks. Our results indicate perirectal fat on post-CRT MRI may be highly relevant for evaluating CRT response and informing follow-on interventions in rectal cancers.
Yim Michael C; Wei Zhouping; Antunes Jacob; Sehgal Neil K R; Bera Kaustav; Brady Justin T; Friedman Kenneth; Willis Joseph; Purysko Andrei; Paspulati Raj; Madabhushi Anant; Viswanath Satish E
Medical Imaging 2019: Image-guided Procedures, Robotic Interventions, And Modeling
2019
2019
Conference Paper
<a href="http://doi.org/10.1117/12.2512612" target="_blank" rel="noreferrer noopener">10.1117/12.2512612</a>