BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.3.1//EN TZID:Asia/Kolkata X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT UID:205@cds.iisc.ac.in DTSTART;TZID=Asia/Kolkata:20260612T110000 DTEND;TZID=Asia/Kolkata:20260612T120000 DTSTAMP:20260602T071858Z URL:https://cds.iisc.ac.in/events/m-tech-research-thesis-defense-cds-012-j une-2026-an-adaptive-learning-framework-for-brain-tissue-extraction-from-m ri-with-explainability/ SUMMARY:M.Tech Research: Thesis Defense: CDS: 012 June 2026 "An Adaptive Le arning Framework for Brain Tissue Extraction from MRI with Explainability" DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nM.Tech Research Thesis Defense\n\n\n\nSpeaker : Mr. Akhil\nS.R. Number : 06-18-01-10-22-23 -1-23206\nTitle : "An Adaptive Learning Framework for Brain Tissue Extract ion from MRI with Explainability"\nThesis examiner : Prof. Soma Biswas\, D epartment of Electrical Engineering\, IISc.\nResearch Supervisor : Prof. D ebnath Pal\nDate &\; Time : June 12\, 2026 (Friday) at 11:00 AM\nVenue : # 102 CDS Seminar Hall\n================================================ ==========================\nABSTRACT\nBrain extraction\, or skull-strippin g\, is a critical first step in nearly every neuroimaging pipeline. When t his step fails\, the errors propagate downward\, degrading the accuracy of volumetric and surface analyses. While neural networks have pushed perfor mance benchmarks higher\, they remain computationally expensive\, opaque\, and often struggle when applied to data from different scanners (domain s hift). This thesis investigates whether a carefully designed\, fully inter pretable classical learning framework can match the accuracy of deep learn ing while offering better explainability and lower deployment costs. We pr esent a registration-guided\, slice-aware framework for extracting the bra in from multi-contrast Magnetic resonance imaging (MRI) (T1/T2/FLAIR). The pipeline operates in three stages: (i) it uses registration to characteri ze patient brain texture and generate reliable training examples\; (ii) it employs a gradient boosted classifier to identify the high-confidence cor e brain tissue\; and (iii) it models boundary uncertainty using a gradient -boosted regressor that outputs continuous inclusion scores. The complete feature set\, including intensity\, rotationally invariant Local Binary Pa tterns (LBP)\, Gray-Level Co-occurrence Matrix (GLCM) Haralick features\, Gabor responses and wavelets. Empirical evaluation was performed on the Ne urofeedback Skull-stripped (NFBS) and Calgary-Campinas-359 (CC-359) datase ts\, as well as an institutional AIIMS clinical cohort. On NFBS\, the prop osed method achieved a mean Dice of 0.9521 ± 0.0069 and a mean HD95 of 5. 1871 ± 1.5544 mm. On CC359\, it attained a mean Dice of 0.9555 ± 0.0073 and a mean HD95 of 4.5633 ± 0.8910 mm. On the AIIMS clinical dataset\, it reached a mean Dice of 0.9434 ± 0.0107 and a mean HD95 of 6.4078 ± 0.87 50 mm. All values are reported as mean ± per-subject standard deviation. These results demonstrate that this feature-driven pipeline achieves overl ap and boundary fidelity comparable to contemporary Convolutional Neural N etwork (CNN) baselines. Crucially\, it does so without requiring large\, a nnotated training corpora and provides fully inspectable decision rules. Q ualitative analysis suggests fewer extreme failure cases in challenging sl ices and more anatomically plausible behavior when the model is uncertain. Current limitations include conservative under-segmentation in regions of complex cortical folding and reduced accuracy in cases of severe anatomic al deformation\, such as large lesions. Future directions aimed at address ing these limitations include proposals for multi-centre validation and pa thology-targeted extensions.\n\n\n\nALL ARE WELCOME CATEGORIES:Events,Thesis Defense END:VEVENT BEGIN:VTIMEZONE TZID:Asia/Kolkata X-LIC-LOCATION:Asia/Kolkata BEGIN:STANDARD DTSTART:20250612T110000 TZOFFSETFROM:+0530 TZOFFSETTO:+0530 TZNAME:IST END:STANDARD END:VTIMEZONE END:VCALENDAR