Malaria diagnosis remains a critical challenge in endemic regions, where precise identification of Plasmodium species in thick blood smears determines treatment and survival. Current deep learning systems classify detections without providing morphological evidence, limiting the ability of microscopists to audit them. To bridge this gap, SGMCE (Segment-Grounded Morphological Concept Explanation) emerges as a post-hoc explanation framework that requires no retraining or additional morphological annotations. This approach generates natural-language explanations based on thick-smear morphology, offering transparency at a level previously inaccessible.
SGMCE extracts mask-guided crop thumbnails for each detection, computes fourteen handcrafted computer-vision morphological features (shape, color, chromatin, hemozoin pigment) using adaptive within-mask thresholds, and queries GPT-4o with both visual evidence and computed measurements. All this is conditioned on a thick-smear-specific knowledge base compiled from World Health Organization bench aids. The output is a structured explanation identifying which morphological features support the detected species and why competing species are excluded. Validation through metrics such as Knowledge-Base Consistency (KBC), CV-Claim Faithfulness (CCF), Discriminativeness Score (DS), and LLM-as-Judge (LLMj) shows mean KBC of 0.91, DS of 0.99, and CCF of 0.97 across 737 detections from 139 images covering four species and white blood cells.
Beyond healthcare, SGMCE’s architecture has profound implications for explainable AI software development. At Q2BSTUDIO, a company specialized in custom software applications, we see such techniques as a cornerstone for building solutions that end users can understand and validate. The ability to generate post-hoc explanations without additional annotations drastically reduces deployment costs in real clinical environments, where trust is as important as accuracy.
From a technical perspective, SGMCE combines classical computer vision with state-of-the-art language models, an approach that can be replicated in fields such as industrial inspection, precision agriculture, or cybersecurity. For instance, in network anomaly detection, an AI agent could explain why a packet is malicious based on extracted traffic features, similar to how SGMCE uses morphological traits. For these solutions to be viable, they need robust infrastructure: here, cloud services from AWS/Azure provide the necessary scalability and security. Q2BSTUDIO offers AI services that integrate explainable models into cloud platforms, ensuring regulatory compliance and protection of sensitive data through advanced cybersecurity practices.
Moreover, the information generated by these explanations can feed business intelligence systems like Power BI, enabling medical and management teams to analyze trends, model performance, and diagnostic patterns through interactive dashboards. The combination of explainable AI, cloud, and BI transforms decision-making into an evidence-based process, not a black box. At Q2BSTUDIO we develop custom AI agents that adapt to specific workflows, connecting explainability with real action.
The SGMCE case illustrates how computational morphology and LLMs can bridge the gap between artificial intelligence and clinical practice. But the same principle applies to any domain where interpretability is a critical non-functional requirement. By outsourcing the development of these capabilities to a technology partner like Q2BSTUDIO, organizations can accelerate AI adoption without compromising transparency. Our team integrates agile methodologies, cloud expertise, and deep knowledge of sector regulations to deliver turnkey solutions.
In conclusion, SGMCE represents a significant step toward AI systems that not only predict but also explain. For companies looking to implement similar technologies, having a partner that masters both the algorithmic and infrastructure sides is key. At Q2BSTUDIO we offer exactly that: custom software applications that integrate explainable AI, cloud, and cybersecurity, ready to transform sectors as diverse as healthcare, logistics, and finance.
