Enhancement pattern mapping technique for improving contrast-to-noise ratios and detectability of hepatobiliary tumors on multiphase computed tomography

Measurable imaging-based changes in enhancement of intrahepatic cholangiocarcinoma after radiotherapy reflect physical mechanisms of response

Background

Although escalated doses of radiation therapy (RT) for intrahepatic cholangiocarcinoma (iCCA) are associated with durable local control (LC) and prolonged survival, uncertainties persist regarding personalized RT based on biological factors. Compounding this knowledge gap, the assessment of RT response using traditional size-based criteria via computed tomography (CT) imaging correlates poorly with outcomes. We hypothesized that quantitative measures of enhancement would more accurately predict clinical outcomes than size-based assessment alone and developed a model to optimize RT.

Methods

Pre-RT and post-RT CT scans of 154 patients with iCCA were analyzed retrospectively for measurements of tumor dimensions (for RECIST) and viable tumor volume using quantitative European Association for Study of Liver (qEASL) measurements. Binary classification and survival analyses were performed to evaluate the ability of qEASL to predict treatment outcomes, and mathematical modeling was performed to identify the mechanistic determinants of treatment outcomes and to predict optimal RT protocols.

Results

Multivariable analysis accounting for traditional prognostic covariates revealed that percentage change in viable volume following RT was significantly associated with OS, outperforming stratification by RECIST. Binary classification identified ≥33% decrease in viable volume to optimally correspond to response to RT. The model-derived, patient-specific tumor enhancement growth rate emerged as the dominant mechanistic determinant of treatment outcome and yielded high accuracy of patient stratification (80.5%), strongly correlating with the qEASL-based classifier.

Conclusion

Following RT for iCCA, changes in viable volume outperformed radiographic size-based assessment using RECIST for OS prediction. CT-derived tumor-specific mathematical parameters may help optimize RT for resistant tumors.

Unsupervised low-dose CT denoising using bidirectional contrastive network

BACKGROUND AND OBJECTIVE

Low-dose computed tomography (LDCT) scans significantly reduce radiation exposure, but introduce higher levels of noise and artifacts that compromise image quality and diagnostic accuracy. Supervised learning methods have proven effective in denoising LDCT images, but are hampered by the need for large, paired datasets, which pose significant challenges in data acquisition. This study aims to develop a robust unsupervised LDCT denoising method that overcomes the reliance on paired LDCT and normal-dose CT (NDCT) samples, paving the way for more accessible and practical denoising techniques.

METHODS

We propose a novel unsupervised network model, Bidirectional Contrastive Unsupervised Denoising (BCUD), for LDCT denoising. This model innovatively combines a bidirectional network structure with contrastive learning theory to map the precise mutual correspondence between the noisy LDCT image domain and the clean NDCT image domain. Specifically, we employ dual encoders and discriminators for domain-specific data generation, and use unique projection heads for each domain to adaptively learn customized embedded representations. We then align corresponding features across domains within the learned embedding spaces to achieve effective noise reduction. This approach fundamentally improves the model's ability to match features in latent space, thereby improving noise reduction while preserving fine image detail.

RESULTS

Through extensive experimental validation on the AAPM-Mayo public dataset and real-world clinical datasets, the proposed BCUD method demonstrated superior performance. It achieved a peak signal-to-noise ratio (PSNR) of 31.387 dB, a structural similarity index measure (SSIM) of 0.886, an information fidelity criterion (IFC) of 2.305, and a visual information fidelity (VIF) of 0.373. Notably, subjective evaluation by radiologists resulted in a mean score of 4.23, highlighting its advantages over existing methods in terms of clinical applicability.

CONCLUSIONS

This paper presents an innovative unsupervised LDCT denoising method using a bidirectional contrastive network, which greatly improves clinical applicability by eliminating the need for perfectly matched image pairs. The method sets a new benchmark in unsupervised LDCT image denoising, excelling in noise reduction and preservation of fine structural details.

Enhancement Pattern Mapping for Early Detection of Hepatocellular Carcinoma in Patients with Cirrhosis

Background and Aims

Limited methods exist to accurately characterize the risk of malignant progression of liver lesions. Enhancement pattern mapping (EPM) measures voxel-based root mean square deviation (RMSD) of parenchyma and the contrast-to-noise (CNR) ratio enhances in malignant lesions. This study investigates the utilization of EPM to differentiate between HCC versus cirrhotic parenchyma with and without benign lesions.

