Semi-automatic Method Research Articles

Comparative evaluation of the accuracy of 3D TMJ analysis performed by different methods of processing computed tomograms

The aim of this study. Comparison of the accuracy of segmentation of TMJ elements in different ways and assessment of the suitability of the data obtained for the diagnosis of TMJ dysfunction. To study the segmentation of the bone elements of the TMJ (articular fossa, head of the LF), 60 computed tomograms of the maxillofacial region of patients were randomly selected in various ways (archival material). In group 1, the results of CT processing by AI diagnostics algorithms (Russia) were collected; in group 2, the results of CT processing based on the semi-automatic segmentation method in the Avantis3D program. The results of CT processing by Avantis3D AI algorithms (Russia) with different probability modes - 0.4 and 0.9, respectively, were selected for the third and fourth groups. Visually, the coincidence of the contours of the LF heads and articular pits isolated using different methods with their contours on all possible sections of the original CT itself was evaluated. The time spent on TMJ segmentation according to CT data was determined and compared using the methods described above. Of the 240 objects, only 7.5% of the cases showed a slight discrepancy between the contours of the original CT in group b1, which was the lowest of all. A slight discrepancy in the TMJ contours to be corrected is characteristic of the semi-automatic method of segmentation by optical density was detected in 50.4% (group 2). The largest percentage of significant errors not subject to correction was noted in the first group, which made it impossible to perform a full 3D analysis of the TMJ, and the smallest in the second and fourth. The magnitude of the error in determining the width of the articular gap in different groups is comparable to the size of one voxel per CT. When segmentation is carried out using AI, the difference between segmented objects is close to zero values. The average time spent on TMJ segmentation in group 1 was 10.2±1.23 seconds, in group 2 - 12.6±1.87 seconds, in groups 3 and 4 - 0.46±0.12 seconds and 0.46±0.13 seconds, respectively. The developed automated method for segmenting TMJ elements using AI is obviously more suitable for practical work, since it requires minimal time, and is almost as accurate as other methods under consideration.

Point of Care Ultrasound Bladder Volume Calculation on the Acute Medical Unit.

Nursing staff, healthcare assistants (HCAs) and other healthcare professionals on the Acute Medical Unit (AMU) at Royal Berkshire Hospital (RBH) were taught a Point of Care Ultrasound (POCUS) skill during a twenty minute session. Practitioners learned how to take bladder volume measurements with the Butterfly iQ, a portable ultrasound device which provides a visually-aided method of volume measurement. A Likert scale was used to measure the confidence that staff had in performing volume measurements with the AMU automated scanners, and with the semi-automated Butterfly iQ. After the teaching session, confidence reported by practitioners in using the semi-automated visual method was significantly higher than confidence reported in using the automated non-visual scanners (t < 0.001). Minimal time and expense was required to teach practitioners how to perform this skill. Training nurses in POCUS for bladder visualisation and bladder volume calculation is easy and practicable.

SA-ODG platform: a semi-automated and PCR-free method to analyse microRNAs in solid tissues.

Over the past two decades, numerous techniques have been developed for analysing microRNAs in body fluids and tissues. However, these techniques still face technical challenges, particularly when compared to well-established techniques for proteins and metabolites. Recently, the ODG platform was introduced, which is an innovative technology that allows for the direct detection and quantification of microRNAs in liquid biopsies without requiring extraction or amplification. This study presents the implementation of the ODG platform within a semi-automated protocol to create the "SA-ODG" platform, enhancing the efficiency and precision of microRNA testing while reducing hands-on time required by laboratory staff. For the first time, the SA-ODG platform has been used to directly quantify microRNAs in solid tissues. The results demonstrate precise analysis of miR-122-5p in mouse liver tissues using SA-ODG. These developments represent a crucial step forward in advancing the field of extraction and amplification-free microRNA detection and quantification.

