Evaluating the relationship between magnetic resonance image quality metrics and deep learning-based segmentation accuracy of brain tumors

BACKGROUND

Magnetic resonance imaging (MRI) scans are known to suffer from a variety of acquisition artifacts as well as equipment-based variations that impact image appearance and segmentation performance. It is still unclear whether a direct relationship exists between magnetic resonance (MR) image quality metrics (IQMs) (e.g., signal-to-noise, contrast-to-noise) and segmentation accuracy.

PURPOSE

Deep learning (DL) approaches have shown significant promise for automated segmentation of brain tumors on MRI but depend on the quality of input training images. We sought to evaluate the relationship between IQMs of input training images and DL-based brain tumor segmentation accuracy toward developing more generalizable models for multi-institutional data.

METHODS

We trained a 3D DenseNet model on the BraTS 2020 cohorts for segmentation of tumor subregions enhancing tumor (ET), peritumoral edematous, and necrotic and non-ET on MRI; with performance quantified via a 5-fold cross-validated Dice coefficient. MRI scans were evaluated through the open-source quality control tool MRQy, to yield 13 IQMs per scan. The Pearson correlation coefficient was computed between whole tumor (WT) dice values and IQM measures in the training cohorts to identify quality measures most correlated with segmentation performance. Each selected IQM was used to group MRI scans as "better" quality (BQ) or "worse" quality (WQ), via relative thresholding. Segmentation performance was re-evaluated for the DenseNet model when (i) training on BQ MRI images with validation on WQ images, as well as (ii) training on WQ images, and validation on BQ images. Trends were further validated on independent test sets derived from the BraTS 2021 training cohorts.

RESULTS

For this study, multimodal MRI scans from the BraTS 2020 training cohorts were used to train the segmentation model and validated on independent test sets derived from the BraTS 2021 cohort. Among the selected IQMs, models trained on BQ images based on inhomogeneity measurements (coefficient of variance, coefficient of joint variation, coefficient of variation of the foreground patch) and the models trained on WQ images based on noise measurement peak signal-to-noise ratio (SNR) yielded significantly improved tumor segmentation accuracy compared to their inverse models.

CONCLUSIONS

Our results suggest that a significant correlation may exist between specific MR IQMs and DenseNet-based brain tumor segmentation performance. The selection of MRI scans for model training based on IQMs may yield more accurate and generalizable models in unseen validation.

Abstract 850: Patient-specific, organ-scale forecasting of prostate cancer growth in active surveillance

Active surveillance (AS) is an established clinical strategy for the management of prostate cancer (PCa) exhibiting low to intermediate risk. In AS, treatment is delayed until progression to higher-risk disease is detected during the close monitoring of patients via longitudinal multiparametric magnetic resonance imaging (mpMRI) scans, biopsies, and Prostate Specific Antigen (PSA) tests. Thus, AS has been regarded as a promising strategy to address the current rates of overtreatment and undertreatment in PCa. However, current AS protocols rely on an observational paradigm, which may delay the detection of tumor progression, and the testing frequency is largely fixed according to population studies, which impedes the design of personalized monitoring plans. To address these issues, we propose to advance AS towards a predictive patient-specific paradigm by leveraging computational tumor forecasts obtained with a biomechanistic model informed by the imaging and clinical data collected during standard AS for each individual patient.

Here, we present a preliminary study in a cohort of eight PCa patients who enrolled in AS and had three mpMRI scans over a period of 2.6 to 5.6 years. Our model describes PCa growth in terms of the dynamics of tumor cell density as a combination of tumor cell mobility and net proliferation, which are formulated as a diffusion process and logistic growth, respectively. The model is implemented in each patient’s prostate geometry, which is segmented on T2-weighted MRI data. Tumor cell density estimates are derived from Apparent Diffusion Coefficient (ADC) maps obtained from diffusion-weighted MRI data. To facilitate modeling, the longitudinal imaging datasets are non-rigidly co-registered for each patient. We initialize the model with the tumor cell density map obtained from the ADC map of the first mpMRI scan. Then, the model is parameterized by minimizing the model-data mismatch in tumor cell density at the date of a second mpMRI scan. Finally, we perform a tumor forecast up to the date of a third mpMRI scan, which we use to assess the model-data agreement of our predictions of PCa growth.

