Novel AI Combining CNN and SVM to Predict Colorectal Cancer Prognosis and Mutational Signatures from HE Images.

Reducing recurrence following radical resection of colon cancer without over- or under-treatment remains a challenge. Postoperative adjuvant chemotherapy (Adj) is currently administered based solely on pathological tumor, node, and metastasis (TNM) stage. However, prognosis can vary significantly among patients with the same disease stage. Therefore, novel classification systems in addition to the TNM are necessary to inform decision-making regarding postoperative treatment strategies, especially stage II and III disease, and to minimize overtreatment and undertreatment with Adj. We developed a prognostic prediction system for colorectal cancer by using a combined convolutional neural network (CNN) and support vector machine (SVM) approach to extract features from hematoxyling and eosin staining (HE) images. We combined the TNM and our AI-based classification system into a TNM-AI (mTNM-AI) classification system with high discriminative power for recurrence-free survival (RFS). Furthermore, the cancer cell population recognized by this system as low risk of recurrence exhibited the mutational signature SBS87 as a genetic phenotype. The novel AI-based classification system developed here is expected to play an important role in prognostic prediction and personalized treatment selection in oncology.

Preoperative evaluation of visceral pleural invasion in peripheral lung cancer utilizing deep learning technology.

PURPOSE: This study aimed to assess the efficiency of artificial intelligence (AI) in the detection of visceral pleural invasion (VPI) of lung cancer using high-resolution computed tomography (HRCT) images, which is challenging for experts because of its significance in T-classification and lymph node metastasis prediction. METHODS: This retrospective analysis was conducted on preoperative HRCT images of 472 patients with stage I non-small cell lung cancer (NSCLC), focusing on lesions adjacent to the pleura to predict VPI. YOLOv4.0 was utilized for tumor localization, and EfficientNetv2 was applied for VPI prediction with HRCT images meticulously annotated for AI model training and validation. RESULTS: Of the 472 lung cancer cases (500 CT images) studied, the AI algorithm successfully identified tumors, with YOLOv4.0 accurately localizing tumors in 98% of the test images. In the EfficientNet v2-M analysis, the receiver operating characteristic curve exhibited an area under the curve of 0.78. It demonstrated powerful diagnostic performance with a sensitivity, specificity, and precision of 76.4% in VPI prediction. CONCLUSION: AI is a promising tool for improving the diagnostic accuracy of VPI for NSCLC. Furthermore, incorporating AI into the diagnostic workflow is advocated because of its potential to improve the accuracy of preoperative diagnosis and patient outcomes in NSCLC.

Development of a 3D Vascular Network Visualization Platform for One-Dimensional Hemodynamic Simulation.

Recent advancements in computational performance and medical simulation technology have made significant strides, particularly in predictive diagnosis. This study focuses on the blood flow simulation reduced-order models, which provide swift and cost-effective solutions for complex vascular systems, positioning them as practical alternatives to 3D simulations in resource-limited medical settings. The paper introduces a visualization platform for patient-specific and image-based 1D-0D simulations. This platform covers the entire workflow, from modeling to dynamic 3D visualization of simulation results. Two case studies on, respectively, carotid stenosis and arterial remodeling demonstrate its utility in blood flow simulation applications.

Development of multiple AI pipelines that predict neoadjuvant chemotherapy response of breast cancer using H&E-stained tissues.

In recent years, the treatment of breast cancer has advanced dramatically and neoadjuvant chemotherapy (NAC) has become a common treatment method, especially for locally advanced breast cancer. However, other than the subtype of breast cancer, no clear factor indicating sensitivity to NAC has been identified. In this study, we attempted to use artificial intelligence (AI) to predict the effect of preoperative chemotherapy from hematoxylin and eosin images of pathological tissue obtained from needle biopsies prior to chemotherapy. Application of AI to pathological images typically uses a single machine-learning model such as support vector machines (SVMs) or deep convolutional neural networks (CNNs). However, cancer tissues are extremely diverse and learning with a realistic number of cases limits the prediction accuracy of a single model. In this study, we propose a novel pipeline system that uses three independent models each focusing on different characteristics of cancer atypia. Our system uses a CNN model to learn structural atypia from image patches and SVM and random forest models to learn nuclear atypia from fine-grained nuclear features extracted by image analysis methods. It was able to predict the NAC response with 95.15% accuracy on a test set of 103 unseen cases. We believe that this AI pipeline system will contribute to the adoption of personalized medicine in NAC therapy for breast cancer.