A Semi-Automated Molecularly Imprinted Polymer Solid-Phase Extraction Method for Catechin Analysis from Jaboticaba (Plinia sp.) Peel Extract Samples by UHPLC-DAD

Jaboticaba (Plinia sp.), a nutritionally significant Brazilian fruit, is rich in phenolic compounds, such as catechin, known for its biological and pharmaceutical properties. Extracting and studying these compounds is an economic and environmental strategy to fully explore the great potential that these chemical compounds have. Therefore, this study focuses on developing a semi-automated solid-phase extraction method using molecularly imprinted polymer for catechin analysis in jaboticaba peel extracts, coupled with an ultra-high-performance liquid chromatography with a diode array detector. The reported polymer demonstrated superior extractor capacity compared to the non-imprinted polymer. Different extraction parameters were optimized, and the method exhibited a linear range of 10 to 100 µg mL-1 for catechin. The obtained precision with the coefficient of variation was below 7.3%, and the limits of quantification and detection were 12.4 and 4.1 µg mL-1 , respectively. The developed sorbent maintained analytical performance through approximately 40 injections. Results suggest that the reported method could efficiently extract catechin from jaboticaba peels in less than 10 min, providing a promising tool for the rapid investigation of natural products. This selective and sustainable approach demonstrated here contributes to the economic and environmental aspects of catechin extraction and analysis in jaboticaba fruit.

A semi-automatic design method of handwriting self-help devices for individuals with upper limb dysfunctions

A semi-automatic design method of handwriting self-help devices for individuals with upper limb dysfunctions

Annotation method of risk data in a certain field based on pattern matching

With the development of information technology and the increasing complexity of industrial technology, there is an urgent need for a certain field to use big data and artificial intelligence to improve the management and decision-making level. In order to classify the field’s risk text data through intelligent algorithms, analysing the risk distribution and the major problems, this paper researches on the annotation methods of training data in this field. The proposed data annotation method is based on pattern matching, addressing the special problems of risk data annotation in this field (such as strong professionalism, small data volume, high accuracy requirement and timeliness requirements). A new matching pattern is generated through the steps of text segmentation, keyword extraction, pattern preliminary generation, pattern relation tree construction, pattern optimization, pattern generalization, pattern verification, classification and annotation, and final classification and annotation are performed after pattern matching. Performance tests in terms of accuracy, recall rate, and annotation time have shown that the overall performance of the proposed method outperforms that of traditional item-by-item manual annotation, and semi-automatic annotation methods through machine learning. The method described in this paper has strong application value for risk data annotation in this field, and also has certain reference significance for high-density, high-accuracy and high-timeliness data annotation in other fields.

High-throughput Analysis of Capillary Density in Skeletal Muscle Cross Sections.

Capillary density in skeletal muscles is key to estimate exercise capacity in healthy individuals, athletes, and those with muscle-related pathologies. Here, we present a step-by-step, high-throughput semi-automated method for quantifying capillary density from whole human skeletal muscle cross-sections, in areas of the muscle occupied by myofibers. We provide a detailed protocol for immunofluorescence staining, image acquisition, processing, and quantification. Image processing is performed in ImageJ, and data analysis is conducted in R. The provided protocol allows high-throughput quantification of capillary density. Key features • This protocol builds upon the method and results described in Abbassi-Daloii et al. (2023b). • It includes step-by-step details on image acquisition and image processing of the entire muscle section. • It enables high-throughput and semi-automated image quantification of capillary density. • It provides a robust analysis for determining capillary density over the entire muscle cross section.

Rapid three-dimensional quantification of high-intensity plaques from coronary atherosclerosis T1-weighted characterization to predict periprocedural myocardial injury