We obtained a concordance correlation coefficient (CCC) for tumor volume and global tumor cell count of 0.87 and 0.95 during model calibration and of 0.91 and 0.87 at forecasting horizon, respectively. For each patient, the spatial fit of tumor cell density yielded a median Dice score and CCC of 0.77 and 0.50 at the second mpMRI date, respectively. Likewise, the tumor cell density predictions at the third scan date resulted in a median Dice coefficient and CCC of 0.74 and 0.51 across the patient cohort, respectively. Thus, while further model development and performance assessment over lager cohorts are required, these results suggest that our forecasting technology is a promising tool to predict PCa progression in AS and, hence, identify patients who require treatment early during the course of AS.

Citation Format: Guillermo Lorenzo Gomez, Chengyue Wu, Joshua P. Yung, John F. Ward, Hector Gomez, Alessandro Reali, Thomas E. Yankeelov, Aradhana M. Venkatesan, Thomas J. Hughes. Patient-specific, organ-scale forecasting of prostate cancer growth in active surveillance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 850.

Temperature mapping of exothermic in situ chemistry: imaging of thermoembolization via MR

Purpose: MR temperature imaging (MRTI) was employed for visualizing the spatiotemporal evolution of the exotherm of thermoembolization, an investigative transarterial treatment for solid tumors. Materials and methods: Five explanted kidneys were injected with thermoembolic solutions, and monitored by MRTI. In three nonselective experiments, 5 ml of 4 mol/l dichloroacetyl chloride (DCA-Cl) solution in a hydrocarbon vehicle was injected via the main renal artery. For two of these three, MRTI temperature data were compared to fiber optic thermal probes. Another two kidneys received selective injections, treating only portions of the kidneys with 1 ml of 2 mol/l DCA-Cl. MRTI data were acquired and compared to changes in pre- and post-injection CT. Specimens were bisected and photographed for gross pathology 24 h post-procedure. Results: MRTI temperature estimates were within ±1 °C of the probes. In experiments without probes, MRTI measured increases of 30 °C. Some regions had not reached peak temperature by the end of the >18 min acquisition. MRTI indicated the initial heating occurred in the renal cortex, gradually spreading more proximally toward the main renal artery. Gross pathology showed the nonselective injection denatured the entire kidney whereas in the selective injections, only the treated territory was coagulated. Conclusion: The spatiotemporal evolution of thermoembolization was visualized for the first time using noninvasive MRTI, providing unique insight into the thermodynamics of thermoembolization. Précis Thermoembolization is being investigated as a novel transarterial treatment. In order to begin to characterize delivery of this novel treatment modality and aid translation from the laboratory to patients, we employ MR temperature imaging to visualize the spatiotemporal distribution of temperature from thermoembolization in ex vivo tissue.

Referenceless magnetic resonance temperature imaging using Gaussian process modeling

PURPOSE

During magnetic resonance (MR)-guided thermal therapies, water proton resonance frequency shift (PRFS)-based MR temperature imaging can quantitatively monitor tissue temperature changes. It is widely known that the PRFS technique is easily perturbed by tissue motion, tissue susceptibility changes, magnetic field drift, and modality-dependent applicator-induced artifacts. Here, a referenceless Gaussian process modeling (GPM)-based estimation of the PRFS is investigated as a methodology to mitigate unwanted background field changes. The GPM offers a complementary trade-off between data fitting and smoothing and allows prior information to be used. The end result being the GPM provides a full probabilistic prediction and an estimate of the uncertainty.

METHODS

GPM was employed to estimate the covariance between the spatial position and MR phase measurements. The mean and variance provided by the statistical model extrapolated background phase values from nonheated neighboring voxels used to train the model. MR phase predictions in the heating ROI are computed using the spatial coordinates as the test input. The method is demonstrated in ex vivo rabbit liver tissue during focused ultrasound heating with manually introduced perturbations (n = 6) and in vivo during laser-induced interstitial thermal therapy to treat the human brain (n = 1) and liver (n = 1).

RESULTS

Temperature maps estimated using the GPM referenceless method demonstrated a RMS error of <0.8°C with artifact-induced reference-based MR thermometry during ex vivo heating using focused ultrasound. Nonheated surrounding areas were <0.5°C from the artifact-free MR measurements. The GPM referenceless MR temperature values and thermally damaged regions were within the 95% confidence interval during in vivo laser ablations.

CONCLUSIONS

A new approach to estimation for referenceless PRFS temperature imaging is introduced that allows for an accurate probabilistic extrapolation of the background phase. The technique demonstrated reliable temperature estimates in the presence of the background phase changes and was demonstrated useful in the in vivo brain and liver ablation scenarios presented.

