Hepatocellular adenoma subtyping by qualitative MRI features and machine learning algorithm of integrated qualitative and quantitative features: a proof-of-concept study

AIM

To evaluate hepatocellular adenoma (HCA) subtyping using qualitative magnetic resonance imaging (MRI) features and feasibility of differentiating HCA subtypes using machine learning (ML) of qualitative and quantitative MRI features with histopathology as the reference standard.

MATERIALS AND METHODS

This retrospective study included 39 histopathologically subtyped HCAs (13 hepatocyte nuclear factor (HNF)-1-alpha mutated [HHCA], 11 inflammatory [IHCA], one beta-catenin-mutated [BHCA], and 14 unclassified [UHCA]) in 36 patients. HCA subtyping by two blinded radiologists using the proposed schema of qualitative MRI features and using the random forest algorithm was compared against histopathology. For quantitative features, 1,409 radiomic features were extracted after segmentation and reduced to 10 principle components. Support vector machine and logistic regression was applied to assess HCA subtyping.

RESULTS

Qualitative MRI features with proposed flow chart yielded diagnostic accuracies of 87%, 82%, and 74% for HHCA, IHCA, and UHCA respectively. The ML algorithm based on qualitative MRI features showed AUCs (area under the receiver operating characteristic curve [ROC] curve) of 0.846, 0.642, and 0.766 for diagnosing HHCA, IHCA, and UHCA, respectively. Quantitative radiomic features from portal venous and hepatic venous phase MRI demonstrated AUCs of 0.83 and 0.82, with a sensitivity of 72% and a specificity of 85% in predicting HHCA subtype.

CONCLUSIONS

The proposed schema of integrated qualitative MRI features with ML algorithm provided high accuracy for HCA subtyping while quantitative radiomic features provide value for diagnosis of HHCA. The key qualitative MRI features for differentiating HCA subtypes were concordant between the radiologists and the ML algorithm. These approaches appear promising to better inform clinical management for patients with HCA.

Multiparametric Magnetic Resonance Imaging in the Diagnosis of Clinically Significant Prostate Cancer: an Updated Systematic Review

There has been growing utilisation of multiparametric magnetic resonance imaging (MPMRI) as a non-invasive tool to diagnose and localise clinically significant prostate cancer (CSPCa). This updated systematic review examines the use of MPMRI in patients with an elevated risk of CSPCa who have had a prior negative transrectal ultrasound systematic biopsy (TRUS-SB) and who were biopsy naïve. MEDLINE, EMBASE and the Cochrane Database of Systematic Reviews were searched for existing systematic reviews published up to September 2020. The literature search of the electronic databases combined disease-specific terms (prostate cancer, prostate carcinoma, etc.) and treatment-specific terms (magnetic resonance, etc.). Studies were included if they were randomised controlled trials (RCTs) comparing MPMRI to template transperineal mapping biopsy (TPMB) or to TRUS-SB. Thirty-six RCTs were eligible. For biopsy-naïve men, accuracy of diagnosis of CSPCa showed sensitivities from 87 to 96% and specificities ranging from 29 to 45%. Meta-analyses for CSPCa showed increased detection favouring MPMRI-targeted biopsy over TRUS-SB by 3% (95% confidence interval 0-7%, P = 0.03) and decreased detection of clinically insignificant prostate cancer (CISPCa) favouring MPMRI by 8% (95% confidence interval -11 to 5%, P < 0.00001). Accuracy of MPMRI for men with prior negative biopsy showed sensitivities of 78-100% and specificities of 30-100%. Meta-analyses comparing MPMRI to TRUS-SB showed increased detection of 5% (95% confidence interval 3-7%, P < 0.0001) with a reduction of CISPCa detection of 7% (95% confidence interval 4-9%, P < 0.00001). The growing acceptance of MPMRI utilisation internationally and the recent publication of several RCTs regarding MPMRI in reducing CISPCa detection rates, particularly in biopsy-naïve men, without loss of sensitivity for CSPCa necessitates the synthesis of updated evidence examining MPMRI in the diagnosis of CSPCa.

