-
An AI-enabled Bias-Free Respiratory Disease Diagnosis Model using Cough Audio: A Case Study for COVID-19
Authors:
Tabish Saeed,
Aneeqa Ijaz,
Ismail Sadiq,
Haneya N. Qureshi,
Ali Rizwan,
Ali Imran
Abstract:
Cough-based diagnosis for Respiratory Diseases (RDs) using Artificial Intelligence (AI) has attracted considerable attention, yet many existing studies overlook confounding variables in their predictive models. These variables can distort the relationship between cough recordings (input data) and RD status (output variable), leading to biased associations and unrealistic model performance. To addr…
▽ More
Cough-based diagnosis for Respiratory Diseases (RDs) using Artificial Intelligence (AI) has attracted considerable attention, yet many existing studies overlook confounding variables in their predictive models. These variables can distort the relationship between cough recordings (input data) and RD status (output variable), leading to biased associations and unrealistic model performance. To address this gap, we propose the Bias Free Network (RBFNet), an end to end solution that effectively mitigates the impact of confounders in the training data distribution. RBFNet ensures accurate and unbiased RD diagnosis features, emphasizing its relevance by incorporating a COVID19 dataset in this study. This approach aims to enhance the reliability of AI based RD diagnosis models by navigating the challenges posed by confounding variables. A hybrid of a Convolutional Neural Networks (CNN) and Long-Short Term Memory (LSTM) networks is proposed for the feature encoder module of RBFNet. An additional bias predictor is incorporated in the classification scheme to formulate a conditional Generative Adversarial Network (cGAN) which helps in decorrelating the impact of confounding variables from RD prediction. The merit of RBFNet is demonstrated by comparing classification performance with State of The Art (SoTA) Deep Learning (DL) model (CNN LSTM) after training on different unbalanced COVID-19 data sets, created by using a large scale proprietary cough data set. RBF-Net proved its robustness against extremely biased training scenarios by achieving test set accuracies of 84.1%, 84.6%, and 80.5% for the following confounding variables gender, age, and smoking status, respectively. RBF-Net outperforms the CNN-LSTM model test set accuracies by 5.5%, 7.7%, and 8.2%, respectively
△ Less
Submitted 4 January, 2024;
originally announced January 2024.
-
Mythological Medical Machine Learning: Boosting the Performance of a Deep Learning Medical Data Classifier Using Realistic Physiological Models
Authors:
Ismail Sadiq,
Erick A. Perez-Alday,
Amit J. Shah,
Ali Bahrami Rad,
Reza Sameni,
Gari D. Clifford
Abstract:
Objective: To determine if a realistic, but computationally efficient model of the electrocardiogram can be used to pre-train a deep neural network (DNN) with a wide range of morphologies and abnormalities specific to a given condition - T-wave Alternans (TWA) as a result of Post-Traumatic Stress Disorder, or PTSD - and significantly boost performance on a small database of rare individuals.
App…
▽ More
Objective: To determine if a realistic, but computationally efficient model of the electrocardiogram can be used to pre-train a deep neural network (DNN) with a wide range of morphologies and abnormalities specific to a given condition - T-wave Alternans (TWA) as a result of Post-Traumatic Stress Disorder, or PTSD - and significantly boost performance on a small database of rare individuals.
Approach: Using a previously validated artificial ECG model, we generated 180,000 artificial ECGs with or without significant TWA, with varying heart rate, breathing rate, TWA amplitude, and ECG morphology. A DNN, trained on over 70,000 patients to classify 25 different rhythms, was modified the output layer to a binary class (TWA or no-TWA, or equivalently, PTSD or no-PTSD), and transfer learning was performed on the artificial ECG. In a final transfer learning step, the DNN was trained and cross-validated on ECG from 12 PTSD and 24 controls for all combinations of using the three databases.
