Deep learning for anatomical interpretation of video bronchoscopy images

AbstractAnesthesiologists commonly use video bronchoscopy to facilitate intubation or confirm the location of the endotracheal tube; however, depth and orientation in the bronchial tree can often be confused because anesthesiologists cannot trace the airway from the oropharynx when it is performed using an endotracheal tube. Moreover, the decubitus position is often used in certain surgeries. Although it occurs rarely, the misinterpretation of tube location can cause accidental extubation or endobronchial intubation, which can lead to hyperinflation. Thus, video bronchoscopy with a decision supporting system using artificial intelligence would be useful in the anesthesiologic process. In this study, we aimed to develop an artificial intelligence model robust to rotation and covering using video bronchoscopy images. We collected video bronchoscopic images from an institutional database. Collected images were automatically labeled by an optical character recognition engine as the carina and left/right main bronchus. Except 180 images for the evaluation dataset, 80% were randomly allocated to the training dataset. The remaining images were assigned to the validation and test datasets in a 7:3 ratio. Random image rotation and circular cropping were applied. Ten kinds of pretrained models with < 25 million parameters were trained on the training and validation datasets. The model showing the best prediction accuracy for the test dataset was selected as the final model. Six human experts reviewed the evaluation dataset for the inference of anatomical locations to compare its performance with that of the final model. In the experiments, 8688 images were prepared and assigned to the evaluation (180), training (6806), validation (1191), and test (511) datasets. The EfficientNetB1 model showed the highest accuracy (0.86) and was selected as the final model. For the evaluation dataset, the final model showed better performance (accuracy, 0.84) than almost all human experts (0.38, 0.44, 0.51, 0.68, and 0.63), and only the most-experienced pulmonologist showed performance comparable (0.82) with that of the final model. The performance of human experts was generally proportional to their experiences. The performance difference between anesthesiologists and pulmonologists was marked in discrimination of the right main bronchus. Using bronchoscopic images, our model could distinguish anatomical locations among the carina and both main bronchi under random rotation and covering. The performance was comparable with that of the most-experienced human expert. This model can be a basis for designing a clinical decision support system with video bronchoscopy.

Download Full-text

Artificial Intelligence in Metrology Data Collection

10.51843/wsproceedings.2021.03 ◽

2021 ◽

Author(s):

Michael Schwartz ◽

Keyword(s):

Artificial Intelligence ◽

Data Collection ◽

Character Recognition ◽

Optical Character Recognition ◽

Lessons Learned ◽

Continuous Learning ◽

Automate Data ◽

Optical Character ◽

Automate Data Collection ◽

Continual Learning

Many companies have tried to automate data collection for handheld Digital Multimeters (DMM) using Optical Character Recognition (OCR). Only recently have companies tried to perform this task using Artificial Intelligence (AI) technology, Cal Lab Solutions being one of them in 2020. But when we developed our first prototype application, we discovered the difficulties of getting a good value with every measurement and test point.A year later, lessons learned and equipped with better software, this paper is a continuation of that AI project. In Beta-,1 we learned the difficulties of AI reading segmented displays. There are no pre-trained models for this type of display, so we needed to train a model. This required the testing of thousands of images, so we changed the scope of the project to a continual learning AI project. This paper will cover how we built our continuous learning AI model to show how any lab with a webcam can start automating those handheld DMMS with software that gets smarter over time.

Download Full-text

Improve OCR Accuracy with Advanced Image Preprocessing using Machine Learning with Python

International Journal of Innovative Technology and Exploring Engineering - Special Issue ◽

10.35940/ijitee.g5745.059720 ◽

2020 ◽

Vol 9 (7) ◽

pp. 1026-1030

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Neural Networks ◽

Character Recognition ◽

Optical Character Recognition ◽

Image Preprocessing ◽

Optical Character ◽

Handwritten Text ◽

Printed Text ◽

Learning Machine

Optical Character Recognition or Optical Character Reader (OCR) is a pattern-based method consciousness that transforms the concept of electronic conversion of images of handwritten text or printed text in a text compiled. Equipment or tools used for that purpose are cameras and apartment scanners. Handwritten text is scanned using a scanner. The image of the scrutinized document is processed using the program. Identification of manuscripts is difficult compared to other western language texts. In our proposed work we will accept the challenge of identifying letters and letters and working to achieve the same. Image Preprocessing techniques can effectively improve the accuracy of an OCR engine. The goal is to design and implement a machine with a learning machine and Python that is best to work with more accurate than OCR's pre-built machines with unique technologies such as MatLab, Artificial Intelligence, Neural networks, etc.

