Machine Learning

Liver cancer treatment, especially for metastatic cases, poses significant challenges in accurately targeting tumours while sparing healthy tissue. Radioembolisation with yttrium-90 (Y-90) microspheres is a promising technique, but precise imaging of microsphere distribution is crucial. This study utilises T-PEPT, a novel Positron Emission Particle Tracking (PEPT) algorithm that combines topological data analysis with machine learning to identify Y-90 microsphere clusters in a digital twin of a patient's liver.

We collected programming problems and their solutions from previous studies. After applying some pre-processing steps, we queried advanced LLMs, such as GPT4, with the collected problems to produce machine-generated codes, while the original solutions were labeled as human-written codes. Finally, the entire collected dataset was divided into training, validation, and test sets, ensuring that there is no overlap among these sets, meaning no solutions in two different sets that solve the same programming problem.

The study focused on two regions in Rupnagar district, India, with an area of 216 km² as shown in Fig. 1a, using satellite data from June to November 2023. The upper region predominantly features paddy and maize, while the lower region includes paddy and sugarcane. Satellite images were obtained from PlanetScope’s 130-satellite constellation, with a spatial resolution of 3 meter. A total of 32 images, captured between late May and mid-November 2023, were used, all with less than 15% cloud cover.

The Facial Expression Dataset (Sri Lankan) is a culturally specific dataset created to enhance the accuracy of emotion recognition models in Sri Lankan contexts. Existing datasets, often based on foreign samples, fail to account for cultural differences in facial expressions, affecting model performance. This dataset bridges that gap, using high-quality data sourced from over 100 video clips of professional Sri Lankan actors to ensure expressive and clear facial imagery.