Methods

Patients with cirrhosis undergoing MRI surveillance were studied prospectively. Cases (n=48) were defined as patients with LI-RADS 3 and 4 lesions who developed HCC during surveillance. Controls (n=99) were patients with and without LI-RADS 3 and 4 lesions who did not develop HCC. Manual and automated EPM signals of liver parenchyma between cases and controls were quantitatively validated on an independent patient set using cross validation with manual methods avoiding parenchyma with artifacts or blood vessels.

Results

With manual EPM, RMSD of 0.37 was identified as a cutoff for distinguishing lesions that progress to HCC from background parenchyma with and without lesions on pre-diagnostic scans (median time interval 6.8 months) with an area under the curve (AUC) of 0.83 (CI: 0.73-0.94) and a sensitivity, specificity, and accuracy of 0.65, 0.97, and 0.89, respectively. At the time of diagnostic scans, a sensitivity, specificity, and accuracy of 0.79, 0.93, and 0.88 were achieved with manual EPM with an AUC of 0.89 (CI: 0.82-0.96). EPM RMSD signals of background parenchyma that did not progress to HCC in cases and controls were similar (case EPM: 0.22 ± 0.08, control EPM: 0.22 ± 0.09, p=0.8). Automated EPM produced similar quantitative results and performance.

Conclusion

With manual EPM, a cutoff of 0.37 identifies quantifiable differences between HCC cases and controls approximately six months prior to diagnosis of HCC with an accuracy of 89%.

CoreDiff: Contextual Error-Modulated Generalized Diffusion Model for Low-Dose CT Denoising and Generalization

Low-dose computed tomography (CT) images suffer from noise and artifacts due to photon starvation and electronic noise. Recently, some works have attempted to use diffusion models to address the over-smoothness and training instability encountered by previous deep-learning-based denoising models. However, diffusion models suffer from long inference time due to a large number of sampling steps involved. Very recently, cold diffusion model generalizes classical diffusion models and has greater flexibility. Inspired by cold diffusion, this paper presents a novel COntextual eRror-modulated gEneralized Diffusion model for low-dose CT (LDCT) denoising, termed CoreDiff. First, CoreDiff utilizes LDCT images to displace the random Gaussian noise and employs a novel mean-preserving degradation operator to mimic the physical process of CT degradation, significantly reducing sampling steps thanks to the informative LDCT images as the starting point of the sampling process. Second, to alleviate the error accumulation problem caused by the imperfect restoration operator in the sampling process, we propose a novel ContextuaL Error-modulAted Restoration Network (CLEAR-Net), which can leverage contextual information to constrain the sampling process from structural distortion and modulate time step embedding features for better alignment with the input at the next time step. Third, to rapidly generalize the trained model to a new, unseen dose level with as few resources as possible, we devise a one-shot learning framework to make CoreDiff generalize faster and better using only one single LDCT image (un)paired with normal-dose CT (NDCT). Extensive experimental results on four datasets demonstrate that our CoreDiff outperforms competing methods in denoising and generalization performance, with clinically acceptable inference time. Source code is made available at https://github.com/qgao21/CoreDiff.

Radiomics, deep learning and early diagnosis in oncology

Medical imaging, including X-ray, computed tomography (CT), and magnetic resonance imaging (MRI), plays a critical role in early detection, diagnosis, and treatment response prediction of cancer. To ease radiologists' task and help with challenging cases, computer-aided diagnosis has been developing rapidly in the past decade, pioneered by radiomics early on, and more recently, driven by deep learning. In this mini-review, I use breast cancer as an example and review how medical imaging and its quantitative modeling, including radiomics and deep learning, have improved the early detection and treatment response prediction of breast cancer. I also outline what radiomics and deep learning share in common and how they differ in terms of modeling procedure, sample size requirement, and computational implementation. Finally, I discuss the challenges and efforts entailed to integrate deep learning models and software in clinical practice.

