AI-based Classification Research Articles

Considerations on Baseline Generation for Imaging AI Studies Illustrated on the CT-Based Prediction of Empyema and Outcome Assessment.

For AI-based classification tasks in computed tomography (CT), a reference standard for evaluating the clinical diagnostic accuracy of individual classes is essential. To enable the implementation of an AI tool in clinical practice, the raw data should be drawn from clinical routine data using state-of-the-art scanners, evaluated in a blinded manner and verified with a reference test. Three hundred and thirty-five consecutive CTs, performed between 1 January 2016 and 1 January 2021 with reported pleural effusion and pathology reports from thoracocentesis or biopsy within 7 days of the CT were retrospectively included. Two radiologists (4 and 10 PGY) blindly assessed the chest CTs for pleural CT features. If needed, consensus was achieved using an experienced radiologist’s opinion (29 PGY). In addition, diagnoses were extracted from written radiological reports. We analyzed these findings for a possible correlation with the following patient outcomes: mortality and median hospital stay. For AI prediction, we used an approach consisting of nnU-Net segmentation, PyRadiomics features and a random forest model. Specificity and sensitivity for CT-based detection of empyema (n = 81 of n = 335 patients) were 90.94 (95%-CI: 86.55–94.05) and 72.84 (95%-CI: 61.63–81.85%) in all effusions, with moderate to almost perfect interrater agreement for all pleural findings associated with empyema (Cohen’s kappa = 0.41–0.82). Highest accuracies were found for pleural enhancement or thickening with 87.02% and 81.49%, respectively. For empyema prediction, AI achieved a specificity and sensitivity of 74.41% (95% CI: 68.50–79.57) and 77.78% (95% CI: 66.91–85.96), respectively. Empyema was associated with a longer hospital stay (median = 20 versus 14 days), and findings consistent with pleural carcinomatosis impacted mortality.

Abstract PD11-01: An artificial intelligence-based predictor of CDH1 biallelic mutations and invasive lobular carcinoma

Abstract Introduction: Invasive lobular carcinoma (ILC) is the most frequent special histologic subtype of breast cancer (BC). ILC is identifiable by pathologic assessment given its distinctive discohesive growth pattern, largely caused by CDH1 inactivation. Compared to common forms of BC, ILCs display lower responses to chemotherapy and selective estrogen receptor modulators. The low interobserver agreement for the diagnosis of ILC, however, renders the inclusion of histologic subtyping in therapeutic decision-making challenging. Artificial intelligence (AI)-based algorithms hold promise for improving pathologic diagnosis; their performance, however, depends on the ground truth labeling used. Here, we seek to develop an AI-based methodology for detection of ILC using ‘CDH1 biallelic mutations’ (i.e., mutation + loss-of-heterozygosity of the wild-type allele or two pathogenic somatic mutations) as ground truth, reasoning that in BC, >95% of CDH1 bi-allelic inactivation is found in ILCs.Materials and methods: We developed a convolutional neural network system to detect CDH1 biallelic genetic inactivation (AI-CDH1) using whole slide images (WSI) of 1,100 primary BCs with available targeted sequencing data. The model was trained using a 10-fold cross-validation method to detect biallelic mutations. The mean number of positive and negative samples in the training set was 85.2 (SD=2.57) and 562.8 (SD=10.51) per fold, respectively. The evaluation set consisted of a mean of 14.2 (SD=2.04) positive and 93.8 (SD=9.13) negative samples. We evaluated the performance of the AI-CDH1 classifier to predict the lobular phenotype and CDH1 status using original and revised labels, following a histopathologic re-review of the histologic type and CDH1 status curation. The latter was conducted by incorporating information on biallelic CDH1 inactivation beyond CDH1 mutations (homozygous deletions, deleterious structural rearrangements, and loss-of-heterozygosity and gene promoter methylation).Results: The AI-CDH1 classifier predicted biallelic CDH1 mutations with an area under the curve (AUC)=0.944 (95 CI: 0.925-0.963), sensitivity=91.6% and specificity=85.9%, PPV=49.8%, NPV=98.5% and accuracy=86.7%, and the original ‘lobular phenotype’ with an AUC=0.941 (95 CI: 0.922-0.960), sensitivity=89%, specificity=86.7%, PPV=55.6%, NPV=97.7% and accuracy=87.1%. Review of the CDH1 gene status revealed that 7/957 BCs lacking CDH1 biallelic mutations harbored biallelic CDH1 inactivation by promoter methylation, homozygous deletions or structural rearrangements. The AI-CDH1 classifier detected all seven reclassified BCs and predicted the revised CDH1 biallelic inactivation with an AUC=0.948 (95 CI: 0.930-0.966), sensitivity=92%, specificity=86.5%, PPV=52.3%, NPV=98.5% and accuracy=87.2%. Upon histologic re-review, which resulted in reclassification of 36/927 non-lobular BCs as ‘lobular’ and 5/173 ‘lobular’ BCs as ‘non-lobular’, the AI-CDH1 classifier detected the ‘lobular phenotype’ with an AUC=0.953 (95 CI: 0.935-0.971), sensitivity=90.7%, specificity=89.7%, PPV=66.8%, NPV=97.7% and accuracy=89.9%. Using the revised histologic re-classification and CDH1 biallelic inactivation status labels, the AI-CDH1 classifier predicted the lobular phenotype irrespective of CDH1 status (P>0.05).Conclusions: By training a machine learning system to detect ‘CDH1 biallelic mutations’, as ground truth rather than histologic diagnosis of lobular carcinoma, which might be confounded by human subjectivity, we developed an AI-based system that can detect ILCs accurately, providing a new paradigm for the development of AI-based cancer classification systems. Citation Format: Jorge S Reis-Filho, Fresia Pareja, Fatemeh Derakhshan, David N Brown, Jillian Sue, Pier Selenica, Yi Kan Wang, Arnaud Da Cruz Paula, Monami Banerjee, Zahra Ebrahimzadeh, Manuel Isava, Matthew Lee, Ran Godrich, Adam Casson, Ruben Padron, George Shaikovski, Alexander van Eck, Antonio Marra, Higinio Dopeso, Hannah Y Wen, Edi Brogi, Matthew G Hanna, Chris Kanan, Jeremy D Kunz, Felipe C Geyer, Carla Leibowitz, David Klimstra, Leo Grady, Thomas J Fuchs. An artificial intelligence-based predictor of CDH1 biallelic mutations and invasive lobular carcinoma [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr PD11-01.

