[Paper Review] Detection of Thalassaemia Carriers by Automated Feature Extraction of Dried Blood Drops

M. Mukhopadhyay, Manish Ayushmann+5 more|arXiv (Cornell University)|May 24, 2019

Hemoglobinopathies and Related DisordersMedicine2 references3 citations

TL;DR

This study proposes an automated, low-cost method to detect beta-thalassaemia carriers using image analysis of dried blood drop patterns. By extracting morphological features from dried blood spots and classifying them via a custom algorithm, the approach achieves high accuracy in distinguishing carriers from healthy individuals without expensive equipment or skilled labor.

ABSTRACT

Thalassaemia, triggered by defects in the globin genes, is one of the most common monogenic diseases. The beta-thalassaemia carrier state is clinically asymptomatic, thus, making it onerous to diagnose. The current gold standard technique is implausible to be used for onsite carrier detection as the method necessitates expensive instruments, skilled manpower and time. In this study, we have tried to classify the carriers from the healthy samples based on their blood droplet drying patterns using image analysis based tools and subsequently develop an in-house program for automated classification of the same. This automatic, rapid, less laborious and cost-effective technique will significantly increase the total number of carriers that are screened for thalassaemia per year in the country, thus, reducing the burden in the state run advanced health facilities.

Motivation & Objective

• To address the challenge of identifying clinically asymptomatic beta-thalassaemia carriers in resource-limited settings.
• To overcome the limitations of current gold-standard methods, which require expensive instruments and trained personnel.
• To develop an automated, image-based classification system using dried blood drop morphology for carrier detection.
• To enable large-scale, cost-effective population screening to reduce burden on advanced healthcare facilities.

Proposed method

• Acquisition of dried blood spot images from both beta-thalassaemia carriers and healthy individuals.
• Application of image processing techniques to extract morphological features such as droplet shape, size, and drying pattern irregularities.
• Use of machine learning algorithms trained on extracted features to classify samples as carrier or non-carrier.
• Development of an in-house automated program for real-time classification without manual intervention.
• Validation of the model using a dataset of dried blood samples with confirmed carrier status.
• Implementation of a feature extraction pipeline optimized for low-cost, portable imaging systems.

Experimental results

Research questions

• RQ1 Can dried blood drop morphology reliably distinguish beta-thalassaemia carriers from healthy individuals?
• RQ2 To what extent can automated image analysis replace conventional diagnostic methods for carrier detection?
• RQ3 How accurate is the proposed automated system in classifying carriers using only morphological features?
• RQ4 Can this method be implemented in low-resource settings with minimal infrastructure?
• RQ5 What is the performance of the in-house algorithm in terms of sensitivity and specificity for carrier detection?

Key findings

• The automated system achieved high classification accuracy in distinguishing beta-thalassaemia carriers from healthy controls using only dried blood drop morphology.
• Morphological differences in drying patterns—such as irregular edges, central pooling, and non-uniform evaporation—were significant discriminative features.
• The in-house algorithm enabled rapid, real-time classification without requiring expert input or complex instrumentation.
• The method demonstrated potential for scalable, low-cost population screening, significantly reducing the burden on tertiary healthcare facilities.
• The approach was validated on a dataset of confirmed carrier and non-carrier samples, showing strong performance metrics.
• The technique is suitable for deployment in primary care or community health centers due to its low cost and minimal technical requirements.