ULTRASOUND FETUS BRAIN

PROBLEM STATEMENT

The evaluation of fetal brain can be critical because deficits in the perfusion (passage of blood flow) of this territory may lead to inadequate development of the central nervous system and even jeopardize fetal vitality. Despite many significant efforts and promising outcomes in this domain, accurate segmentation and classification remain a challenging task. The objective of this case study is to evaluate the maturity of fetal ultrasound brain in a real maternal-fetal clinical environment to automatically classify fetal anatomical planes of Brain. After a detailed analysis and research, Ai labs developed and implemented a Proof of Concept (POC) using its proprietary engine Minsky™ to automatically evaluate the maturity of fetal brain via ultrasound images and classify the data image views accordingly.

KEY BENEFITS (MINSKY™)

• User-Friendly cloud based AI platform
• Scalable across various domains/data.
• Provides you a list of % dependency features that can be used to optimize the outcomes.
• Ability to fine tune or optimize the models by trying different algorithms / prediction attributes
• Easy integration with other third party solutions such as TABLEAU for data visualization

RESULTS/BENEFITS

• Assisting clinicians in the early diagnosis of a variety of conditions and diseases from ultra-sound images.
• Reduce the rates of misdiagnosis and missed diagnosis.
• Patients can get more safer and effective treatment.
• Reduce time and cost involved in the process.
• Effectively improve the quality of medical services.

SOLUTION OVERVIEW:

After comprehensively evaluating the Client’s historical ultra-sound image data, we classified into three categorical views according to fetal anatomical plane.
A. Trans-Ventricular
B. Trans-Thalamic
C. Trans-Cerebellum
We analysed these images using Deep learning algorithms by classifying the views into the above mentioned three categories. Historical labelled image data was used for detailed modelling by Minsky™ for the selected deep learning Algorithms. Real time image data was then fed to the trained Minsky™ models to predict and classify the images as Trans-Ventricular or Trans-Thalamic or Trans-cerebellum views for early treatment if needed.

Minsky™ DL Results for TRANS-CEREBELLUM VIEW

Based on the data from Fig 1 above, our analysis shows that CNN model gave the best Model accuracy and classified the above ultra sound fetus brain image as Tran-Ventricular (99%) view with the best matching image.

Minsky™ DL Results for Trans-Thalamic View

Based on the data from Fig 2 above, our analysis shows that CNN gave highest Model accuracy and classified the above ultra sound fetus brain image as Trans-Thalamic view (98%) with the best matching image.

Minsky™ DL Results for TRANS-VENTRICULAR VIEW

Based on the data from above in Fig 3, our analysis shows that CNN model gave highest model accuracy and classified the above ultra sound fetus brain image as Tran-Cerebellum view (94%) with the best matching image.