A hand-held, real-time, AI-assisted capillary convection PCR system for point-of-care diagnosis of African swine fever virus

A hand-held, low-cost and real-time convection polymerase chain reaction (CPCR) system is developed for point-of-care diagnosis of African swine fever (ASF) virus. In CPCR amplification, the reagent is transferred to denaturation, annealing and extension stages of the reaction repeatedly due to thermal convection. Two different modes of fiber-based optical detection are compared, and it shows that the detection sensitivity, repeatability, and consistency can be significantly improved with bottom collection comparing to sidewall collection. Each capillary tube is illuminated by its own LED from the top through a common optical filter, and meanwhile, a camera equipped with another optical filter is adopted to detect the fluorescence signal from optical fibers. To handle the untypical real-time fluorescence curves of negative tests due to non-specific amplification, AI-based classification methods, for example, artificial neural network (ANN) is adopted to improve the detection specificity up to 97.50% comparing to 47.5% or 75.0% which is achieved with Ct-based method. A smartphone is adopted to run the AI algorithm and custom software, as well as to report an ASF event with the geographical information. Experimental results show that with the hand-held CPCR system, within 30 min, ASF viruses can be successfully detected with acceptable sensitivity and specificity.

Ensemble Computational Intelligent for Insomnia Sleep Stage Detection via the Sleep ECG Signal