Mesoscopic-scale grain formation in HfO2-based ferroelectric thin films and its impact on electrical characteristics.

Ferroelectric memory devices are expected for low-power and high-speed memory applications. HfO2-based ferroelectric is attracting attention for its CMOS-compatibility and high scalability. Mesoscopic-scale grains, of which size is almost comparable to device size, are formed in HfO2-based ferroelectric poly-crystalline thin films, which largely influences electrical characteristics in memory devices. It is important to study the impact of mesoscopic-scale grain formation on the electrical characteristics. In this work, first, we have studied the thickness dependence of the polarization switching kinetics in HfO2-based ferroelectric. While static low-frequency polarization is comparable for different thickness, dynamic polarization switching speed is slower in thin Hf0.5Zr0.5O2 (HZO) capacitors. Based on the analysis using the NLS model and physical characterization, thinner HZO contains smaller grains with orientation non-uniformity and more grain boundaries than thicker HZO, which can impede macroscopic polarization switching. We have also theoretically and experimentally studied the polar-axis alignment of a HfO2-based ferroelectric thin film. While in-plane polar orientation is stable in as-grown HZO, out-of-plane polarization can be dominant by applying electric field, which indicates the transition from in-plane polar to out-of-plane polar orientation in the ferroelectric phase grains. This is confirmed by calculating kinetic pathway using ab-initio calculation.

Prediction of early recurrence of hepatocellular carcinoma after resection using digital pathology images assessed by machine learning.

Hepatocellular carcinoma (HCC) is a representative primary liver cancer caused by long-term and repetitive liver injury. Surgical resection is generally selected as the radical cure treatment. Because the early recurrence of HCC after resection is associated with low overall survival, the prediction of recurrence after resection is clinically important. However, the pathological characteristics of the early recurrence of HCC have not yet been elucidated. We attempted to predict the early recurrence of HCC after resection based on digital pathologic images of hematoxylin and eosin-stained specimens and machine learning applying a support vector machine (SVM). The 158 HCC patients meeting the Milan criteria who underwent surgical resection were included in this study. The patients were categorized into three groups: Group I, patients with HCC recurrence within 1 year after resection (16 for training and 23 for test); Group II, patients with HCC recurrence between 1 and 2 years after resection (22 and 28); and Group III, patients with no HCC recurrence within 4 years after resection (31 and 38). The SVM-based prediction method separated the three groups with 89.9% (80/89) accuracy. Prediction of Groups I was consistent for all cases, while Group II was predicted to be Group III in one case, and Group III was predicted to be Group II in 8 cases. The use of digital pathology and machine learning could be used for highly accurate prediction of HCC recurrence after surgical resection, especially that for early recurrence. Currently, in most cases after HCC resection, regular blood tests and diagnostic imaging are used for follow-up observation; however, the use of digital pathology coupled with machine learning offers potential as a method for objective postoprative follow-up observation.

A penalized spline fitting method to optimize geometric parameters of arterial centerlines extracted from medical images.

In order to grasp the spatial and temporal evolution of vascular geometry, three-dimensional (3D) arterial bending structure and geometrical changes of arteries and stent grafts (SG) must be quantified using geometrical parameters such as curvature and torsion along the vasculature centerlines extracted from medical images. Here, we develop a robust method for constructing smooth centerlines based on a spline fitting method (SFM) such that the optimized geometric parameters of curvature and torsion can be obtained independently of digitization noise in the images. Conventional SFM consists of the 3rd degree spline basis function and 2nd derivative penalty term. In contrast, the present SFM uses the 5th degree spline basis function and 3rd and 4th derivative penalty terms, the coefficients of which are derived by the Akaike information criterion. The results show that the developed SFM can reduce the errors of curvature and torsion compared to conventional SFM. We then apply the present SFM to the centerline of the SG in an abdominal aortic aneurysm (AAA), and those of bilateral internal carotid arteries (ICA) in 6 cases: 3 cases with aneurysms and 3 cases without any aneurysm. The SG centerlines were obtained from temporal medical images at three scan times. The strong peak of the curvature could be clearly observed in the distal area of the SG, the inversion of the torsion at 0 months in the middle area of SG disappeared over time, and the torsions around the SG bifurcation at the three time periods were inverted. The curvature-torsion graphs along the ICA centerlines superimposing five aneurysmal positions were useful for investigating the relationship between arterial bending structure and aneurysmal positions. Both ICAs had curvature peak values higher than 0.4 within the ICA syphons. The ICA torsion graphs indicated that left and right ICA tended to be a right- and left-handed helix, respectively. In the left ICA syphon, the biggest aneurysm could be observed downstream of the salient torsion inversion. All aneurysms for 3 cases were positioned at the downstream of the inverted torsion.

