The challenge was to make predictions based on X-ray, MRI and CT images of patients; for example, the predictions could coronavirus patients based on chest X-ray images, and microscopic images of cells. The predictions or object detections then assisted decision-making about medical treatments.
One of the biggest challenges of working with medical images is that the number of case images is limited compared to the control images, and this leads to imbalanced datasets. To overcome this issue, we implemented measurements such as different class weights and data augmentation.
- Our team analyzed clinical images, all radiographs were initially screened for quality control and all low-quality or unreadable scans were removed. A set of 5,000 high-quality chest X-ray images were analyzed for the coronavirus project. A series of 10,000 microscopic images were analyzed for heart attack research.
- By developing a ConvNET algorithm for object recognition/tracking from image data, we used machine learning techniques to predict the nature of particles.
- We developed a deep neural network (U-net) to automate image segmentation.
- Our team used various Machine Learning (ML) techniques for different classification tasks such as SVM, random forest and logistic regression, and checked for better evaluation metrics.
- Costs were reduced by lowering the numbers of required lab tests/experiments; for example, the goal of the coronavirus project was to predict based on inexpensive chest X-rays and as a result, saved the cost of expensive tests such as CT and also reduced the cost of the Patients’ length of stay in ICUs by early diagnosis (On average this model could save approximately $2,000 and 7 days of ICU stay per patient).
- Clinicians and scientists were provided with more insights to assist them in making better decisions about medical treatments and underlying conditions.
- We made case-control study predictions based on the medical images with encouraging accuracy; for example, the prediction of serious coronavirus illness based on a chest X-ray without requiring more lab tests.
We use open source technology to leverage the full potential of your data. Predictive and prescriptive results are actioned using AI and Machine Learning (ML).
We deliver scientific results that drive clinical and translational research decisions. Our Bioinformatics team has extensive experience designing, optimizing, executing and analyzing pre-clinical and clinical research projects using next-generation sequencing technologies.
What can we help you with?
T: +1 425-624-7532
Alternatively, simply fill in this form and we’ll be in touch.