Insomnia is a common sleep disorder in which patients cannot sleep properly. Accurate detection of insomnia disorder is a crucial step for disease analysis in the early stages. The disruption in getting quality sleep is one of the big sources of cardiovascular syndromes such as blood pressure and stroke. The traditional insomnia detection methods are time-consuming, cumbersome, and more expensive because they demand a long time from a trained neurophysiologist, and they are prone to human error, hence, the accuracy of diagnosis gets compromised. Therefore, the automatic insomnia diagnosis from the electrocardiogram (ECG) records is vital for timely detection and cure. In this paper, a novel hybrid approach based on the power spectral density (PSD) of the heart rate variability (HRV) is proposed to detect insomnia in three classification scenarios: (1) subject-based classification scenario (normal Vs. insomnia), (2) sleep stage-based classification (REM Vs. W. stage), and (3) the combined classification scenario using both subject-based and sleep stage-based features. The ensemble learning of random forest (RF) and decision tree (DT) classifiers are used to perform the first and second classification scenarios, while the linear discriminant analysis (LDA) classifier is used to perform the third combined scenario. The proposed framework includes data collection, investigation of the ECG signals, extraction of the signal HRV, estimation of the PSD, and AI-based classification via hybrid machine learning classifiers. The proposed framework is fine-tuned and evaluated using the free public Physio Net dataset over fivefold trails cross-validation. For the subject-based classification scenario, the detection performance in terms of sensitivity, specificity, and accuracy is recorded to be 96.0%, 94.0%, and 96.0%, respectively. For the sleep stage-based classification scenario, the detection evaluation results are recorded equally with 96.0% for ceiling level accuracy, sensitivity, and specificity. For the combined classification scenario, the LDA classifier have achieved the best insomnia detection accuracy of 99.0% of the three cases as discussed. In future, the proposed approach could be applicable for mobile observation schemes to automatically detect insomnia disorder.

Cluster Analysis of Cell Nuclei in H&E-Stained Histological Sections of Prostate Cancer and Classification Based on Traditional and Modern Artificial Intelligence Techniques.

Biomarker identification is very important to differentiate the grade groups in the histopathological sections of prostate cancer (PCa). Assessing the cluster of cell nuclei is essential for pathological investigation. In this study, we present a computer-based method for cluster analyses of cell nuclei and performed traditional (i.e., unsupervised method) and modern (i.e., supervised method) artificial intelligence (AI) techniques for distinguishing the grade groups of PCa. Two datasets on PCa were collected to carry out this research. Histopathology samples were obtained from whole slides stained with hematoxylin and eosin (H&E). In this research, state-of-the-art approaches were proposed for color normalization, cell nuclei segmentation, feature selection, and classification. A traditional minimum spanning tree (MST) algorithm was employed to identify the clusters and better capture the proliferation and community structure of cell nuclei. K-medoids clustering and stacked ensemble machine learning (ML) approaches were used to perform traditional and modern AI-based classification. The binary and multiclass classification was derived to compare the model quality and results between the grades of PCa. Furthermore, a comparative analysis was carried out between traditional and modern AI techniques using different performance metrics (i.e., statistical parameters). Cluster features of the cell nuclei can be useful information for cancer grading. However, further validation of cluster analysis is required to accomplish astounding classification results.

An accurate artificial intelligence system for the detection of pulmonary and extra pulmonary Tuberculosis

Tuberculosis (TB) is the greatest irresistible illness in humans, caused by microbes Mycobacterium TB (MTB) bacteria and is an infectious disease that spreads from one individual to another through the air. It principally influences lung, which is termed Pulmonary TB (PTB). However, it can likewise influence other parts of the body such as the brain, bones and lymph nodes. Hence, it is also referred to as Extra Pulmonary TB (EPTB). TB has normal symptoms, so without proper testing, it is hard to detect if a patient has TB or not. In this paper, an accurate and novel system for diagnosing TB (PTB and EPTB) has been designed using image processing and AI-based classification techniques. The designed system is comprised of two phases. Firstly, the X-Ray image is processed using preprocessing, segmentation and features extraction and then, three different AI-based techniques are applied for classification. For image processing, ‘Histogram Filter’ and ‘Median Filter’ are applied with the CLAHE process to retrieve the segmented image. Then, classification based on AI techniques is done. The designed system produces the accuracy of 98%, 83%, and 89% for Decision Tree, SVM, and Naïve Bayes Classifier, respectively and has been validated by the doctors of the Jalandhar, India.

The utility of color normalization for AI-based diagnosis of hematoxylin and eosin-stained pathology images.