Geometric analysis of ruptured and nonruptured abdominal aortic aneurysms.

OBJECTIVE: The objective of this study was to use parameters to determine the geometric differences between ruptured abdominal aortic aneurysms (AAAs) and nonruptured AAAs. METHODS: Computed tomography data of 38 ruptured AAAs and 215 electively repaired (nonruptured) AAAs were collected from multiple institutes. We compared the ruptured AAA group and nonruptured AAA group with 1:1 matching by using the Mahalanobis distance, which was calculated using the patient's age, sex, and AAA diameter. We selected the longitudinal AAA image in multiplanar reconstruction view, placed a hypothetical ellipse on the aneurysm's protruded curve, and placed a circle on the portion connecting the aneurysm and the aorta. We then measured the aspect ratio (the vertical diameter divided by the horizontal diameter) and fillet radius (the radius of arc). RESULTS: The aspect ratio was significantly lower in the ruptured group than in the nonruptured group (2.02 ± 0.53 vs 2.60 ± 1.02; P = .002), as was the fillet radius (0.28 ± 0.18 vs 0.81 ± 0.44; P < .001). Receiver operating characteristic analysis revealed that the area under the curve of the aspect ratio was 0.688, and the optimal cutoff point was 2.23, with sensitivity and specificity of 0.55 and 0.76, respectively. The area under the curve of the fillet radius was 0.933, and the optimal cutoff was 0.347, with sensitivity and specificity of 0.97 and 0.87, respectively. CONCLUSIONS: The geometric analysis performed in this study revealed that ruptured AAAs had a smaller fillet radius and smaller aspect ratio than nonruptured AAAs did.

Morphological analysis using geometric parameters for splenic aneurysms.

Background Considering the unique characteristics of splenic artery aneurysms, we hypothesized that hemodynamic forces could play an important role in splenic artery aneurysm formation and that splenic artery geometry should be correlated with aneurysm development. Methods Tortuosity of the splenic artery was evaluated three-dimensionally by calculating the curvature using software and the original modeling system. We selected 54 splenic artery aneurysm patients who had undergone thin-slice computed tomography imaging with contrast. We compared the splenic artery aneurysm group to non-vascular patients via propensity-score matching (35 patients in each group). The splenic artery length index, average curvature, and maximum curvature were analyzed. Results Splenic artery aneurysm patients tended to have a longer splenic artery and the curvature was more severe compared to the non-vascular control patients. The average curvature of splenic artery aneurysm patients was associated with the dilatation rate in female patients. Conclusion Females with a tortuous splenic artery may have an increased risk of aneurysm formation.

Morphological Changes and Device Migration After Stent Graft Insertion　- Clinical Application of an Image-Based Modeling System and Analysis With Geometric Parameters.

BACKGROUND: Previously, we developed an image-based modeling system (V-Modeler) to investigate geometric changes in stent grafts (SGs) following their implantation for abdominal aortic aneurysms (AAAs). The aims of the present study were to improve this system for clinical use, to chronologically analyze postoperative morphological changes in SGs, and to demonstrate scenarios of SG migration.Methods and Results:Contrast-enhanced computed tomography data from 36 patients who underwent endovascular aneurysm repair (EVAR) for AAAs were used, with 72 centerline paths, in total, analyzed for bilateral SG legs. The existing V-modeler system was modified by introducing a penalty term, optimizing the number of control points using Akaike's information criterion, and changing the degree of the function from 3 to 5. Geometric parameters were then analyzed immediately, as well as >1 year after EVAR. Eight migrations were found and although overall SG curvature and curvature at the distal (leg) site did not change, curvature at the proximal (trunk) site of SGs decreased over time. Subanalysis revealed that SGs with severe curvature showed the same trend, whereas distal curvature increased in the non-severe curvature group. In addition, proximal curvature decreased more in Excluder than Zenith devices. CONCLUSIONS: The present study demonstrates SG behavior after implantation with numerical values for SG length and curvature.

Development of a Numerical Method for Patient-Specific Cerebral Circulation Using 1D-0D Simulation of the Entire Cardiovascular System with SPECT Data.

