Vinay Duddalwar

Study Design: Narrative review. Objectives: Back pain is one of the leading causes of disability worldwide. While conventional imaging interpretation remains subjective and expertise-dependent, radiomics offers quantitative, data-driven analysis of medical images. We aimed to evaluate the current literature for the application of radiomics in: (1) soft tissue characterization, (2) hard tissue analysis, and (3) treatment outcome prediction in back pain conditions. Methods: We conducted a PRISMA-style literature search across PubMed, Google Scholar, Wiley, Springer, and IEEE Xplore, focusing on studies from the past 4 years. From 296 identified articles, 22 met inclusion criteria based on their use of radiomic methods and association with pain outcomes. Results: Current literature demonstrates that in many, but not all cases, using radiomics improves clinical models for soft and hard tissue diagnostics as well as for prognosis and treatment prediction. However, the improvements can be minor. There also exist limitations that prevent widespread clinical adoption of radiomics, including a lack of standardization in image acquisition/analysis protocols, homogeneity of patient populations studied, and inadequate integration with existing clinical imaging systems. Additionally, much current work is based on retrospective data instead of real-world data, where there is often an added complexity. Yet, there is increasing work in developing combined models where clinical features, demographics, and patient history are used to enhance the output and accuracy of radiomics. Conclusions: Radiomics can improve back pain diagnosis and treatment. Future directions should focus on developing generalizable radiomics models applicable to broad patient populations, imaging systems, and clinician-interpretable interfaces.

Clear cell renal cell carcinoma (ccRCC) is the most common type of renal cancer. Extensive efforts have been made to utilize radiomics from computed tomography (CT) imaging to predict tumor immune microenvironment (TIME) measurements. This study proposes a Green Learning (GL) framework for approximating tissue-based biomarkers from CT scans, focusing on the PD-L1 expression and CD68 tumor-associated macrophages (TAMs) in ccRCC. Our approach includes radiomic feature extraction, redundancy removal, and supervised feature selection through a discriminant feature test (DFT), a relevant feature test (RFT), and least-squares normal transform (LNT) for robust feature generation. For the PD-L1 expression in 52 ccRCC patients, treated as a regression problem, our GL model achieved a 5-fold cross-validated mean squared error (MSE) of 0.0041 and a Mean Absolute Error (MAE) of 0.0346. For the TAM population (CD68+/PanCK+), analyzed in 78 ccRCC patients as a binary classification task (at a 0.4 threshold), the model reached a 10-fold cross-validated Area Under the Receiver Operating Characteristic (AUROC) of 0.85 (95% CI [0.76, 0.93]) using 10 LNT-derived features, improving upon the previous benchmark of 0.81. This study demonstrates the potential of GL in radiomic analyses, offering a scalable, efficient, and interpretable framework for the non-invasive approximation of key biomarkers.

As artificial intelligence advances in medical image analysis, its environmental impact remains largely overlooked. This study analyzes the energy demands of AI workflows for medical image segmentation using the popular Kidney Tumor Segmentation-2019 (KiTS-19) dataset. It examines how training and inference differ in energy consumption, focusing on factors that influence resource usage, such as computational complexity, memory access, and I/O operations. To address these aspects, we evaluated three variants of convolution—Standard Convolution, Depthwise Convolution, and Group Convolution—combined with optimization techniques such as Mixed Precision and Gradient Accumulation. While training is energy-intensive, the recurring nature of inference often results in significantly higher cumulative energy consumption over a model’s life cycle. Depthwise Convolution with Mixed Precision achieves the lowest energy consumption during training while maintaining strong performance, making it the most energy-efficient configuration among those tested. In contrast, Group Convolution fails to achieve energy efficiency due to significant input/output overhead. These findings emphasize the need for GPU-centric strategies and energy-conscious AI practices, offering actionable guidance for designing scalable, sustainable innovation in medical image analysis.