The color variation of hematoxylin and eosin (H&E)-stained tissues has presented a challenge for applications of artificial intelligence (AI) in digital pathology. Many color normalization algorithms have been developed in recent years in order to reduce the color variation between H&E images. However, previous efforts in benchmarking these algorithms have produced conflicting results and none have sufficiently assessed the efficacy of the various color normalization methods for improving diagnostic performance of AI systems. In this study, we systematically investigated eight color normalization algorithms for AI-based classification of H&E-stained histopathology slides, in the context of using images both from one center and from multiple centers. Our results show that color normalization does not consistently improve classification performance when both training and testing data are from a single center. However, using four multi-center datasets of two cancer types (ovarian and pleural) and objective functions, we show that color normalization can significantly improve the classification accuracy of images from external datasets (ovarian cancer: 0.25 AUC increase, p = 1.6e-05; pleural cancer: 0.21 AUC increase, p = 1.4e-10). Furthermore, we introduce a novel augmentation strategy by mixing color-normalized images using three easily accessible algorithms that consistently improves the diagnosis of test images from external centers, even when the individual normalization methods had varied results. We anticipate our study to be a starting point for reliable use of color normalization to improve AI-based, digital pathology-empowered diagnosis of cancers sourced from multiple centers. © 2021 The Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Recurrence and Quality-of-Life Following Involved Node Radiotherapy for Head and Neck Squamous Cell Carcinoma: Initial Results From the Phase II INRT-Air Trial

Elective neck irradiation (ENI) has long been considered mandatory when treating head and neck squamous cell carcinoma (HNSCC) with definitive radiotherapy (RT). However, ENI is often responsible for the majority of the dose to the salivary glands and swallowing organs-at-risk (OAR), and recent prospective studies have supported reduced elective neck volume and dose. Moreover, with sensitive staging studies and artificial intelligence (AI), our ability to identify occult lymphadenopathy is greatly improving. In this prospective phase II study, we investigated the efficacy and tolerability of eliminating ENI and strictly treating involved and suspicious lymph nodes (LN) with intensity modulated radiation therapy.Patients with newly-diagnosed squamous cell carcinoma of the oropharynx, larynx and hypopharynx were eligible for enrollment, excluding stage I-II glottic larynx. Each LN was characterized as involved or suspicious based on CT (e.g., cross-sectional size, enhancement pattern) and PET criteria and an in-house AI-based classification model. Gross disease and the primary clinical target volume were irradiated to 70 and 63 Gy in 35 fractions, respectively, and suspicious LN were treated to 66.5 Gy. Concurrent chemotherapy was required for stage T3N0-1 and IVA/B patients. The primary endpoint of the study was risk of solitary elective volume LN recurrence (i.e., LN recurrence in a classic ENI field without synchronous in-field or distant failure), with secondary endpoints including patterns-of-failure and patient reported outcomes.Sixty-seven patients were enrolled, with 18 larynx/hypopharynx and 49 oropharynx cancer (88% p16+). In total, 10% and 90% of patients were treated with accelerated RT and chemoRT, respectively. Average contralateral/ipsilateral mean parotid and contralateral submandibular doses were 11.9/18.4 Gy and 34.1 Gy, respectively, and average mean larynx/inferior constrictor/esophagus doses for oropharynx patients were 18.9/12.2/6.7 Gy. With a median follow-up of 12.8 months for surviving patients, there have been no solitary elective volume LN recurrences. Four patients have died from unrelated causes, and one died from in-field local progression. Gastrostomy was placed in 14 (21%), with median removal after 2.8 months for disease-free patients; only one patient is still dependent. Grade 1/2 dermatitis was seen in 90%/10% patients. There was no significant difference in MDADI score between baseline and months 3 and 6 post-RT; for patients with baseline MDADI score > 75, the scores at 3/6 months were 84/83.Preliminary results suggest that eliminating ENI is oncologically sound for HNSCC, with highly favorable OAR dosimetry and promising quality-of-life outcomes. Longer follow-up is needed to confirm these results.

Deep Learning Algorithms Correctly Classify Brassica rapa Varieties Using Digital Images.

Efficient and accurate methods of analysis are needed for the huge amount of biological data that have accumulated in various research fields, including genomics, phenomics, and genetics. Artificial intelligence (AI)-based analysis is one promising method to manipulate biological data. To this end, various algorithms have been developed and applied in fields such as disease diagnosis, species classification, and object prediction. In the field of phenomics, classification of accessions and variants is important for basic science and industrial applications. To construct AI-based classification models, three types of phenotypic image data were generated from 156 Brassica rapa core collections, and classification analyses were carried out using four different convolutional neural network architectures. The results of lateral view data showed higher accuracy compared with top view data. Furthermore, the relatively low accuracy of ResNet50 architecture suggested that definition and estimation of similarity index of phenotypic data were required before the selection of deep learning architectures.