Multiparametric Magnetic Resonance Imaging in the Diagnosis of Prostate Cancer: A Systematic Review

A systematic review was conducted to investigate the use of multiparametric magnetic resonance imaging (MPMRI) followed by targeted biopsy in the diagnosis of clinically significant prostate cancer (CSPC) and to compare it with transrectal ultrasound-guided (TRUS-guided) systematic biopsy in patients with an elevated risk of prostate cancer who are either biopsy-naive or who have a previous negative TRUS-guided biopsy. MEDLINE, PubMed and EMBASE (1997 to April 2014), the Cochrane Library and six relevant conferences were searched to find eligible studies. Search terms indicative of 'prostate cancer' and 'magnetic resonance imaging' with their alternatives were used. Twelve systematic reviews, 52 full texts and 28 abstracts met the preplanned study selection criteria; data from 15 articles were extracted. In patients with an elevated risk of prostate cancer who were biopsy-naive, MPMRI followed by targeted biopsy could detect 2-13% of CSPC patients whom TRUS-guided systematic biopsy missed; TRUS-guided systematic biopsy could detect 0-7% of CSPC patients whom MPMRI followed by targeted biopsy missed. In patients with an elevated risk of prostate cancer who had a previous negative TRUS-guided biopsy, MPMRI followed by targeted biopsy detected more CSPC patients than repeated TRUS-guided systematic biopsy in all four studies, with a total of 516 patients, but only one study reached a statistically significant difference. In patients with an elevated risk of prostate cancer who are biopsy-naive, there is insufficient evidence for MPMRI followed by targeted biopsy to be considered the standard of care. In patients who had a prior negative TRUS-guided systematic biopsy and show a growing risk of having CSPC, MPMRI followed by targeted biopsy may be helpful to detect more CSPC cases as opposed to a repeat TRUS-guided systematic biopsy.

Early experience in MRI-guided therapies of prostate cancer: HIFU, laser and photodynamic treatment

Prostate cancer screening has resulted in earlier diagnosis with lower-grade disease, leading to over-detection and over-treatment in a significant number of patients. Current whole-gland radical treatments are associated with significant rates of morbidity. The high prevalence of low-risk disease together with an inability to accurately identify those men harboring more aggressive cancers has led to tremendous research in low-morbidity focal therapies for prostate cancer. This review summarizes the early experiences with focal therapy with emphasis on early applications of laser, high-intensity focuses ultrasound, and photodynamic approaches.

Linear and nonlinear viscoelastic modeling of ovine aortic biomechanical properties under in vivo and ex vivo conditions

This study uses linear and nonlinear viscoelastic models to describe the dynamic distention of the aorta induced by time-varying arterial blood pressure. We employ an inverse mathematical modeling approach on a four-parameter (linear) Kelvin viscoelastic model and two five-parameter nonlinear viscoelastic models (arctangent and sigmoid) to infer vascular biomechanical properties under in vivo and ex vivo experimental conditions in ten and eleven male Merino sheep, respectively. We used the Akaike Information Criterion (AIC) as a goodness-of-fit measure. Results show that under both experimental conditions, the nonlinear models generally outperform the linear Kelvin model, as judged by the AIC. Furthermore, the sigmoid nonlinear viscoelastic model consistently achieves the lowest AIC and also matches the zero-stress vessel radii measured ex vivo. Based on these observations, we conclude that the sigmoid nonlinear viscoelastic model best describes the biomechanical properties of ovine large arteries under both experimental conditions considered in this study.

Monitored steady-state excitation and recovery (MSSER) radiation force imaging using viscoelastic models

Acoustic radiation force imaging methods distinguish tissue structure and composition by monitoring tissue responses to applied radiation force excitations. Although these responses are a complex, multidimensional function of the geometric and viscoelastic nature of tissue, simplified discrete biomechanical models offer meaningful insight to the physical phenomena that govern induced tissue motion. Applying Voigt and standard linear viscoelastic tissue models, we present a new radiation force technique - monitored steady-state excitation and recovery (MSSER) imaging - that tracks both steady-state displacement during prolonged force application and transient response following force cessation to estimate tissue mechanical properties such as elasticity and viscosity. In concert with shear wave elasticity imaging (SWEI) estimates for Young's modulus, MSSER methods are useful for estimating tissue mechanical properties independent of the applied force magnitude. We test our methods in gelatin phantoms and excised pig muscle, with confirmation through mechanical property measurement. Our results measured 10.6 kPa, 14.7 kPa, and 17.1 kPa (gelatin) and 122.4 kPa (pig muscle) with less than 10% error. This work demonstrates the feasibility of MSSER imaging and merits further efforts to incorporate relevant mechanical tissue models into the development of novel radiation force imaging techniques.