Download Full-text

Robotic Automation Process – The operating system for the digital enterprise

Proceedings of the International Conference on Business Excellence ◽

10.2478/picbe-2018-0007 ◽

2018 ◽

Vol 12 (1) ◽

pp. 54-69 ◽

Cited By ~ 5

Author(s):

Sorin Anagnoste

Keyword(s):

Artificial Intelligence ◽

Character Recognition ◽

Optical Character Recognition ◽

Big Data Analytics ◽

Process Automation ◽

Cognitive Services ◽

Accounts Payable ◽

Ibm Watson ◽

Digital Enterprise ◽

Robotic Automation

Abstract Robotic Process Automation (RPA) is going into a “maturity market”. The main vendor providers surpassed USD 1 billion in evaluation and the research they are launching these days on the market will change again radically the business landscape. It can be seen already what is coming next to RPA: intelligent optical character recognition (IOCR), chat-bots, machine learning, big data analytics, cognitive platforms, anomaly detection, pattern analysis, voice recognition, data classification and many more. As a result the top vendors developed partnerships with the main leading artificial intelligence providers, such as: IBM Watson, Microsoft Artificial Intelligence, Microsoft Cognitive services, blockchain, Google etc. On the business part, the consulting companies who are implementing the RPA solution are moving from developing Proof-of-Concepts (POCs) and Pilots to helping clients with RAP global roll-outs and developing Centre of Excellences (CoE). As a result, the experiences gathered so far by the author on this kind of projects will be tackled also in this paper. In this article will we will present also some data related to automation for different business areas (eg. Accounts Payable, Accounts Receivable etc) and how an assessment can be done correctly in order to decide if a process can be automatized and, if yes, up to which extent (ie. percent). Moreover, through the case studies we will provide (1) how now the RPA is integrated with Artificial Intelligence and Cloud, (2) how can be scaled in order to face hypes, (3) how can interpret data and (4) what savings these technologies can bring to the organizations. All the aforementioned services made Robotics Process Automation a very powerful tool since a year ago when the author did the last research. A process that was mainly not recommended for automation or was partially automated can be now fully automated with more advantages, such as: money, non-FTE savings and fulfillment time.

Download Full-text

Optical character recognition (OCR) using partial least square (PLS) based feature reduction: an application to artificial intelligence for biometric identification

Journal of Enterprise Information Management ◽

10.1108/jeim-02-2020-0076 ◽

2020 ◽

Vol ahead-of-print (ahead-of-print) ◽

Author(s):

Zainab Akhtar ◽

Jong Weon Lee ◽

Muhammad Attique Khan ◽

Muhammad Sharif ◽

Sajid Ali Khan ◽

...

Keyword(s):

Artificial Intelligence ◽

Character Recognition ◽

Optical Character Recognition ◽

Feature Reduction ◽

Partial Least Square ◽

Least Square ◽

License Plate ◽

Content Type ◽

Optical Character ◽

Machine Readable

PurposeIn artificial intelligence, the optical character recognition (OCR) is an active research area based on famous applications such as automation and transformation of printed documents into machine-readable text document. The major purpose of OCR in academia and banks is to achieve a significant performance to save storage space.Design/methodology/approachA novel technique is proposed for automated OCR based on multi-properties features fusion and selection. The features are fused using serially formulation and output passed to partial least square (PLS) based selection method. The selection is done based on the entropy fitness function. The final features are classified by an ensemble classifier.FindingsThe presented method was extensively tested on two datasets such as the authors proposed and Chars74k benchmark and achieved an accuracy of 91.2 and 99.9%. Comparing the results with existing techniques, it is found that the proposed method gives improved performance.Originality/valueThe technique presented in this work will help for license plate recognition and text conversion from a printed document to machine-readable.

Download Full-text

The impact of optical character recognition artificial intelligence on the labour market

International Journal of Engineering and Management Sciences ◽

10.21791/ijems.2021.4.2. ◽

2021 ◽

Vol 6 (4) ◽

Author(s):

Zsófia Riczu ◽

Zsolt Krutilla

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Information Technology ◽

Labour Market ◽

Character Recognition ◽

Optical Character Recognition ◽

Employment Policy ◽

Labour Law ◽

Employee Competencies ◽

The Impact

Because of present day information technology, there is neither need to plant complicated computers for more millions price if we would like to process and store big amounts of data, nor modelling them. The microprocessors and CPUs produced nowadays by that kind of technology and calculating capacity could not have been imagined 10 years before. We can store, process and display more and more data. In addition to this level of data processing capacity, programs and applications using machine learning are also gaining ground. During machine learning, biologically inspired simulations are performed by using artificial neural networks to able to solve any kind of problems that can be solved by computers. The development of information technology is causing rapid and radical changes in technology, which require not only the digital adaptation of users, but also the adaptation of certain employment policy and labour market solutions. Artificial intelligence can fundamentally question individual labour law relations: in addition to reducing the living workforce, it forces new employee competencies. This is also indicated by the Supiot report published in 1998, the basic assumption of which was that the social and economic regulatory model on which labour law is based is in crisis.

Download Full-text

Segmentation of Handwritten Text Document Written in Devanagri Script for Simple character, skewed character and broken character

INTERNATIONAL JOURNAL OF COMPUTERS & TECHNOLOGY ◽

10.24297/ijct.v8i1.3427 ◽

2013 ◽

Vol 8 (1) ◽

pp. 686-691

Author(s):

Vneeta Rani ◽

Dr.Vijay Laxmi

Keyword(s):

Artificial Intelligence ◽

Character Recognition ◽

Optical Character Recognition ◽

Character Segmentation ◽

Research Areas ◽

Text Document ◽

Optical Character ◽

Handwritten Text ◽

Recognition Phase ◽

Simple Character

OCR (optical character recognition) is a technology that is commonly used for recognizing patterns artificial intelligence & computer machine. With the help of OCR we can convert scanned document into editable documents which can be further used in various research areas. In this paper, we are presenting a character segmentation technique that can segment simple characters, skewed characters as well as broken characters. Character segmentation is very important phase in any OCR process because output of this phase will be served as input to various other phase like character recognition phase etc. If there is some problem in character segmentation phase then recognition of the corresponding character is very difficult or nearly impossible.

