- Computer Vision with Deep Learning
to Identify Risk Factors
- State-of-the-art sensitivity
- Detection sensitivity increased to 60% while cutting detection time by half
- Increased patient reach and fostered long-term relationships with lower capex
- Early detection
- 60% increase in early andaccurate detection in localize & metastasis cases
- Cost effective
- 15% reduction in costwhile accommodatingmore requests
- Let’s Get Started
Computer Vision with Deep Learning
to Identify Risk Factors
The healthcare client aspired to revive and enhance its diagnosis and screening approach and chose Sparity for developing a solution that would enable early and accurate diagnosis, thus advancing effective treatment.
The Healthcare client is highly renowned multi-specialty facility that offers tertiary and quaternary care services, highly specialized treatments along with exceptional and compassionate patient care. Driven by innovative clinical research, cutting-edge technologies, modular operation theatre, modernized monitoring systems and advanced treatments, including targeted therapies, genomic-based treatment and immunotherapy offer highly sophisticated diagnostic and treatment options to the patients who visit them. The healthcare client brings together a variety of wellness care, diagnostic, and treatment services, all provided within a single location.
The healthcare client was looking for a solution that could address their severe need for effective diagnosis of screening results where the naked human eye could miss the early signs of cancer. They needed a solution that could initially detect screening results and offer an accurate diagnosis. In response, Sparity developed a deep-learning AI model based on convolutional neural networks (CNNs) and deep learning algorithms that are capable of visual object detection, that detect patterns against large datasets of images from both healthy and cancerous tissue and further offers metastasis detection to facilitate a more accurate and efficient diagnosis, saving the patient time and speeding up the treatment.
The healthcare client was facing challenges with their current diagnosis and screening approach as trained radiologists can miss evidence in screening results —including high false-positive rates and difficulty predicting treatment success. Furthermore, the manual assessment is prone to subjectivity and sometimes error-prone. As early diagnosis and detection will benefit from potentially curative treatment. The healthcare client needed a solution that could analyze the collected structured data and screening test results in specifically identifying early metastasis in screening mammography, early diabetic retinopathy, colonoscopy, prostate-specific antigen, and cervical cytology.
Designed the framework
that can automatically
detect patterns and localize tumors
Trained the AI model
on a pre-sampled
set of image patches
Trained a random forest
classifier for identification
detection and localization.
Sparity’s healthcare product experts, based on initial requirements, developed a deep-learning AI model with convolutional neural network (CNN) architecture with visual object recognition and detection from whole-slide images from the screening results. The accuracy of various computer vision tasks such as image recognition, object detection, and semantic segmentation is greatly enhanced by the use of CNNs. The deep learning model used an image segmentation approach where images were broken down into multiple segments to make them easier to analyze. The algorithms are trained using a vast database of labelled or pre-identified images and videos of both healthy and cancerous tissue. Furthermore, the deep-learning algorithm searches exhaustively for abnormal patterns and identifies cancer patterns, dramatically reducing the uncertainty of a false negative or missed case of cancer.