The detailed flow information in the circle of Willis (CoW) can facilitate a better understanding of disease progression, and provide useful references for disease treatment. We have been developing a one-dimensional-zero-dimensional (1D-0D) simulation method for the entire cardiovascular system to obtain hemodynamics information in the CoW. This paper presents a new method for applying 1D-0D simulation to an individual patient using patient-specific data. The key issue is how to adjust the deviation of physiological parameters, such as peripheral resistance, from literature data when patient-specific geometry is used. In order to overcome this problem, we utilized flow information from single photon emission computed tomography (SPECT) data. A numerical method was developed to optimize physiological parameters by adjusting peripheral cerebral resistance to minimize the difference between the resulting flow rate and the SPECT data in the efferent arteries of the CoW. The method was applied to three cases using different sets of patient-specific data in order to investigate the hemodynamics of the CoW. The resulting flow rates in the afferent arteries were compared to those of the phase-contrast magnetic resonance angiography (PC-MRA) data. Utilization of the SPECT data combined with the PC-MRA data showed a good agreement in flow rates in the afferent arteries of the CoW with those of PC-MRA data for all three cases. The results also demonstrated that application of SPECT data alone could provide the information on the ratios of flow distributions among arteries in the CoW.

Development of an Image-Based Modeling System to Investigate Evolutional Geometric Changes of a Stent Graft in an Abdominal Aortic Aneurysm.

BACKGROUND: Quantification of geometric changes of the stent graft (SG) in abdominal aortic aneurysm has been required for follow up of endovascular aneurysm repair (EVAR). The aim was to develop an image-based modeling system (V-Modeler) to investigate these changes over time. METHODS AND RESULTS: V-Modeler was applied to investigate the migration of the SG. Three sets of computed tomography images were taken at 3 different times: (1) 5 days after the implantation; (2) 7 months later when the unilateral leg migrated upward; and (3) 10 months later when the limb had migrated into the common iliac aneurysm resulting in a type 1b endoleak. A spline function was used to represent the center lines of the SG to track its evolutional geometric changes in a three-dimensional manner. The characteristics of vascular geometry, as well as the SG geometry using geometric parameters such as length, curvature, torsion, angle of tangent vector (ATV), and migrated length, was evaluated. It was observed that the strong peak of the curvature in the distal area appeared, and a conversion of the torsion disappeared chronologically. CONCLUSIONS: The V-Modeler was developed, which not only can extract vascular geometry but also can identify geometric parameter, such as curvature, torsion, and ATV, to predict adverse events following EVAR.

Quantitative measurement of blood flow volume in the major intracranial arteries by using 123i-iodoamphetamine SPECT.

PURPOSE: The aim of this study was to establish the novel automatic method to quantify blood flow volumes of the major intracranial arteries by using SPECT. METHODS: We created the vascular templates to cover the territory supplied by the major intracranial arteries. Each blood flow volume was calculated as the regional cerebral blood flow on SPECT using this template × volume size of the template. In this study, we evaluated the volume flows in 22 cerebral hemispheres with normal perfusion and 28 hemispheres with severe stenosis in the internal cerebral artery (ICA) or middle cerebral artery (MCA) and that at acetazolamide test in 16 normal hemispheres and 20 hemispheres with stenosis. RESULTS: The mean blood flow volumes of the ICA and MCA in the normal hemispheres increased to more than 40% after acetazolamide test (161-228 mL/min for ICA and 111-157 mL/min for MCA), although those in the hemispheres with stenosis increased to less than 35% (158-192 mL/min for ICA and 107-127 mL/min for MCA). The receiver operating characteristic analyses revealed that the simple difference between the blood flow volume at acetazolamide test and that at rest using the new MCA template was superior to detecting reduction of cerebrovascular reactivity (CVR), compared with the conventional percent CVR using the original template. CONCLUSIONS: Blood flow volumes of the intracranial arteries had been able to be quantified automatically on SPECT, and difference of CVR was available for predicting the blood demand-supply balance.

Cell-type specificity of the expression of Os BOR1, a rice efflux boron transporter gene, is regulated in response to boron availability for efficient boron uptake and xylem loading.

We describe a boron (B) transporter, Os BOR1, in rice (Oryza sativa). Os BOR1 is a plasma membrane-localized efflux transporter of B and is required for normal growth of rice plants under conditions of limited B supply (referred to as -B). Disruption of Os BOR1 reduced B uptake and xylem loading of B. The accumulation of Os BOR1 transcripts was higher in roots than that in shoots and was not affected by B deprivation; however, Os BOR1 was detected in the roots of wild-type plants under -B conditions, but not under normal conditions, suggesting regulation of protein accumulation in response to B nutrition. Interestingly, tissue specificity of Os BOR1 expression is affected by B treatment. Transgenic rice plants containing an Os BOR1 promoter-beta-glucuronidase (GUS) fusion construct grown with a normal B supply showed the strongest GUS activity in the steles, whereas after 3 d of -B treatment, GUS activity was elevated in the exodermis. After 6 d of -B treatment, GUS activity was again strong in the stele. Our results demonstrate that Os BOR1 is required both for efficient B uptake and for xylem loading of B. Possible roles of the temporal changes in tissue-specific patterns of Os BOR1 expression in response to B condition are discussed.