AI-based detection and classification of distal radius fractures using low-effort data labeling: evaluation of applicability and effect of training set size

ObjectivesTo evaluate the performance of a deep convolutional neural network (DCNN) in detecting and classifying distal radius fractures, metal, and cast on radiographs using labels based on radiology reports. The secondary aim was to evaluate the effect of the training set size on the algorithm’s performance.MethodsA total of 15,775 frontal and lateral radiographs, corresponding radiology reports, and a ResNet18 DCNN were used. Fracture detection and classification models were developed per view and merged. Incrementally sized subsets served to evaluate effects of the training set size. Two musculoskeletal radiologists set the standard of reference on radiographs (test set A). A subset (B) was rated by three radiology residents. For a per-study-based comparison with the radiology residents, the results of the best models were merged. Statistics used were ROC and AUC, Youden’s J statistic (J), and Spearman’s correlation coefficient (ρ).ResultsThe models’ AUC/J on (A) for metal and cast were 0.99/0.98 and 1.0/1.0. The models’ and residents’ AUC/J on (B) were similar on fracture (0.98/0.91; 0.98/0.92) and multiple fragments (0.85/0.58; 0.91/0.70). Training set size and AUC correlated on metal (ρ = 0.740), cast (ρ = 0.722), fracture (frontal ρ = 0.947, lateral ρ = 0.946), multiple fragments (frontal ρ = 0.856), and fragment displacement (frontal ρ = 0.595).ConclusionsThe models trained on a DCNN with report-based labels to detect distal radius fractures on radiographs are suitable to aid as a secondary reading tool; models for fracture classification are not ready for clinical use. Bigger training sets lead to better models in all categories except joint affection.Key Points• Detection of metal and cast on radiographs is excellent using AI and labels extracted from radiology reports.• Automatic detection of distal radius fractures on radiographs is feasible and the performance approximates radiology residents.• Automatic classification of the type of distal radius fracture varies in accuracy and is inferior for joint involvement and fragment displacement.

AI-based localization and classification of skin disease with erythema

Although computer-aided diagnosis (CAD) is used to improve the quality of diagnosis in various medical fields such as mammography and colonography, it is not used in dermatology, where noninvasive screening tests are performed only with the naked eye, and avoidable inaccuracies may exist. This study shows that CAD may also be a viable option in dermatology by presenting a novel method to sequentially combine accurate segmentation and classification models. Given an image of the skin, we decompose the image to normalize and extract high-level features. Using a neural network-based segmentation model to create a segmented map of the image, we then cluster sections of abnormal skin and pass this information to a classification model. We classify each cluster into different common skin diseases using another neural network model. Our segmentation model achieves better performance compared to previous studies, and also achieves a near-perfect sensitivity score in unfavorable conditions. Our classification model is more accurate than a baseline model trained without segmentation, while also being able to classify multiple diseases within a single image. This improved performance may be sufficient to use CAD in the field of dermatology.

Hidden Variables in Deep Learning Digital Pathology and Their Potential to Cause Batch Effects: Prediction Model Study.

BackgroundAn increasing number of studies within digital pathology show the potential of artificial intelligence (AI) to diagnose cancer using histological whole slide images, which requires large and diverse data sets. While diversification may result in more generalizable AI-based systems, it can also introduce hidden variables. If neural networks are able to distinguish/learn hidden variables, these variables can introduce batch effects that compromise the accuracy of classification systems.ObjectiveThe objective of the study was to analyze the learnability of an exemplary selection of hidden variables (patient age, slide preparation date, slide origin, and scanner type) that are commonly found in whole slide image data sets in digital pathology and could create batch effects.MethodsWe trained four separate convolutional neural networks (CNNs) to learn four variables using a data set of digitized whole slide melanoma images from five different institutes. For robustness, each CNN training and evaluation run was repeated multiple times, and a variable was only considered learnable if the lower bound of the 95% confidence interval of its mean balanced accuracy was above 50.0%.ResultsA mean balanced accuracy above 50.0% was achieved for all four tasks, even when considering the lower bound of the 95% confidence interval. Performance between tasks showed wide variation, ranging from 56.1% (slide preparation date) to 100% (slide origin).ConclusionsBecause all of the analyzed hidden variables are learnable, they have the potential to create batch effects in dermatopathology data sets, which negatively affect AI-based classification systems. Practitioners should be aware of these and similar pitfalls when developing and evaluating such systems and address these and potentially other batch effect variables in their data sets through sufficient data set stratification.