Main results: The best performing approach (AUROC = 0.77, Accuracy = 0.72, F1-score = 0.64) was found by performing both transfer learning steps, using the pre-trained arrhythmia DNN, the artificial data and the real PTSD-related ECG data. Removing the artificial data from training led to the largest drop in performance. Removing the arrhythmia data from training provided a modest, but significant, drop in performance. The final model showed no significant drop in performance on the artificial data, indicating no overfitting.
Significance: In healthcare, it is common to only have a small collection of high-quality data and labels, or a larger database with much lower quality (and less relevant) labels. The paradigm presented here, involving model-based performance boosting, provides a solution through transfer learning on a large realistic artificial database, and a partially relevant real database.
△ Less
Submitted 28 December, 2021;
originally announced December 2021.
-
Systematic spatial and stoichiometric screening towards understanding the surface of ultrasmall oxygenated silicon nanocrystal
Authors:
Shanawer Niaz,
Aristides D. Zdetsis,
Emmanuel N. Koukaras,
Oguz Gulseren,
Imran Sadiq
Abstract:
In most of the realistic ab initio and model calculations which have appeared on the emission of light from Si nanocrystals, the role of surface oxygen has been usually ignored, underestimated or completely ruled out. We investigate theoretically, by density functional theory (DFT/B3LYP) possible modes of oxygen bonding in hydrogen terminated silicon quantum dots using as a representative case of…
▽ More
In most of the realistic ab initio and model calculations which have appeared on the emission of light from Si nanocrystals, the role of surface oxygen has been usually ignored, underestimated or completely ruled out. We investigate theoretically, by density functional theory (DFT/B3LYP) possible modes of oxygen bonding in hydrogen terminated silicon quantum dots using as a representative case of the Si29 nanocrystal. We have considered Bridge-bonded oxygen (BBO), Doubly-bonded oxygen (DBO), hydroxyl (OH) and Mix of these oxidizing agents. Due to stoichiometry, all comparisons performed are unbiased with respect to composition whereas spatial distribution of oxygen species pointed out drastic change in electronic and cohesive characteristics of nanocrytals. From an overall perspective of this study, it is shown that bridge bonded oxygenated nanocrystals accompanied by Mix have higher binding energies and large electronic gap compared to nanocrystals with doubly bonded oxygen atoms. In addition, it is observed that the presence of OH along with BBO, DBO and mixed configurations further lowers electronic gaps and binding energies and trends. It is also demonstrated that oxidizing constituent besides their spatial distribution significantly alters binding energy and highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) gap (HOMO-LUMO gap up to 1.48 eV) within same composition.
△ Less
Submitted 12 April, 2016;
originally announced April 2016.
-
Optical Properties of Pure and Mixed Germanium and Silicon Quantum Dots
Authors:
Shanawer Niaz,
Aristides D. Zdetsis,
Manzoor Ahmad Badar,
Safdar Hussain,
Imran Sadiq,
Muhammad Aslam Khan
Abstract:
We study the optical properties of hydrogen passivated silicon, germanium and mixed Ge/Si core/shell quantum dots (QDs) using high accuracy Density Functional Theory (DFT) and time-dependent DFT (TDFT). We employ the hybrid DFT functional of Becke, Lee, Yang and Parr (B3LYP) in combination with good quality basis sets. As we have shown in our previous work, this combination is an accurate and comp…
▽ More
We study the optical properties of hydrogen passivated silicon, germanium and mixed Ge/Si core/shell quantum dots (QDs) using high accuracy Density Functional Theory (DFT) and time-dependent DFT (TDFT). We employ the hybrid DFT functional of Becke, Lee, Yang and Parr (B3LYP) in combination with good quality basis sets. As we have shown in our previous work, this combination is an accurate and computationally efficient way for such calculations. The mixed quantum dots, as would be expected, are more versatile and offer more possibilities for band gap engineering, with gap values (electronic and optical) between those of the corresponding Si and Ge dots. Our results support the quantum confinement theory for all three types of QDs.
△ Less
Submitted 17 February, 2016;
originally announced February 2016.