BackgroundHigh-intensity plaque (HIP) on magnetic resonance imaging (MRI) has been documented as a powerful predictor of periprocedural myocardial injury (PMI) following percutaneous coronary intervention (PCI). Despite the recent proposal of three-dimensional HIP quantification to enhance the predictive capability, the conventional pulse sequence, which necessitates the separate acquisition of anatomical reference images, hinders accurate three-dimensional segmentation along the coronary vasculature. Coronary atherosclerosis T1-weighted characterization (CATCH) enables the simultaneous acquisition of inherently coregistered dark-blood plaque and bright-blood coronary artery images. We aimed to develop a novel HIP quantification approach using CATCH and to ascertain its superior predictive performance compared to the conventional two-dimensional assessment based on plaque-to-myocardium signal intensity ratio (PMR). MethodsIn this prospective study, CATCH MRI was conducted before elective stent implantation in 137 lesions from 125 patients. On CATCH images, dedicated software automatically generated tubular three-dimensional volumes of interest on the dark-blood plaque images along the coronary vasculature, based on the precisely matched bright-blood coronary artery images, and subsequently computed PMR and HIP volume (HIPvol). Specifically, HIPvol was calculated as the volume of voxels with signal intensity exceeding that of the myocardium, weighted by their respective signal intensities. PMI was defined as post-PCI cardiac troponin-T > 5 × the upper reference limit. ResultsThe entire analysis process was completed within 3 min per lesion. PMI occurred in 44 lesions. Based on the receiver operating characteristic curve analysis, HIPvol outperformed PMR for predicting PMI (C-statistics, 0.870 [95% CI, 0.805–0.936] vs. 0.787 [95% CI, 0.706–0.868]; p = 0.001). This result was primarily driven by the higher sensitivity HIPvol offered: 0.886 (95% CI, 0.754–0.962) vs. 0.750 for PMR (95% CI, 0.597–0.868; p = 0.034). Multivariable analysis identified HIPvol as an independent predictor of PMI (odds ratio, 1.15 per 10-μL increase; 95% CI, 1.01–1.30, p = 0.035). ConclusionsOur semi-automated method of analyzing coronary plaque using CATCH MRI provided rapid HIP quantification. Three-dimensional assessment using this approach had a better ability to predict PMI than conventional two-dimensional assessment.

IoT-enable smart agriculture using multiple sensors for sprinkle irrigation systems

The Internet of Things (IoT) has experienced considerable expansion and development in recent years, offering a multitude of prospects for technological progress in the agricultural sector, with the aim of improving efficiency and providing help to farmers on a global scale. This paper outlines the conceptualization and development of an smart sprinkle irrigation system that incorporates the Internet of Things (IoT) technology, leveraging the ESP32 platform. The aim of this study is to develop a smart irrigation-based internet of things utilizing multiple sensors for sprinkle irrigation system. This will be achieved by integrating several sensors to monitor and control the irrigation process. The facilitation of communication between the ESP32 device and the Blynk application simplifies the acquisition of irrigation data, initiation of plant watering, deactivation of the automated watering functionality, and generation of graphical representations derived from sensor measurements. The irrigation system that has been created encompasses three distinct ways, particularly manual, semi-automatic, and automatic. The manual irrigation method enables users to directly administer irrigation, but the semi-automatic method necessitates users to select the timing and length of watering. The automated approach employs irrigation systems that are capable of responding to soil moisture levels, triggering the activation of the pump when the moisture content decreases beyond a predetermined threshold. The device is equipped with a pH sensor that enables the monitoring of soil pH values, a critical factor in facilitating optimal plant growth and development. The system has the capability to present the rate at which water flows and the total amount of water utilized in a singular irrigation event, as well as to aggregate the water consumption across many irrigation instances. In general, the experiments demonstrate the satisfactory performance of the three irrigation systems, including the capability to manually initiate the pump.

Diastolic function assessment with four-dimensional flow cardiovascular magnetic resonance using automatic deep learning E/A ratio analysis

