BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Cambridge MedAI Seminar Series  - Jack Dixon\, Gift Mungmeeprued\,
  Bevis Drury
DTSTART:20240611T110000Z
DTEND:20240611T120000Z
UID:TALK217432@talks.cam.ac.uk
CONTACT:Ines Machado
DESCRIPTION:The next seminar will be held on *11 June 2024\, 12-1pm at the
  Jeffrey Cheah Biomedical Centre (Main Lecture Theatre)\, University of Ca
 mbridge* and streamed online via Zoom. A light *lunch from Aromi will be s
 erved from 11:50*. This month will feature the following talks:\n\n*How Lo
 w Can We Go? - Investigating the interaction between cancer-detecting AI a
 nd low-dose quantum noise in CT images - Jack Dixon\, Master’s student\,
  Department of Physics\, University of Cambridge*\n\nJack is an undergradu
 ate student at the University of Cambridge currently studying for a Master
 's Degree in Natural Sciences. He specialises in physics\, and is particul
 arly interested in statistical and computational physics. As part of his d
 egree\, he undertook a research project within the Early Cancer Institute 
 under the supervision of William McGough\, Dr Mireia Crispin-Ortuzar and D
 r Ander Biguri\, focused on low-dose CT scan simulation and deep-learning 
 based segmentation.\n\nAbstract: Renal cancers (RC) are associated with mo
 re than 140\,000 deaths annually. Mortality rates for RC could be reduced 
 if a suitable screening program to allow for early diagnosis was construct
 ed. Trials into screening\, such as the Yorkshire Kidney Screening Trial\,
  use non-contrast enhanced CT scans and ideally seek to lower the dose of 
 ionising radiation as much as is feasible. In order for a screening progra
 m to be effective\, it must be both cost-effective and (relatively) safe. 
 To this end\, my Master's project focused on assessing the performance of 
 automatic renal segmentation models as the incident radiation dose of the 
 input CT scans is decreased. This involved first attempting to construct a
 nd validate a low-dose CT scan simulation technique that can be applied re
 troactively\, and then assessing segmentation performance as the dose is d
 ecreased. The renal and cancer segmentation models produced both displayed
  strong positive ranked correlation between Dice similarity coefficient an
 d incident dose to a significance level of 2.5%. We conclude that renal se
 gmentation performance in non-contrast enhanced CT scans is correlated wit
 h the incident dose.\n\n\n*Deformable registration to assess tumour progre
 ssion in ovarian cancer patients - Gift Mungmeeprued - Master's student\, 
 Department of Physics\, University of Cambridge*\n\nGift Mungmeeprued is a
  master's student in the Department of Physics at the University of Cambri
 dge. She is interested in machine learning to make healthcare more accessi
 ble and affordable.\n\nAbstract: High-grade serous ovarian carcinoma (HGSO
 C) is the most common and deadliest subtype of ovarian cancer\, often char
 acterised by multi-site and heterogeneous tumours. The standard line of tr
 eatment for HGSOC in the UK is neoadjuvant chemotherapy (NACT) followed by
  delayed primary surgery. Response Evaluation Criteria in Solid Tumours (R
 ECIST 1.1) is the current standardised criteria to assess the tumour respo
 nse to NACT based on measurements of tumour diameters in pre- and post-NAC
 T CT scans. While RECIST is designed to be relatively quick for radiologis
 ts to evaluate\, it only captures 1-dimensional global change in tumour si
 ze. In this talk\, we explored the use of deformable image registration as
  an automated tool to assess tumour response to NACT. Registration between
  pre- and post-NACT CT scans reveals spatial heterogeneity of changes with
 in the tumour and across multiple disease sites.\n\n\n*Automating Segmenta
 tion and Chemotherapy Response Measurement in Ovarian Cancer with Multitas
 k Deep Learning - Bevis Drury\, Master’s student\, Department of Physics
 \, University of Cambridge*\n\nBevis is a Part III student studying physic
 s at the University of Cambridge. He is interested in applying machine lea
 rning to all areas of research\, from physics to medicine.\n\nAbstract: Hi
 gh Grade Serous Ovarian Cancer (HGSOC) is the most common type of ovarian 
 cancer. Often diagnosed at advanced stages\, HGSOC presents significant ch
 allenges due to its heterogeneity and metastatic nature. Treatment of HGSO
 C begins with either immediate primary surgery\, or neoadjuvant chemothera
 py prior to delayed primary surgery. To track disease progression\, radiol
 ogists routinely use abdominopelvic CT imaging. The patient’s radiologic
 al response to treatment can be measured using the Response Evaluation Cri
 teria in Solid Tumours (RECIST)\, which compares CT scans taken before and
  after treatment. Manual calculation of RECIST is time-consuming and often
  inconsistent between radiologists\, impacting the accuracy and reliabilit
 y of treatment assessments. This paper develops a multitask deep learning 
 architecture for automating the segmentation and chemotherapy response pre
 diction of HGSOC patients. The model combines features from two identical 
 U-Net architectures\, which are then used to predict binarised RECIST labe
 ls. We use a training cohort of 99 HGSOC cases with pre- and post-treatmen
 t CT scans\, and an external validation cohort of 49 cases. For the valida
 tion cohort\, we predict binarised RECIST labels with an AUC of 0.78. We a
 re the first to predict RECIST labels for HGSOC patients using multitask d
 eep learning\, establishing this research as a benchmark for future work. 
 RECIST measurements are not currently used in clinical practice\, so this 
 framework aims to provide radiologists with real-time segmentations and RE
 CIST labels leading to more informed decisions.
LOCATION:Jeffrey Cheah Biomedical Centre (Main Lecture Theatre)\, Universi
 ty of Cambridge
END:VEVENT
END:VCALENDAR