Saccharomyces cerevisiae Bor1p is a boron exporter and a key determinant of boron tolerance.

Boron is toxic to living organisms when present in excess. Saccharomyces cerevisiae Bor1p is a plasma membrane protein that decreases the intracellular concentration of boron and confers boron tolerance in yeasts. We investigated the detailed characteristics of boron transport by Bor1p and its roles in boron tolerance. Boron transport assays showed that the bor1 deletion mutant (bor1Delta) accumulates higher intracellular concentrations of boron and has a lower rate of boron export. The bor1Delta showed greater susceptibility to high concentrations of boron than the wild-type strain, and the growth rates of both strains were negatively correlated with the intracellular concentrations of boron. With normal to toxic levels of external boron, green fluorescent protein (GFP)-tagged Bor1p localized to the plasma membrane irrespective of the concentration of boron in the medium. Taken together, these results establish Bor1p as a plasma membrane boron exporter and a key determinant of boron tolerance.

Roles of BOR1, DUR3, and FPS1 in boron transport and tolerance in Saccharomyces cerevisiae.

The roles of three membrane proteins, BOR1, DUR3, and FPS1, in boron (B) transport in yeast were examined. The boron concentration in yeast cells lacking BOR1 was elevated upon exposure to 90 mM boric acid, whereas cells lacking DUR3 or FPS1 showed lower boron concentrations. Compared with control cells, cells overexpressing BOR1 or FPS1 had a lower boron concentration, and cells overexpressing DUR3 had a higher boron concentration. These results suggest that, in addition to the efflux boron transporter BOR1, DUR3 and FPS1 play important roles in regulating the cellular boron concentration. Analysis of the yeast transformants for tolerance to a high boric acid concentration revealed an apparent negative correlation between the protoplasmic boron concentration and the degree of tolerance to a high external boron concentration. Thus, BOR1, DUR3, and FPS1 appear to be involved in tolerance to boric acid and the maintenance of the protoplasmic boron concentration.

Isolation of Arabidopsis thaliana cDNAs that confer yeast boric acid tolerance.

An Arabidopsis thaliana cDNA library was introduced into a Saccharomyces cerevisiae mutant that lacks ScBOR1 (YNL275W), a boron (B) efflux transporter. Five cDNAs were identified that confer tolerance to high boric acid. The nucleotide sequence analysis identified the clones as a polyadenylate-binding protein, AtPAB2; a ribosomal small subunit protein, AtRPS20B; an RNA-binding protein, AtRBP47c'; and two Myb transcription factors, AtMYB13 and AtMYB68. The expression of these five genes also conferred boric acid tolerance on wild-type yeast. Two yeast genes, ScRPS20 and ScHRB1, that are similar to the isolated clones, were necessary for this boric acid tolerance. The possible roles of these A. thaliana and S. cerevisiae genes in boric acid tolerance are discussed.

Arabidopsis boron transporter for xylem loading.

Boron deficiency hampers the productivity of 132 crops in more than 80 countries. Boron is essential in higher plants primarily for maintaining the integrity of cell walls and is also beneficial and might be essential in animals and in yeast. Understanding the molecular mechanism(s) of boron transport is crucial for alleviating boron deficiency. Here we describe the molecular identification of boron transporters in biological systems. The Arabidopsis thaliana mutant bor1-1 is sensitive to boron deficiency. Uptake studies indicated that xylem loading is the key step for boron accumulation in shoots with a low external boron supply and that the bor1-1 mutant is defective in this process. Positional cloning identified BOR1 as a membrane protein with homology to bicarbonate transporters in animals. Moreover, a fusion protein of BOR1 and green fluorescent protein (GFP) localized to the plasma membrane in transformed cells. The promoter of BOR1 drove GFP expression in root pericycle cells. When expressed in yeast, BOR1 decreased boron concentrations in cells. We show here that BOR1 is an efflux-type boron transporter for xylem loading and is essential for protecting shoots from boron deficiency.