BackgroundDiastolic left ventricular dysfunction is a powerful contributor to the symptoms and prognosis of patients with heart failure. In patients with depressed left ventricular systolic function the EA ratio is the first step to defining the grade of diastolic dysfunction. Doppler echocardiography is the preferred imaging technique for diastolic function assessment, while CMR is less established as a method. Previous 4D Flow-based studies have looked at the EA ratio proximally to the mitral valve, requiring manual interaction. In this study we compare an automated, deep learning-based, and two semiautomated approaches for 4D Flow CMR-based EA ratio assessment to conventional, gold standard echo-based methods. MethodsNinety-seven subjects with chronic ischemic heart disease underwent a cardiac echo followed by an MRI investigation. 4D Flow-based EA ratio values were computed using three different approaches; two semi-automated, assessing the EA ratio by measuring the inflow velocity (MVvel) and the inflow volume (MVflow) at the mitral valve plane, and one fully automated, creating a full LV segmentation using a deep learning-based method within which the EA ratio could be assessed without constraint to the mitral plane (LVvel). ResultsMVvel, MVflow, and LVvel EA ratios were strongly associated with Echo EA ratio (R2= 0.60, 0.58, 0.72). LVvel peak E and A showed moderate association to Echo peak E and A, while MVvel values were weakly associated. MVvel and MVflow EA ratios were very strongly associated with LVvel (R2= 0.84, 0.86). MVvel peak E was moderately associated with LVvel, while peak A showed strong association (R2= 0.26, 0.57). Discussion and ConclusionPeak E, peak A and EA ratio are integral to the assessment of diastolic dysfunction and may expand the utility of CMR studies in patients with cardiovascular disease. While underestimation of absolute peak E and A velocities was noted, the EA ratio measured with all three 4D Flow methods was strongly associated with the gold standard Doppler echocardiography. The automatic, deep learning-based method performed best, with the most favorable runtime of ~40seconds. As both semiautomatic methods associated very strongly to LVvel, they could be employed as an alternative for estimation of EA ratio.

Prediction of 177Lu-DOTATATE Therapy Outcomes in Neuroendocrine Tumor Patients Using Semi-Automatic Tumor Delineation on 68Ga-DOTATATE PET/CT.

Treatment of metastatic neuroendocrine tumors (NET) with 177Lu-DOTATATE peptide receptor radionuclide therapy (PRRT) results in favorable response only in a subset of patients. We investigated the prognostic value of quantitative pre-treatment semi-automatic 68Ga-DOTATATE PET/CT analysis in NET patients treated with PRRT. The medical records of 94 NET patients who received at least one cycle of PRRT at a single institution were retrospectively reviewed. On each pre-treatment 68Ga-DOTATATE PET/CT, the total tumor volume (TTV), maximum tumor standardized uptake value for the patient (SUVmax), and average uptake in the lesion with the lowest radiotracer uptake (SUVmin) were determined with a semi-automatic tumor delineation method. Progression-free survival (PFS) and overall survival (OS) among the patients were compared based on optimal cutoff values for the imaging parameters. On Kaplan-Meier analysis and univariate Cox regression, significantly shorter PFS was observed in patients with lower SUVmax, lower SUVmin, and higher TTV. On multivariate Cox regression, lower SUVmin and higher TTV remained predictive of shorter PFS. Only higher TTV was found to be predictive of shorter OS on Kaplan-Meier and Cox regression analyses. In a post hoc Kaplan-Meier analysis, patients with at least one high-risk feature (low SUVmin or high TTV) showed shorter PFS and OS, which may be the most convenient parameter to measure in clinical practice. The tumor volume and lowest lesion uptake on 68Ga-DOTATATE PET/CT can predict disease progression following PRRT in NET patients, with the former also predictive of overall survival. NET patients at risk for poor outcomes following PRRT can be identified with semi-automated quantitative analysis of 68Ga-DOTATATE PET/CT.

Full virtual patient generated by artificial intelligence-driven integrated segmentation of craniomaxillofacial structures from CBCT images