A literature review on COVID-19 disease diagnosis from respiratory sound data

The World Health Organization (WHO) has announced a COVID-19 was a global pandemic in March 2020. It was initially started in china in the year 2019 December and affected an expanding number of nations in various countries in the last few months. In this particular situation, many techniques, methods, and AI-based classification algorithms are put in the spotlight in reacting to fight against it and reduce the rate of such a global health crisis. COVID-19's main signs are heavy temperature, different cough, cold, breathing shortness, and a combination of loss of sense of smell and chest tightness. The digital world is growing day by day, in this context digital stethoscope can read all of these symptoms and diagnose respiratory disease. In this study, we majorly focus on literature reviews of how SARS-CoV-2 is spreading and in-depth analysis of the diagnosis of COVID-19 disease from human respiratory sounds like cough, voice, and breath by analyzing the respiratory sound parameters. We hope this review will provide an initiative for the clinical scientists and researcher's community to initiate open access, scalable, and accessible work in the collective battle against COVID-19.

The Concept of AI-Based Algorithm: Analysis of CEUS Images and HSPs for Identification of Early Parenchymal Changes in Severe Acute Pancreatitis

(1) Background: Identifying early pancreas parenchymal changes remains a challenging radiologic diagnostic task. In this study, we hypothesized that applying artificial intelligence (AI) to contrast-enhanced ultrasound (CEUS) along with measurement of Heat Shock Protein (HSP)-70 levels could improve detection of early pancreatic necrosis in acute pancreatitis. (2) Methods: Acute pancreatitis $(n=146)$ and age- and sex matched healthy controls $(n=50)$ were enrolled in the study. The severity of acute pancreatitis was determined according to the revised Atlanta classification. The selected severe acute pancreatitis (AP) patient and an age/sex-matched healthy control were analysed for the algorithm initiation. Peripheral blood samples from the pancreatitis patient were collected on admission and HSP-70 levels were measured using ELISA. A CEUS device acquired multiple mechanical index contrast-specific mode images. Manual contour selection of the two-dimensional (2D) spatial region of interest (ROI) followed by calculations of the set of quantitative parameters. Image processing calculations and extraction of quantitative parameters from the CEUS diagnostic images were performed using algorithms implemented in the MATLAB software. (3) Results: Serum HSP-70 levels were 100.246 ng/ml (mean 76.4 ng/ml) at the time of the acute pancreatitis diagnosis. The CEUS Peek value was higher (155.5) and the mean transit time was longer (40.1 s) for healthy pancreas than in parenchyma affected by necrosis (46.5 and 34.6 s, respectively). (4) Conclusions: The extracted quantitative parameters and HSP-70 biochemical changes are suitable to be used further for AI-based classification of pancreas pathology cases and automatic estimation of pancreatic necrosis in AP.

Innovative microfossil (radiolarian) analysis using a system for automated image collection and AI-based classification of species

Microfossils are a powerful tool in earth sciences, and they have been widely used for the determination of geological age and in paleoenvironmental studies. However, the identification of fossil species requires considerable time and labor by experts with extensive knowledge and experience. In this study, we successfully automated the acquisition of microfossil data using an artificial intelligence system that employs a computer-controlled microscope and deep learning methods. The system was used to calculate changes in the relative abundance (%) of Cycladophora davisiana, a siliceous microfossil species (Radiolaria) that is widely used as a stratigraphic tool in studies on Pleistocene sediments in the Southern Ocean. The estimates obtained using this system were consistent with the results obtained by a human expert (< ± 3.2%). In terms of efficiency, the developed system was capable of performing the classification tasks approximately three times faster than a human expert performing the same task.

A Novel Approach to Blood Smear Analysis Based on Specimen Topology: Implications for Human and Artificial Intelligence Decision Making