Predictive Modeling for Voxel-Based Quantification of Imaging-Based Subtypes of Pancreatic Ductal Adenocarcinoma (PDAC): A Multi-Institutional Study

Previously, we characterized qualitative imaging-based subtypes of pancreatic ductal adenocarcinoma (PDAC) on computed tomography (CT) scans. Conspicuous (high delta) PDAC tumors are more likely to have aggressive biology and poorer clinical outcomes compared to inconspicuous (low delta) tumors. Here, we developed a quantitative classification of this imaging-based subtype (quantitative delta; q-delta). Retrospectively, baseline pancreatic protocol CT scans of three cohorts (cohort#1 = 101, cohort#2 = 90 and cohort#3 = 16 [external validation]) of patients with PDAC were qualitatively classified into high and low delta. We used a voxel-based method to volumetrically quantify tumor enhancement while referencing normal-pancreatic-parenchyma and used machine learning-based analysis to build a predictive model. In addition, we quantified the stromal content using hematoxylin- and eosin-stained treatment-naïve PDAC sections. Analyses revealed that PDAC quantitative enhancement values are predictive of the qualitative delta scoring and were used to build a classification model (q-delta). Compared to high q-delta, low q-delta tumors were associated with improved outcomes, and the q-delta class was an independent prognostic factor for survival. In addition, low q-delta tumors had higher stromal content and lower cellularity compared to high q-delta tumors. Our results suggest that q-delta classification provides a clinically and biologically relevant tool that may be integrated into ongoing and future clinical trials.

Imaging-Based Subtypes of Pancreatic Ductal Adenocarcinoma Exhibit Differential Growth and Metabolic Patterns in the Pre-Diagnostic Period: Implications for Early Detection

BACKGROUND

Previously, we characterized subtypes of pancreatic ductal adenocarcinoma (PDAC) on computed-tomography (CT) scans, whereby conspicuous (high delta) PDAC tumors are more likely to have aggressive biology and poorer clinical outcomes compared to inconspicuous (low delta) tumors. Here, we hypothesized that these imaging-based subtypes would exhibit different growth-rates and distinctive metabolic effects in the period prior to PDAC diagnosis.

MATERIALS AND METHODS

Retrospectively, we evaluated 55 patients who developed PDAC as a second primary cancer and underwent serial pre-diagnostic (T0) and diagnostic (T1) CT-scans. We scored the PDAC tumors into high and low delta on T1 and, serially, obtained the biaxial measurements of the pancreatic lesions (T0-T1). We used the Gompertz-function to model the growth-kinetics and estimate the tumor growth-rate constant (α) which was used for tumor binary classification, followed by cross-validation of the classifier accuracy. We used maximum-likelihood estimation to estimate initiation-time from a single cell (10^-6 mm³) to a 10 mm³ tumor mass. Finally, we serially quantified the subcutaneous-abdominal-fat (SAF), visceral-abdominal-fat (VAF), and muscles volumes (cm³) on CT-scans, and recorded the change in blood glucose (BG) levels. T-test, likelihood-ratio, Cox proportional-hazards, and Kaplan-Meier were used for statistical analysis and p-value <0.05 was considered significant.

RESULTS

Compared to high delta tumors, low delta tumors had significantly slower average growth-rate constants (0.024 month^-1 vs. 0.088 month^-1, p<0.0001) and longer average initiation-times (14 years vs. 5 years, p<0.0001). α demonstrated high accuracy (area under the curve (AUC)=0.85) in classifying the tumors into high and low delta, with an optimal cut-off of 0.034 month^-1. Leave-one-out-cross-validation showed 80% accuracy in predicting the delta-class (AUC=0.84). High delta tumors exhibited accelerated SAF, VAF, and muscle wasting (p <0.001), and BG disturbance (p<0.01) compared to low delta tumors. Patients with low delta tumors had better PDAC-specific progression-free survival (log-rank, p<0.0001), earlier stage tumors (p=0.005), and higher likelihood to receive resection after PDAC diagnosis (p=0.008), compared to those with high delta tumors.

CONCLUSION

Imaging-based subtypes of PDAC exhibit distinct growth, metabolic, and clinical profiles during the pre-diagnostic period. Our results suggest that heterogeneous disease biology may be an important consideration in early detection strategies for PDAC.