Machine learning model development for quantitative analysis of CT heterogeneity in canine hepatic masses may predict histologic malignancy

Abstract

Tumor heterogeneity is a well‐established marker of biologically aggressive neoplastic processes and is associated with local recurrence and distant metastasis. Quantitative analysis of CT textural features is an indirect measure of tumor heterogeneity and therefore may help predict malignant disease. The purpose of this retrospective, secondary analysis study was to quantitatively evaluate CT heterogeneity in dogs with histologically confirmed liver masses to build a predictive model for malignancy. Forty dogs with liver tumors and corresponding histopathologic evaluation from a previous prospective study were included. Triphasic image acquisition was standardized across dogs and whole liver and liver mass were contoured on each precontrast and delayed postcontrast dataset. First‐order and second‐order indices were extracted from contoured regions. Univariate analysis identified potentially significant indices that were subsequently used for top‐down model construction. Multiple quadratic discriminatory models were constructed and tested, including individual models using both postcontrast and precontrast whole liver or liver mass volumes. The best performing model utilized the CT features voxel volume and uniformity from postcontrast mass contours; this model had an accuracy of 0.90, sensitivity of 0.67, specificity of 1.0, positive predictive value of 1.0, negative predictive value of 0.88, and precision of 1.0. Heterogeneity indices extracted from delayed postcontrast CT hepatic mass contours were more informative about tumor type compared to indices from whole liver contours, or from precontrast hepatic mass and whole liver contours. Results demonstrate that CT radiomic feature analysis may hold clinical utility as a noninvasive method of predicting hepatic malignancy and may influence diagnostic or therapeutic approaches.

A deep learning model for CT-based kidney volume determination in dogs and normal reference definition

“…There have been several studies on texture analysis and the development of machine learning algorithms for canine radiographs ( 27 – 29 ) and CT images ( 30 ) in veterinary medicine, but there is no published method to detect the kidney and determine its volume in dogs from CT images using deep learning models.…”

Artificial intelligence in veterinary diagnostic imaging: A literature review

“…The quadratic discriminant analysis model applied to postcontrast images performed best, differentiating benign and malignant masses with an accuracy of 0.9, sensitivity of 0.67, and specificity of 1.0. 62 This investigation and others [12][13][14] show potential for texture analysis to serve as an additional non-invasive assay to supplement subjective imaging evaluation in differentiation of malignant neoplasms from benign masses.…”

“…57 Numerous attempts with conventional imaging such as CT angiography have been made to differentiate benign and malignant hepatic masses using CT angiography with limited success. [58][59][60][61] Shaker et al 62 applied quadratic discriminant analysis (a non-NN classification model) to 40 triple-phase CT angiographic studies to evaluate the heterogeneity of canine hepatic masses and predict malignancy. Masses were manually segmented with hand-drawn ROIs, window width and level were manually adjusted as needed, and precontrast, arterial, venous, and delayed phase images were co-registered, and radiomics features such as diameter and coarseness were extracted.…”

“…Numerous attempts with conventional imaging such as CT angiography have been made to differentiate benign and malignant hepatic masses using CT angiography with limited success 58–61 . Shaker et al 62 . applied quadratic discriminant analysis (a non‐NN classification model) to 40 triple‐phase CT angiographic studies to evaluate the heterogeneity of canine hepatic masses and predict malignancy.…”

The adding value of contrast-enhanced CT radiomics: Differentiating tuberculosis from non-tuberculous infectious lesions presenting as solid pulmonary nodules or masses

“…Shape-based features are mainly used to describe the geometry of the lesion. In general, texture features can indirectly reflect the heterogeneity of the tumor ( 13 , 26 ). In this study, the images are resampled to achieve voxel isotropy, and voxel size resampling can greatly improve the proportion of robust features ( 27 ).…”