Complete blood count (CBC) analysis is one of the most commonly ordered laboratory tests and is a critical first step in patients' clinical evaluation. However, CBC analyzers are limited in their ability to positively identify several types of white blood cells (WBC), and cells with substantial clinical significance, such as immature granulocytes or blasts, are merely marked as flags. Also, CBC analyzers fall short of recognizing informative red blood cell (RBC) morphology, such as schistocytes, and often provide inaccurate platelets count. Flags and clinically non-sufficient CBC-derived data reflex to generation of blood smear (BS), and BS review comprises a substantial portion of the workload in routine hematology laboratories. For accurate identification and classification of WBC, BS analysis (BSA) requires detailed observation of cells with high-magnification objective (60-100X), which provides a relatively narrow Field of View (FOV). This physical limitation restricts current BSA to either low resolution/wide FOV or to high resolution/narrow FOV data generation (Fig. 1A). Hence, key issues of BSA such as the effects of the smearing process on the distribution of blood components, the effects of cells distribution on their morphology and further classification, as well as many other attributes, are addressed only qualitatively or empirically, leaving the real topology of the BS obscure. The computational imaging microscopy system presented herein uses a low resolution and wide FOV objective, and records a plurality of images under different illumination conditions, of the same sample area (Fig. 1B). An algorithm reconstructs a high resolution and aberration free image of whole specimens, as can be observed in the attached link (https://tinyurl.com/Scopio-Labs-X100-ASH-2020). High resolution images are critical not only for manual BSA, but also for artificial intelligence (AI)-derived BSA, since data quality is of prime importance for deep-learning processes, and to a large extent determine their outcome. Thus, the combination of high resolution/wide FOV turns each BS into a big data analytic field, rendering the measurement of yet undetermined cell characteristics. In order to elucidate the basic topology, 60 normal BS (28 females, 32 males) were subjected to analysis utilizing this novel computational imaging microscopy. For convenience of analysis and comparison with current BSA methodology, BS were segmented into strips according to RBC density (Fig. 1C, D). The average length of smear from females (F) was higher by nearly 28% compared with smear from males (M), and the presence of acute inflammation (A) resulted in a significant 33% increase in overall smear length compared to normal (N) average (Fig. 1E). As expected, RBC density formed a linear gradient (Fig. 1C) along the axis of sample smearing, however, RBC morphology was affected by location within the BS. For example, strips 4-5 contained RBC with the appearance of spherocytes (Fig. 1F; arrows), while in strips with increased RBC density, cells aggregated resembling rouleaux formation (Fig. 1F; arrowheads). Platelets distribution was non-linear, with only a few of them reaching the feathered edge of the smear (Fig. 1G). Since the variance of both RBC/FOV and platelets/FOV concentrations drops starting with strip 4, BS-derived platelets number estimates should not be performed in strips 1-3. On average, a normal BS contains 890+399 WBC in the scanned area (strips 1-8). Similar to RBC, the location of individual WBC throughout the BS may affect their morphology, and hence their classification. WBC in the feathered edge (strips 1-3) are generally more stretched, and often squeezed between RBC, rendering their classification by AI-based tools challenging (Fig. 1H). In strips 4-7, WBC morphology is optimal for a classification task, enabling favorable outcomes for either manual or AI cell analysis (Fig. 1H). These data indicate that BSA can be taken to a sensitivity level of at least 10-3 of WBC analysis, provided that a large portion of the BS is scanned. Our system provides a novel combination of computational imaging microscopy and AI-based classification tools to unravel the complex topology of blood smears, and upgrade the data obtained in BSA. This approach enables the establishment of quantitative rules to scientifically direct the objective analysis of cellular blood components both manually, and by AI-tools. Figure Disclosures Katz: Scopio Labs: Consultancy.

Artificial Intelligence-Based Classification of Chest X-Ray Images into COVID-19 and Other Infectious Diseases.

The ongoing pandemic of coronavirus disease 2019 (COVID-19) has led to global health and healthcare crisis, apart from the tremendous socioeconomic effects. One of the significant challenges in this crisis is to identify and monitor the COVID-19 patients quickly and efficiently to facilitate timely decisions for their treatment, monitoring, and management. Research efforts are on to develop less time-consuming methods to replace or to supplement RT-PCR-based methods. The present study is aimed at creating efficient deep learning models, trained with chest X-ray images, for rapid screening of COVID-19 patients. We used publicly available PA chest X-ray images of adult COVID-19 patients for the development of Artificial Intelligence (AI)-based classification models for COVID-19 and other major infectious diseases. To increase the dataset size and develop generalized models, we performed 25 different types of augmentations on the original images. Furthermore, we utilized the transfer learning approach for the training and testing of the classification models. The combination of two best-performing models (each trained on 286 images, rotated through 120° or 140° angle) displayed the highest prediction accuracy for normal, COVID-19, non-COVID-19, pneumonia, and tuberculosis images. AI-based classification models trained through the transfer learning approach can efficiently classify the chest X-ray images representing studied diseases. Our method is more efficient than previously published methods. It is one step ahead towards the implementation of AI-based methods for classification problems in biomedical imaging related to COVID-19.

