Augustin, A. I. | Search bioRxiv

bioRxiv2022

Turing Test Inspired Method for Analysis of Biases Prevalent in Artificial Intelligence-Based Medical Imaging

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSBackgroundC_ST_ABSBecause of the growing need to provide better global healthcare, computer-based and robotic healthcare equipment that depend on artificial intelligence have seen an increase in development. In order to evaluate artificial intelligence (AI) in computer technology, the Turing test was created. For evaluating the future generation of medical diagnostics and medical robots, it remains an essential qualitative instrument. MethodWe propose a novel methodology to assess AI-based healthcare technology that provided verifiable diagnostic accuracy and statistical robustness. In order to run our test, we used a State-of-the-art AI model and compared it against radiologist for checking how generalized of the model is and if any biases are prevalent. ResultsWe achieved results that can evaluate the performance of our chosen model for this study in a clinical setting and we also applied a quantifiable methods for evaluating our modified turing test results using a meta-analytical evaluation framework. ConclusionThis test provides a translational standard for upcoming AI modalities. Our modified Turing Test is a notably strong standard to measure the actual performance of the AI model on a variety of edge cases and normal cases and also helps in detecting if the algorithm is biased towards any one type of case. This method extends the flexibility detect any prevalent biases and also classify the type of bias.

bioinformatics↗

bioRxiv2022

RADGENNETS: DEEP LEARNING-BASED RADIOGENOMICS MODEL FOR GENE MUTATION PREDICTION IN LUNG CANCER

AO_SCPLOWBSTRACTC_SCPLOWIn this paper, we present our methodology that can be used for predicting gene mutation in patients with non-small cell lung cancer (NSCLC). There are three major types of gene mutations that a NSCLC patients gene structure can change to: epidermal growth factor receptor (EGFR), Kirsten rat sarcoma virus (KRAS), and Anaplastic lymphoma kinase (ALK). We worked with the clinical and genomics data for each patient as well CT scans. We preprocessed all of the data and then built a novel pipeline to integrate both the image and tabular data. We built a novel pipeline that used a fusion of Convolutional Neural Networks and Dense Neural Networks. Also, using a search approach, we pick an ensemble of deep learning models to classify the separate gene mutations. These models include EfficientNets, SENet, and ResNeXt WSL, among others. Our model achieved a high area under curve (AUC) score of 94% in detecting gene mutation.

cancer biology↗