Download Full-text

PENDETEKSIAN PLAT NOMOR KENDARAAN MENGGUNAKAN ALGORITMA YOU ONLY LOOK ONCE V3 DAN TESSERACT

Jurnal Ilmiah Teknologi Infomasi Terapan ◽

10.33197/jitter.vol8.iss1.2021.718 ◽

2021 ◽

Vol 8 (1) ◽

pp. 57-62

Author(s):

Muhamad Rizky Fauzan ◽

Ari Purno Wahyu Wibowo

Keyword(s):

Artificial Intelligence ◽

Character Recognition ◽

Optical Character Recognition ◽

Optical Character

Perkembangan teknologi saat ini sangat berkembang pesat. Teknologi yang saat ini sedang dilakukan pengembangan secara besar-besaran yaitu Artificial Intelligence. Artificial Intelligence atau AI memiliki berbagai macam fungsi dan tujuan tergantung dari sistem yang akan dibuat. Salah satunya yaitu pendekteksian objek dan teks dari gambar atau video. Contoh dari pemanfaatan teknologi ini yaitu pada pendeteksian objek dan teks pada plat nomor kendaraan. Pada penelitian ini dilakukan perancangan sistem dengan menggunakan algoritma You Only Look Once V3 sebagai algoritma pendeteksi objek dan Tesseract Optical Character Recognition sebagai pendeteksi teks dalam gambar. Perancangan ini akan dibantu dengan library OpenCV pada bahasa pemrogramanan python dan menggunakan dataset gambar yang sudah tersedia. Penelitian ini bertujuan untuk mengetahui tingkat keakurasian algoritma You Only Look Once V3 yang dikombinasikan dengan Tesseract Optical Character Recognition.

Download Full-text

Abstract 141: Artificial Intelligence to Improve the Detection and Triage of Cerebral Aneurysms

Stroke ◽

10.1161/str.51.suppl_1.141 ◽

2020 ◽

Vol 51 (Suppl_1) ◽

Author(s):

Vitor Mendes Pereira ◽

Yoni Donner ◽

Gil Levi ◽

Nicole Cancelliere ◽

Erez Wasserman ◽

...

Keyword(s):

Artificial Intelligence ◽

Data Augmentation ◽

Pearson Correlation ◽

Model Performance ◽

Cerebral Aneurysms ◽

Roc Curves ◽

Clinical Decision ◽

Training Dataset ◽

Initial Assessment ◽

Validation Dataset

Cerebral Aneurysms (CAs) may occur in 5-10% of the population. They can be often missed because they require a very methodological diagnostic approach. We developed an algorithm using artificial intelligence to assist and supervise and detect CAs. Methods: We developed an automated algorithm to detect CAs. The algorithm is based on 3D convolutional neural network modeled as a U-net. We included all saccular CAs from 2014 to 2016 from a single center. Normal and pathological datasets were prepared and annotated in 3D using an in-house developed platform. To assess the accuracy and to optimize the model, we assessed preliminary results using a validation dataset. After the algorithm was trained, a dataset was used to evaluate final IA detection and aneurysm measurements. The accuracy of the algorithm was derived using ROC curves and Pearson correlation tests. Results: We used 528 CTAs with 674 aneurysms at the following locations: ACA (3%), ACA/ACOM (26.1%), ICA/MCA (26.3%), MCA (29.4%), PCA/PCOM (2.3%), Basilar (6.6%), Vertebral (2.3%) and other (3.7%). Training datasets consisted of 189 CA scans. We plotted ROC curves and achieved an AUC of 0.85 for unruptured and 0.88 for ruptured CAs. We improved the model performance by increasing the training dataset employing various methods of data augmentation to leverage the data to its fullest. The final model tested was performed in 528 CTAs using 5-fold cross-validation and an additional set of 2400 normal CTAs. There was a significant improvement compared to the initial assessment, with an AUC of 0.93 for unruptured and 0.94 for ruptured. The algorithm detected larger aneurysms more accurately, reaching an AUC of 0.97 and a 91.5% specificity at 90% sensitivity for aneurysms larger than 7mm. Also, the algorithm accurately detected CAs in the following locations: basilar(AUC of 0.97) and MCA/ACOM (AUC of 0.94). The volume measurement (mm3) by the model compared to the annotated one achieved a Pearson correlation of 99.36. Conclusion: The Viz.ai aneurysm algorithm was able to detect and measure ruptured and unruptured CAs in consecutive CTAs. The model has demonstrated that a deep learning AI algorithm can achieve clinically useful levels of accuracy for clinical decision support.

Download Full-text