Vascular targeted photodynamic therapy with palladium-bacteriopheophorbide photosensitizer for recurrent prostate cancer following definitive radiation therapy: assessment of safety and treatment response

PURPOSE

Tookad is a novel intravascular photosensitizer. When activated by 763 nm light, it destroys tumors by damaging their blood supply. It then clears rapidly from the circulatory system. To our knowledge we report the first application of Tookad vascular targeted photodynamic therapy in humans. We assessed the safety, pharmacokinetics and preliminary treatment response as a salvage procedure after external beam radiation therapy.

MATERIALS AND METHODS

Patients received escalating drug doses of 0.1 to 2 mg/kg at a fixed light dose of 100 J/cm or escalated light doses of 230 and 360 J/cm at the 2 mg/kg dose. Four optical fibers were placed transperineally in the prostate, including 2 for light delivery and 2 for light dosimetry. Treatment response was assessed primarily by hypovascular lesion formation on contrast enhanced magnetic resonance imaging and transrectal ultrasound guided biopsies targeting areas of lesion formation and secondarily by serum prostate specific antigen changes.

RESULTS

Tookad vascular targeted photodynamic therapy was technically feasible. The plasma drug concentration was negligible by 2 hours after infusion. In the drug escalation arm 3 of 6 patients responded, as seen on magnetic resonance imaging, including 1 at 1 mg/kg and 2 at 2 mg/kg. The light dose escalation demonstrated an increasing volume of effect with 2 of 3 patients in the first light escalation cohort responding and all 6 responding at the highest light dose with lesions encompassing up to 70% of the peripheral zone. There were no serious adverse events, and continence and potency were maintained.

CONCLUSIONS

Tookad vascular targeted photodynamic therapy salvage therapy is safe and well tolerated. Lesion formation is strongly drug and light dose dependent. Early histological and magnetic resonance imaging responses highlight the clinical potential of Tookad vascular targeted photodynamic therapy to manage post-external beam radiation therapy recurrence.

Triple-phase MDCT of hepatocellular carcinoma

OBJECTIVE

The purpose of this study was to describe the appearances of hepatocellular carcinoma including intralesional contrast washout using a triple-phase liver protocol on an MDCT scanner.

MATERIALS AND METHODS

Fifty-one patients with newly diagnosed hepatocellular carcinoma underwent standardized triple-phase CT using a multidetector scanner. Pathologic proof was obtained in 35 patients (69%); in 16 patients (31%), hepatocellular carcinoma was diagnosed on clinical and laboratory findings. Two radiologists independently reviewed the CT studies for the appearance and attenuation of the lesions. Intralesional washout of contrast material was evaluated subjectively and objectively. Statistical analysis was performed using Fisher's exact test to analyze the relationships between tumor appearance and alpha-fetoprotein level, tumor grade, and risk factor. Correlation between tumor size and appearance was analyzed using the Student's t test and Wilcoxon's rank sum test.

RESULTS

The most common enhancement pattern for hepatocellular carcinoma was hypervascularity on hepatic arterial phase images with a mosaic pattern on both arterial and portal venous images; this finding was seen in 86% and 78% of lesions by the two observers, respectively. A hypervascular component was seen in 96% of lesions by both observers, and the observers recorded 86% and 63% of lesions as showing washout, respectively. Objective washout was present in 76% of lesions. Both subjective and objective washout correlated with an elevated alpha-fetoprotein level (p = 0.01).

CONCLUSION

The appearances of hepatocellular carcinoma on images obtained using MDCT scanners are similar to those described for images obtained using single-detector helical scanners. However, the prevalence of hypervascular hepatocellular carcinoma on MDCT images is higher than previously described on single-detector helical images and most lesions showed washout on portal venous MDCT images.

Perineural cysts presenting as complex adnexal cystic masses on transvaginal sonography

OBJECTIVE

This study describes the sonographic features of sacral perineural cysts that initially presented as adnexal complex cystic masses on transvaginal sonography.

CONCLUSION

Perineural cysts may have a complex cystic appearance, including septation and internal debris, on transvaginal sonography. The extraovarian, extraperitoneal, and posterior location on real-time sonography are suggestive features.