Quantitative comparison of thermal dose models in normal canine brain

PURPOSE

Minimally invasive thermal ablative therapies as alternatives to conventional surgical management of solid tumors and other pathologies is increasing owing to the potential benefits of performing these procedures in an outpatient setting with reduced complications and comorbidity. Magnetic resonance temperature imaging (MRTI) measurement allows existing thermal dose models to use the spatiotemporal temperature history to estimate the thermal damage to tissue. However, the various thermal dose models presented in the literature employ different parameters and thresholds, affecting the reliability of thermal dosimetry. In this study, the authors quantitatively compared three thermal dose models (Arrhenius rate process, CEM43, and threshold temperature) using the dice similarity coefficient (DSC).

METHODS

The DSC was used to compare the spatial overlap between the region of thermal damage as predicted by the models for in vivo normal canine brain during thermal therapy to the region of thermal damage as revealed by contrast-enhanced T1-weighted images acquired immediately after therapy (< 20 min). The outer edge of the hyperintense rim of the ablation region was used as the surrogate marker for the limits of thermal coagulation. The DSC was also used to investigate the impact of varying the thresholds on each models' ability to predict the zone of thermal necrosis.

RESULTS

At previously reported thresholds, the authors found that all three models showed good agreement (defined as DSC > 0.7) with post-treatment imaging. All three models examined across the range of commonly applied thresholds consistently showed highly accurate spatial overlap, low variability, and little dependence on temperature uncertainty. DSC values corresponding to cited thresholds were not significantly different from peak DSC values.

CONCLUSIONS

Thus, the authors conclude that the all three thermal dose models can be used as a reliable surrogate for postcontrast tissue damage verification imaging in rapid ablation procedures and can also be used to enhance the capability of MRTI to control thermal therapy in real time.

Disc movements over the condylar head. Radiographical study on autopsy materials

A radiographical study has been performed to evaluate the movement of the disc posterior band over the condylar head during mandibular opening. Six formal free embalmed subjects were selected as 'normal'. Micro stainless steel balls were used as landmarks both into the bone and into the disc for X-ray identification. The data was processed at each phase of movement in selected planes. The true coordinates of each disc landmark gave the potential to calculate the linear length of their path and any distance change between them. It can be concluded that there is a 5% width reduction during the opening movement and it also suggests strongly that a fair amount of translation movement between the disc and the condyle occurs in the lower compartment.

Anatomy of the temporomandibular joint and related structures in the frontal plane

Conclusions about the anatomy and physiology of the temporomandibular joint have too often been based on sagittal representations. The aim of the article is to describe frontal serial cuts of an acrylic embedded articulation of a subject with teeth in intercuspal position. The frontal plane is best to visualise the difference between lateral and medial aspects of structures such as the disk-condyle-complex, the articular capsule, and related muscles.

Temporomandibular joint serial sections made with mandible in intercuspal position

This article presents a method for preparing block sections of the human temporomandibular joint. The purpose of the study was to examine the relationship of the temporomandibular joint components without disturbing the anatomic arrangement of the parts. The mandibles of seven fresh cadavers were locked into a specific occlusal position (intercuspal) throughout all procedures. An acrylic embedding technique allowed serial sections in any plane without displacement or tearing of mineralized and soft tissues. Photographs were obtained using an Olympus transmitted light microscope with various magnifications from 1 X to 40 X.

Insertions of the lateral pterygoid muscle: an anatomic study of the human temporomandibular joint

The objective of this investigation was to evaluate the anatomic relationships of the lateral pterygoid muscle with the disc-condyle complex using an acrylic embedding technique to obtain anatomic serial cuts of a solid block containing the temporomandibular joint (TMJ) and the neighboring structures. The fibers of the upper and lower heads were found to be fused in front of the TMJ and to constitute medially a strong muscular wall. The lateral third of the anterior band of the disc was free of any muscle insertions and related anteriorly with loose connective tissue; the middle third showed fibers that run under the anterior band of the disc and attached in the upper part of the condylar fovea. Only the medial portion demonstrated both fibers running into the disc and fibers inserting into the bone. The fibers inserted into the bone run under those attached into the disc and terminated below the medial pole of the condyle binding the disc tightly over the medial pole. This study demonstrates that the main insertions of the superior head are not into the disc but into the condyle. Considering the anatomic organization of the upper head, the explanation of anterior displacement of the disc due to a spastic activity of this muscle alone is not probable. Hypotonicity, not hyperactivity, of the upper head may contribute to an anterior and medial disc displacement.