Automated Interpretation of Blood Culture Gram Stains by Use of a Deep Convolutional Neural Network

Abstract

24Microscopic interpretation of stained smears is one of the most operator-dependent and time 25 intensive activities in the clinical microbiology laboratory. Here, we investigated application of 26 an automated image acquisition and convolutional neural network (CNN)-based approach for 27 automated Gram stain classification. Using an automated microscopy platform, uncoverslipped 28 slides were scanned with a 40x dry objective, generating images of sufficient resolution for 29 interpretation. We collected 25,488 images from positive blood culture Gram stains prepared 30 during routine clinical workup. These images were used to generate 100,213 crops containing 31Gram-positive cocci in clusters, Gram-positive cocci in chains/pairs, Gram-negative rods, or 32 background (no cells). These categories were targeted for proof-of-concept development as they 33

Machine learning applications in epilepsy

“…20 An alternate approach to overcoming overfitting is transfer learning, wherein a deep learning model previously trained for one task may be retrained using a far smaller dataset for a different task, for instance by retraining an image classification algorithm to classify Gram stain images. 21…”

Automated classification of osteomeatal complex inflammation on computed tomography using convolutional neural networks

“…One promising study demonstrated that the performance of a CNN in diagnosing benign and malignant skin lesions from photographs alone was equivalent and, in some cases, superior to a group of board-certified dermatologists 15 . Another was able to correctly classify bacteria seen on gram-stain slides with 94% accuracy 16 . Within otolaryngology, two pilot studies have used CNNs to detect cancer cells 17 and endolymphatic hydrops in images from mice 18 .…”

Morphologic Classification and Automatic Diagnosis of Bacterial Vaginosis by Deep Neural Networks

“…In the past years, CNN methods had shown lots of examples of successful applications in medical image processing. Typical examples were: automatically classification 26 common skin conditions [39]; using pixels and disease labels as inputs to classify skin lesions [40]; identified prostate cancer in biopsy specimens and detected breast cancer metastasis in sentinel lymph nodes [41]; automated classification of Gram staining of blood samples [33]; automatically diagnosis of H. pylori infection [32]; predicting cardiovascular risk factors from extracted retinal fundus images [31]. Therefore, the CNN methods was suitable for medical image processing.…”

“…Numerous CNN models, including LeNet-5 [20], AlexNet [21], VGGNet [22], ResNet [23], GooLeNet [24], Xception [25], FCN [26], PSPNet [27], and U-net [28], were developed to improve performance on natural image recognition. Many models were proved effective for medical image processing, from identifying diabetic retinopathy in retinal fundus photographs [29][30][31], endoscopic images [32], to microbiology recognitions [33,34]. We hypothesized that CNN based deep learning models can be used to diagnose BV using Nugent score classifications efficiently and accurately.…”