Application of deep learning for medical imaging evaluation
Inventors
Putha, Preetham • Tadepalli, Manoj • Reddy, Bhargava • Nimmada, Tarun • Rao, Pooja • Warier, Prashant
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-10733727-B2
Publication Date
2020-08-04
Expiration Date
Abstract
This disclosure generally pertains to methods and systems for processing electronic data obtained from imaging or other diagnostic and evaluative medical procedures. Certain embodiments relate to methods for the development of deep learning algorithms that perform machine recognition of specific features and conditions in imaging and other medical data. Another embodiment provides systems configured to detect and localize medical abnormalities on medical imaging scans by a deep learning algorithm.
Core Innovation
The invention streamlines tuberculosis detection by automating an X-ray screening and prioritizing process that receives and processes images from chest X-ray scan imaging procedure data. A deep learning system detects and localizes medical abnormalities in the images and outputs recognition scores corresponding to a level of recognition for each medical abnormality. For each detected medical abnormality, the system generates a bounding box representing the precise location and extent of the medical abnormalities.
The deep learning system is developed by selecting medical imaging scans and extracting medical abnormalities using natural language processing (NLP) algorithms. The medical abnormalities comprise blunted costophrenic angle, calcification, cardiomegaly, cavity, cervical rib, consolidation, hyper inflation, fibrosis, prominence in hilar region, opacity, pleural effusion, and scoliosis. The selected medical imaging scans are pre-processed by resizing and tag-specific data augmentations, and a deep learning algorithm is trained with the selected medical imaging scans.
Predictions of the presence or absence of particular medical abnormalities are produced by combining predictions of multiple classification models using different initial conditions, architectures and sampling via a majority ensembling scheme. The multiple classification models are pre-trained on a task of separating chest X-rays from X-rays of other body parts. Based on the medical abnormality scores that suggest typical or atypical tuberculosis, the system generates another tuberculosis recognition score and outputs advice corresponding to the level of recognition of tuberculosis.
Claims Coverage
The independent claim set covers a method and a system that automate an X-ray screening and prioritizing process for tuberculosis detection, using a deep learning system to detect and localize specified medical abnormalities, output per-abnormality recognition scores with bounding boxes, and generate a tuberculosis recognition score and advice based on typical or atypical tuberculosis indicators. The claim coverage includes five inventive features across the independent claims family.
Automated X-ray screening for tuberculosis detection
Receiving and processing images from chest X-ray scan imaging procedure data; detecting and localizing medical abnormalities using a deep learning system; generating a score corresponding to a level of recognition of each medical abnormality and outputting a bounding box representing the precise location and extent of the medical abnormalities; generating another score and advice corresponding to a level of recognition of tuberculosis based on the scores of the medical abnormalities that suggest typical or atypical tuberculosis.
NLP-extracted abnormality set for training and prediction
Selecting medical imaging scans and extracting the medical abnormalities using natural language processing (NLP) algorithms, wherein the medical abnormalities comprise blunted costophrenic angle, calcification, cardiomegaly, cavity, cervical rib, consolidation, hyper inflation, fibrosis, prominence in hilar region, opacity, pleural effusion, and scoliosis.
Image pre-processing via resizing and tag-specific augmentations
Pre-processing the selected medical imaging scans by resizing and tag-specific data augmentations.
Majority-ensembled classification models pre-trained to separate chest X-rays from other body parts
Predicting the presence or absence of a particular type of medical abnormalities by combining the predictions of multiple classification models using different initial conditions, architectures and sampling via a majority ensembling scheme, wherein the multiple classification models are pre-trained on a task of separating chest X-rays from X-rays of other body parts.
Tuberculosis recognition score and advice derived from abnormality recognition
Generating another score and advice that corresponds to a level of recognition of tuberculosis based on the scores of the medical abnormalities that suggest typical or atypical tuberculosis.
Across the independent claims, the core coverage centers on automating tuberculosis detection from chest X-ray imaging by NLP-extracting a defined set of medical abnormalities, training a deep learning system to detect and localize those abnormalities, and combining multiple classification models via majority ensembling with pre-training to separate chest X-rays from other-body-part X-rays. The system then produces per-abnormality recognition scores with bounding boxes and derives a tuberculosis recognition score and advice from abnormalities suggesting typical or atypical tuberculosis.
Stated Advantages
Streamlining tuberculosis detection by automating an X-ray screening and prioritizing process.
Documented Applications
A tuberculosis screening embodiment that provides a tuberculosis screening output with probability scores and advice, based on abnormalities suggesting typical or atypical tuberculosis, and integrates with workflow elements including PACS/VNA and HL7.
Radiologist validation of automated deep learning detection and localization of chest X-ray abnormalities, using a majority gold standard of 3 radiologists and an external dataset (Qure90K).
Interested in licensing this patent?