ObjectivesTo assess the performance, time-efficiency, and consistency of a convolutional neural network (CNN) based automated approach for integrated segmentation of craniomaxillofacial structures compared with semi-automated method for creating a virtual patient using cone beam computed tomography (CBCT) scans. MethodsThirty CBCT scans were selected. Six craniomaxillofacial structures, encompassing the maxillofacial complex bones, maxillary sinus, dentition, mandible, mandibular canal, and pharyngeal airway space, were segmented on these scans using semi-automated and composite of previously validated CNN-based automated segmentation techniques for individual structures. A qualitative assessment of the automated segmentation revealed the need for minor refinements, which were manually corrected. These refined segmentations served as a reference for comparing semi-automated and automated integrated segmentations. ResultsThe majority of minor adjustments with the automated approach involved under-segmentation of sinus mucosal thickening and regions with reduced bone thickness within the maxillofacial complex. The automated and the semi-automated approaches required an average time of 1.1 min and 48.4 min, respectively. The automated method demonstrated a greater degree of similarity (99.6 %) to the reference than the semi-automated approach (88.3 %). The standard deviation values for all metrics with the automated approach were low, indicating a high consistency. ConclusionsThe CNN-driven integrated segmentation approach proved to be accurate, time-efficient, and consistent for creating a CBCT-derived virtual patient through simultaneous segmentation of craniomaxillofacial structures. Clinical relevanceThe creation of a virtual orofacial patient using an automated approach could potentially transform personalized digital workflows. This advancement could be particularly beneficial for treatment planning in a variety of dental and maxillofacial specialties.

A semi-automatic segmentation method for meningioma developed using a variational approach model.

Meningioma is the commonest primary brain tumour. Volumetric post-contrast magnetic resonance imaging (MRI) is recognised as gold standard for delineation of meningioma volume but is hindered by manual processing times. We aimed to investigate the utility of a model-based variational approach in segmenting meningioma. A database of patients with a meningioma (2007-2015) was queried for patients with a contrast-enhanced volumetric MRI, who had consented to a research tissue biobank. Manual segmentation by a neuroradiologist was performed and results were compared to the mathematical model, using a battery of tests including the Sørensen-Dice coefficient (DICE) and JACCARD index. A publicly available meningioma dataset (708 segmented T1 contrast-enhanced slices) was also used to test the reliability of the model. 49 meningioma cases were included. The most common meningioma location was convexity (n = 15, 30.6%). The mathematical model segmented all but one incidental meningioma, which failed due to the lack of contrast uptake. The median meningioma volume by manual segmentation was 19.0cm3 (IQR 4.9-31.2). The median meningioma volume using the mathematical model was 16.9cm3 (IQR 4.6-28.34). The mean DICE score was 0.90 (SD = 0.04). The mean JACCARD index was 0.82 (SD = 0.07). For the publicly available dataset, the mean DICE and JACCARD scores were 0.90 (SD = 0.06) and 0.82 (SD = 0.10), respectively. Segmentation of meningioma volume using the proposed mathematical model was possible with accurate results. Application of this model on contrast-enhanced volumetric imaging may help reduce work burden on neuroradiologists with the increasing number in meningioma diagnoses.

Development of an automated, high-throughput assay to detect angiotensin AT2-receptor agonistic compounds by nitric oxide measurements in vitro

Angiotensin AT2-receptor (AT2R) agonists have shown a wide range of protective effects in many preclinical disease models. However, the availability of AT2R-agonists is very limited due to the lack of high-throughput assays for AT2R-agonist identification. Therefore, we aimed to design and validate an assay for high-throughput screening of AT2R-agonist candidates. The assay is based on nitric oxide (NO) release measurements in primary human aortic endothelial cells (HAEC), in AT2R-transfected CHO cells (AT2R-CHO) or in non-transfected CHO cells (Flp-CHO) using the fluorescent probe DAF-FM diacetate. It is run in 96-well plates and fluorescence signals are semi-automatically quantified. The assay was tested for sensitivity (recognition of true positive results), selectivity (recognition of true negative results), and reliability (by calculating the repeatability coefficient (RC)). The high-throughput, semi-automated method was proven suitable, as the NO-releasing agents C21, CGP42112A, angiotensin-(1−7) and acetylcholine significantly increased NO release from HAEC. The assay is sensitive and selective, since the established AT2R-agonists C21, CGP42112A and angiotensin II significantly increased NO release from AT2R-CHO cells, while the non-AT2R-agonists angiotensin-(1−7) and acetylcholine had no effect. Assay reliability was shown by high-throughput screening of a library comprised of 40 potential AT2R-agonists, of which 39 met our requirements for reliability (RC ≤ 20% different from RC for C21). Our newly developed high-throughput method for detection of AT2R-agonistic activity was proven to be sensitive, selective, and reliable. This method is suitable for the screening of potential AT2R-agonists in future drug development programs.