An axisymmetric boundary integral model for incompressible linear viscoelasticity: application to the micropipette aspiration contact problem

The micropipette aspiration test has been used extensively in recent years as a means of quantifying cellular mechanics and molecular interactions at the microscopic scale. However, previous studies have generally modeled the cell as an infinite half-space in order to develop an analytical solution for a viscoelastic solid cell. In this study, an axisymmetric boundary integral formulation of the governing equations of incompressible linear viscoelasticity is presented and used to simulate the micropipette aspiration contact problem. The cell is idealized as a homogeneous and isotropic continuum with constitutive equation given by three-parameter (E, tau 1, tau 2) standard linear viscoelasticity. The formulation is used to develop a computational model via a "correspondence principle" in which the solution is written as the sum of a homogeneous (elastic) part and a nonhomogeneous part, which depends only on past values of the solution. Via a time-marching scheme, the solution of the viscoelastic problem is obtained by employing an elastic boundary element method with modified boundary conditions. The accuracy and convergence of the time-marching scheme are verified using an analytical solution. An incremental reformulation of the scheme is presented to facilitate the simulation of micropipette aspiration, a nonlinear contact problem. In contrast to the halfspace model (Sato et al., 1990), this computational model accounts for nonlinearities in the cell response that result from a consideration of geometric factors including the finite cell dimension (radius R), curvature of the cell boundary, evolution of the cell-micropipette contact region, and curvature of the edges of the micropipette (inner radius a, edge curvature radius epsilon). Using 60 quadratic boundary elements, a micropipette aspiration creep test with ramp time t* = 0.1 s and ramp pressure p*/E = 0.8 is simulated for the cases a/R = 0.3, 0.4, 0.5 using mean parameter values for primary chondrocytes. Comparisons to the half-space model indicate that the computational model predicts an aspiration length that is less stiff during the initial ramp response (t = 0-1 s) but more stiff at equilibrium (t = 200 s). Overall, the ramp and equilibrium predictions of aspiration length by the computational model are fairly insensitive to aspect ratio a/R but can differ from the half-space model by up to 20 percent. This computational approach may be readily extended to account for more complex geometries or inhomogeneities in cellular properties.

A mathematical approximation for the solution of a static indentation test

The classical contact problem of the indentation of a thin compressible linear elastic layer bonded to a rigid substrate is considered. Closed-form mathematical approximations of the deformation are presented for the cases of plane indentation by a rectangular block and three dimensional indentation by a plane-ended (axisymmetric) cylinder. The approximations are analyzed in the context of a static indentation test by comparison of applied load values to those obtained using a classical integral transform solution. In the case of plane indentation, the mathematical and classical predictions agree to within 2% relative error for aspect ratios between 0.1 and 1.0 and apparent Poisson ratio between 0.0 and 0.3. Comparisons for the axisymmetric case indicate a similar pattern. The main advantage of the new approach is that it yields closed-form approximations of the static indentation solution which can also capture the essential singular behavior.

The role of intraoperative ultrasonography in planning the resection of hepatic neoplasms

OBJECTIVE

Both intraoperative ultrasonography (US) and computed tomography with arterial portography (CTAP) may be used in planning segment-oriented liver resection in patients with primary or secondary hepatic neoplasms. This study was conducted to determine if intraoperative US findings substantially alter surgical plans based on CTAP findings alone.

PATIENTS AND METHODS

Patients with potentially resectable hepatic lesions were considered for the study; judgement as to whether a lesion was resectable was based on a clinical evaluation and CTAP. Over the period Apr. 4, 1991, to Oct. 1, 1993, 24 consecutive patients with hepatic lesions were examined; of these, 22 (13 men and 9 women with a mean age of 60.1 years) underwent US during resection, 1 did not undergo surgery and 1 was found at surgery to have carcinomatosis. The true nature of the lesions was verified pathologically or by follow-up imaging and assays for carcinoembryonic antigen.

RESULTS

The surgical plan based on CTAP findings alone was altered by the intraoperative US findings in 9 (41%) of the 22 patients. A total of 60 intrahepatic abnormalities were evaluated: 49 malignant lesions and 11 artifacts. The specificity (100%) and negative predictive value (73.3%) for intraoperative US were significantly greater than for CTAP (specificity of 9.1% and negative predictive value of 14.1%).

CONCLUSION

In a substantial proportion of patients undergoing hepatic resection, intraoperative US alters the surgical plan based on CTAP and provides additional specificity in the evaluation of liver lesions. This method of imaging is therefore justified for patients undergoing liver resection.