Radiomics involves extracting quantitative features from medical images, resulting in high-dimensional data. Unsupervised clustering has been used to discover patterns in radiomic features, potentially yielding hidden biological insights. However, its effectiveness depends on the selection of dimensionality reduction techniques, clustering methods, and hyperparameter optimization, an area with limited exploration in the literature. We present a novel bootstrapping-based hyperparameter search approach to optimize clustering efficacy, treating dimensionality reduction and clustering as a connected process chain. The hyperparameter search was guided by the Adjusted Rand Index (ARI) and Davies–Bouldin Index (DBI) within a bootstrapping framework of 100 iterations. The cluster assignments were generated through 10-fold cross-validation, and a grid search strategy was used to explore hyperparameter combinations. We evaluated ten unsupervised learning pipelines using both simulation studies and real-world radiomics data derived from multiphase CT images of renal cell carcinoma. In simulations, we found that Non-negative Matrix Factorization (NMF) and Spectral Clustering outperformed the traditional Principal Component Analysis (PCA)-based approach. The best-performing pipeline (NMF followed by K-means clustering) successfully identified all three simulated clusters, achieving a Cramér’s V of 0.9. The simulation also established a reference framework for understanding the concordance patterns among different pipelines under varying strengths of clustering effects. High concordance reflects strong clustering. In the real-world data application, we observed a moderate clustering effect, which aligned with the weak associations to clinical outcomes, as indicated by the highest AUROC of 0.63.

Purpose This study aimed to develop and evaluate a machine learning pipeline using multiphase contrast-enhanced CT images and clinical data to classify renal tumors as benign, malignant-indolent, or malignant-aggressive, while assessing the contribution of each data source to the classification. Methods In this retrospective study, 448 patients (mean age: 60.7±12.6 years, 306 male, 142 female) who underwent nephrectomy and preoperative CECT between June 2008 and July 2018 were included. Tumors were histologically categorized as benign-indolent, malignant-indolent, or malignant-aggressive. Self-supervised feature extraction converted 4-phase CECT images into 512 real-valued features, combined with clinical data and tumor size for classification. Two machine learning classifiers, random forest (RF) and multi-layer perceptron (MLP), were used to predict tumor type. Nested five-fold cross-validation was employed for hyperparameter tuning and model evaluation, and performance was assessed using area under the curve (AUC) analysis. Results The best-performing models achieved an AUC of 0.90 (95% CI: 0.88–0.93) for classifying indolent versus aggressive tumors and 0.76 (95% CI: 0.71–0.81) for benign versus malignant tumors. Models incorporating tumor size significantly improved classification accuracy. RF classifiers excelled in distinguishing indolent from aggressive tumors, while MLP classifiers performed better for benign versus malignant classification. Conclusion The machine learning pipeline demonstrated high accuracy in differentiating aggressive from indolent renal tumors, offering valuable prognostic insights for personalized treatment. Tumor size was a critical factor, complementing CECT images and clinical data. These findings highlight the potential of ML techniques in enhancing renal tumor risk stratification.

Purpose: To create a nomogram to predict the absence of clinically significant prostate cancer (CSPCa) in males with non-suspicion multiparametric magnetic resonance imaging (mpMRI) undergoing prostate biopsy (PBx).

Image imputation refers to the task of generating a type of medical image given images of another type. This task becomes challenging when the difference between the available images, and the image to be imputed is large. In this manuscript, one such application is considered. It is derived from the dynamic contrast enhanced computed tomography (CECT) imaging of the kidneys: given an incomplete sequence of three CECT images, we are required to impute the missing image. This task is posed as one of probabilistic inference and a generative algorithm to generate samples of the imputed image, conditioned on the available images, is developed, trained, and tested. The output of this algorithm is the “best guess” of the imputed image, and a pixel-wise image of variance in the imputation. It is demonstrated that this best guess is more accurate than those generated by other, deterministic deep-learning based algorithms, including ones which utilize additional information and more complex loss terms. It is also shown that the pixel-wise variance image, which quantifies the confidence in the reconstruction, can be used to determine whether the result of the imputation meets a specified accuracy threshold and is therefore appropriate for a downstream task.