Comparative Study of Two Semi-automated Forensic DNA Extraction Methods

Automation in forensic DNA analysis is crucial for analysts to reduce time, improve results, and decrease risk of contamination. With the variety of commercially available automated DNA extraction systems, comes the need for end-users to be informed of what they provide and what they might lack. Thus, this study aimed to evaluate the efficiency of two semi-automated DNA extraction systems used for forensic DNA analysis: Automate Express™ and Hamilton Microlab STAR™ system, for four parameters; reproducibility, stability, sensitivity and contamination. Overall, the results indicated that both semi-automated systems performed similarly in providing robust and reproducible DNA results while maintaining good capability to overcome PCR inhibition with low risk of contamination. The two semi-automated systems showed higher DNA recovery than organic extraction using phenol-chloroform by 22% for semen and 7% for blood samples. In addition, three sample types, blood, saliva, semen were tested to compare the two systems (total samples n=100). Overall, the data showed the average DNA recovery for Hamilton was higher than the DNA recovery by Automate Express™ for the blood and semen sample types indicating better performance of the Hamilton Microlab STAR™ in terms of recovery and sensitivity level.

Assessing the impact of gold mining on forest cover in the Surinamese Amazon from 1997 to 2019: A semi-automated satellite-based approach

The Amazon rainforest has faced extensive deforestation for decades due to urban growth, agricultural expansion, logging and mining. Mining however has been relatively understudied in the Amazon. The objectives of this study were: (1) to develop an effective cloud-based classification approach coupled with an innovative semi-automated reclassification method; (2) to quantify the mining extent and its impacts on forest cover in Suriname from 1997 to 2019; and (3) to evaluate the impact of mining on the structure and health of the forest. Landsat 5 and 8 images were used for the initial land-use/land-cover classification using the classification and regression trees algorithm in Google Earth Engine. The resulting classified maps were reclassified in a semi-automated model to correct misclassifications between mining and urban pixels. The final maps were used to analyse the expansion of mining and its impacts on forest cover, structure (using the fragmentation metric effective mesh size), and forest health (using the phenology metric peak greenness). The combined approach resulted in an improvement in mining detection accuracy from 72% (65% producer, 79% user) to 89.5% (84% producer accuracy, 95% user). The results showed that mining increased from 69.4 km2 in 1997 to 431.6 km2 in 2019, an increase by 522% over 22 years, which led to 421.3 km2 of forest loss, of which 85% was due to artisanal and small-scale mining (ASGM). The loss of forest for ASGM resulted in greater fragmentation, with a decrease in effective mesh size by 122.8 km2 compared to a decrease by 83 km2 caused by industrial mining. Mining also caused a decrease in the health of the surrounding forest, with a larger decrease in peak greenness due to ASGM compared to industrial mining. The results demonstrate the potential of this approach that leverages cloud-based machine learning with a semi-automated reclassification to allow for rapid, accurate, and potentially global mining detection.