A Transfer Learning approach for AI-based classification of brain tumors

Classification of Brain Tumor (BT) is a vital assignment for assessing Tumors and making a suitable treatment. There exist numerous imaging modalities that are utilized to identify tumors in the brain. Magnetic Resonance Imaging (MRI) is generally utilized for such a task because of its unrivaled quality of the image and the reality that it does not depend on ionizing radiations. The relevance of Artificial Intelligence (AI) in the form of Deep Learning (DL) in the area of medical imaging has paved the path to extraordinary developments in categorizing and detecting intricate pathological conditions, like a brain tumor, etc. Deep learning has demonstrated an astounding presentation, particularly in segmenting and classifying brain tumors. In this work, the AI-based classification of BT using Deep Learning Algorithms are proposed for the classifying types of brain tumors utilizing openly accessible datasets. These datasets classify BTs into (malignant and benign). The datasets comprise 696 images on T1-weighted images for testing purposes. The projected arrangement accomplishes a noteworthy performance with the finest accuracy of 99.04%. The achieved outcome signifies the capacity of the proposed algorithm for the classification of brain tumors.

Using AI-Based Classification Techniques to Process EEG Data Collected during the Visual Short-Term Memory Assessment

Visual short-term memory (VSTM) is defined as the ability to remember a small amount of visual information, such as colors and shapes, during a short period of time. VSTM is a part of short-term memory, which can hold information up to 30 seconds. In this paper, we present the results of research where we classified the data gathered by using an electroencephalogram (EEG) during a VSTM experiment. The experiment was performed with 12 participants that were required to remember as many details as possible from the two images, displayed for 1 minute. The first assessment was done in an isolated environment, while the second assessment was done in front of the other participants, in order to increase the stress of the examinee. The classification of the EEG data was done by using four algorithms: Naive Bayes, support vector, KNN, and random forest. The results obtained show that AI-based classification could be successfully used in the proposed way, since we were able to correctly classify the order of the images presented 90.12% of the time and type of the displayed image 90.51% of the time.

A comparison of artificial intelligence-based classification techniques in predicting flow variables in sharp curved channels

Due to the complexity of variable distributions, it is essential to evaluate the flow patterns in sharp curved channels with different angles. It has recently become more common to use classification methods in combination with different artificial intelligence (AI) models to boost AI model performance. Gholami et al. (Appl Soft Comput 48:563–583, 2016a) obtained enhanced results by combining two common AI models with a classifier based on decision trees to estimate the flow velocity and flow depth in a 90° sharp curved channel. Their results represent the superior performance and accuracy of hybrid models over classic AI models alone. However, because the diversion angle in intake and bend channels has a considerable influence on the flow variables, the flow patterns differ in 60° and 90° bend channels. Hence, the present paper evaluates the goodness-of-fit results of the classifier multilayer perceptron (CMLP) and classifier radial basis function (CRBF) models designed based on decision trees. These models were designed to estimate the velocity and flow depth patterns in a 60° sharp curved channel based on 780 observational data for six flow rates: 5, 7.8, 13.6, 19.1, 25.3, and 30.6 l/s. According to the mean absolute relative error (MARE) and determination coefficient (R2) results of 0.0514 and 0.6 for velocity and 0.005 and 0.99 for flow depth prediction, the proposed CMLP model is more accurate than the classic MLP model. The CRBF model performed similar to CMLP, whereby CRBF predicted both velocity (MARE 0.086 and R2 = 0.9) and flow depth (MARE 0.0133 and R2 = 0.9) more accurately than the RBF model alone. Overall, the CMLP and CRBF models exhibited MARE reductions of 3% and 20% in velocity prediction and 36% and 22% in flow depth prediction compared to the individual MLP and RBF models, respectively. Moreover, CMLP and CRBF produce robust relationships based on different categories established according to various hydraulic conditions. The uncertainty analysis results show that, among the models examined, CMLP had the lowest uncertainty with the narrowest width of confidence bounds (WCB) of ± 0.1385 and ± 0.0031 in predicting velocity and flow depth, respectively. Moreover, CMLP and CRBF exhibited the greatest reliability with the lowest uncertainty indices in estimating the two flow variables in curved channels compared to the classic MLP and RBF models. Therefore, the combined AI-based classification models proposed in this study can serve as alternatives to the classic MLP and RBF models in the design and construction of curved channels with 60° and 90° bends.