COMPARISON OF STEREOLOGY METHODS FOR ASSESSING AGE-RELATED EFFECTS ON IMMUNOSTAINED BRAIN CELLS

Abstract The primary benefit of stereology methods is quantification of well-stained biological objects in tissue sections with the ability to adjust sampling intensity to achieve desired levels of precision. The advent of hand-crafted algorithms and artificial intelligence-based deep learning (DL) provides an opportunity for more standardized collection of stereology data with enhanced efficiency and higher reproducibility compared to state-of-the-art manual stereology. We contrasted and compared the performance of four manual, semi-automatic, and fully automatic approaches for generating data for total number of Neu-N immunostained neurons in neocortex (NCTX) in the mouse brain. The gold standard for these studies was manual counts using the state-of-the-art optical fractionator method on 3-D reconstructed serial z-axis image stacks through a known tissue volume (disector stacks). To allow for direct methodological comparisons on the same images, disector stacks were automatically converted into extended depth of field (EDF) images in which all neurons in the disector stack were imaged at each cell’s maximal plane of focus. Total number of Neu-N neurons on the same EDF images were counted by a fully automatic hand-crafted method [automatic segmentation algorithm (ASA)] and a semi-automatic method [ASA counts manually corrected for false positives and negatives]. All comparison counts were done using unbiased frames and counting rules with total counts of NeuN-immunostained neurons by the optical fractionator method. The results were comparable across methods with wide variations in throughput efficiency and inter-rater agreement. These results are discussed with respect to applications to experimental studies of brain aging, neuroinflammation and neurodegenerative disease.

Poracle: Testing Patches under Preservation Conditions to Combat the Overfitting Problem of Program Repair

To date, the users of test-driven program repair tools suffer from the overfitting problem; a generated patch may pass all available tests without being correct. In the existing work, users are treated as merely passive consumers of the tests. However, what if they are willing to modify the test to better assess the patches obtained from a repair tool? In this work, we propose a novel semi-automatic patch-classification methodology named Poracle . Our key contributions are three-fold. First, we design a novel lightweight specification method that reuses the existing test. Specifically, the users extend the existing failing test with a preservation condition —the condition under which the patched and pre-patched versions should produce the same output. Second, we develop a fuzzer that performs differential fuzzing with a test containing a preservation condition. Once we find an input that satisfies a specified preservation condition but produces different outputs between the patched and pre-patched versions, we classify the patch as incorrect with high confidence. We show that our approach is more effective than the four state-of-the-art patch classification approaches. Last, we show through a user study that the users find our semi-automatic patch assessment method more effective and preferable than the manual assessment.

Dataset Creation for Semantic Segmentation Using Colored Point Clouds Considering Shadows on Traversable Area

Semantic segmentation, which provides pixel-wise class labels for an input image, is expected to improve the movement performance of autonomous robots significantly. However, it is difficult to train a good classifier for target applications; public large-scale datasets are often unsuitable. Actually, a classifier trained using Cityscapes is not enough accurate for the Tsukuba Challenge. To generate an appropriate dataset for the target environment, we attempt to construct a semi-automatic method using a colored point cloud obtained with a 3D scanner. Although some degree of accuracy is achieved, it is not practical. Hence, we propose a novel method that creates images with shadows by rendering them in the 3D space to improve the classification accuracy of actual images with shadows, for which existing methods do not output appropriate results. Experimental results using datasets captured around the Tsukuba City Hall demonstrate that the proposed method was superior when appropriate constraints were applied for shadow generation; the mIoU was improved from 0.358 to 0.491 when testing images were obtained at different locations.

Local graph cut in the Image Segmenter app for breast ultrasound images segmentation

Breast cancer is among the most common cancers diagnosed in women globally. To help the breast cancer diagnosis, an important step is to accurately segment the breast lesion. To support clinicians in this important step, we analyze the performance of a semi-automated segmentation method based on the Local Graph Cut technique in the Image Segmenter application. Local graph cuts algorithm has the ability to segment more complicated shape by converting the image into a graph representation. It employs seed points set by the user and a cost function. The user identifies certain pixels as foreground and background. The region properties are identified from these pixels and they allow to specify the probability of a pixel belonging to the background or foreground. The graph cut formulation assigns each pixel to a node in the graph and incorrectly segmented pixels are re-assigned until the desired segmentation is completed. To evaluate the segmentation results, the Dice similarity coefficient and Fréchet distance were calculated between the ground truth images and the segmented images. Results show a Dice score of 0.7754 for malignant lesions and 0.8842 for benign lesions. The average Fréchet distance values were 303.28 for malignant and 290.80 for benign lesions, respectively. The experimental results show that the method achieves the best performance and gets the higher Dice score and Fréchet distance for breast benign lesions against malignant lesions.